diff --git "a/slurmm1.0-20174.out" "b/slurmm1.0-20174.out" new file mode 100644--- /dev/null +++ "b/slurmm1.0-20174.out" @@ -0,0 +1,3069 @@ +*************************************************** + You have loaded a buildenv module +*************************************************** +The buildenv-gcccuda/12.9.1-gcc11 module makes available: + - GCC: 11 + - CUDA: 12.9.1 + - OpenMPI: 5.0.8 + - FFTW: 3.3.10 + - OpenBLAS: 0.3.30 + - ScaLAPACK: 2.2.2 + - MAGMA: 2.9.0 + - Eigen: 5.0.0 + - OpenCV: 4.12.0 + - LLVM: 21.1.3 + +These libraries are accessible via the standard environment +variables CPATH (for headers) and LIBRARY_PATH (for libraries), +which are picked up automatically by the compiler toolchain. + +Reusing existing joint manifests under: data/mix/manifests/joint_openvid_fullmobile_5v1i +Training config: configs/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_v2_bs64_8gpu.yaml +Output root: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu +Checkpoint mode: init +Checkpoint path: output/stage1_bridge_fulldit_full_mobile_o_smolvlm2_500m_lexical_gated_k2_diffonly_init10k_bs64_v2_20260420_8gpu/20260425_135135/checkpoint_latest.pt +W0510 12:43:46.788000 3035434 site-packages/torch/distributed/run.py:792] +W0510 12:43:46.788000 3035434 site-packages/torch/distributed/run.py:792] ***************************************** +W0510 12:43:46.788000 3035434 site-packages/torch/distributed/run.py:792] Setting OMP_NUM_THREADS environment variable for each process to be 1 in default, to avoid your system being overloaded, please further tune the variable for optimal performance in your application as needed. +W0510 12:43:46.788000 3035434 site-packages/torch/distributed/run.py:792] ***************************************** +[2026-05-10 12:44:02,400] [INFO] [real_accelerator.py:239:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2026-05-10 12:44:02,602] [INFO] [real_accelerator.py:239:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2026-05-10 12:44:02,639] [INFO] [real_accelerator.py:239:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2026-05-10 12:44:02,681] [INFO] [real_accelerator.py:239:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2026-05-10 12:44:02,707] [INFO] [real_accelerator.py:239:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2026-05-10 12:44:02,723] [INFO] [real_accelerator.py:239:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2026-05-10 12:44:02,754] [INFO] [real_accelerator.py:239:get_accelerator] Setting ds_accelerator to cuda (auto detect) +[2026-05-10 12:44:02,772] [INFO] [real_accelerator.py:239:get_accelerator] Setting ds_accelerator to cuda (auto detect) +Warning: The cache directory for DeepSpeed Triton autotune, /proj/cvl/users/x_fahkh2/caches, appears to be on an NFS system. While this is generally acceptable, if you experience slowdowns or hanging when DeepSpeed exits, it is recommended to set the TRITON_CACHE_DIR environment variable to a non-NFS path. +Warning: The cache directory for DeepSpeed Triton autotune, /proj/cvl/users/x_fahkh2/caches, appears to be on an NFS system. While this is generally acceptable, if you experience slowdowns or hanging when DeepSpeed exits, it is recommended to set the TRITON_CACHE_DIR environment variable to a non-NFS path. +Warning: The cache directory for DeepSpeed Triton autotune, /proj/cvl/users/x_fahkh2/caches, appears to be on an NFS system. While this is generally acceptable, if you experience slowdowns or hanging when DeepSpeed exits, it is recommended to set the TRITON_CACHE_DIR environment variable to a non-NFS path. +Warning: The cache directory for DeepSpeed Triton autotune, /proj/cvl/users/x_fahkh2/caches, appears to be on an NFS system. While this is generally acceptable, if you experience slowdowns or hanging when DeepSpeed exits, it is recommended to set the TRITON_CACHE_DIR environment variable to a non-NFS path. +Warning: The cache directory for DeepSpeed Triton autotune, /proj/cvl/users/x_fahkh2/caches, appears to be on an NFS system. While this is generally acceptable, if you experience slowdowns or hanging when DeepSpeed exits, it is recommended to set the TRITON_CACHE_DIR environment variable to a non-NFS path. +Warning: The cache directory for DeepSpeed Triton autotune, /proj/cvl/users/x_fahkh2/caches, appears to be on an NFS system. While this is generally acceptable, if you experience slowdowns or hanging when DeepSpeed exits, it is recommended to set the TRITON_CACHE_DIR environment variable to a non-NFS path. +Warning: The cache directory for DeepSpeed Triton autotune, /proj/cvl/users/x_fahkh2/caches, appears to be on an NFS system. While this is generally acceptable, if you experience slowdowns or hanging when DeepSpeed exits, it is recommended to set the TRITON_CACHE_DIR environment variable to a non-NFS path. +Warning: The cache directory for DeepSpeed Triton autotune, /proj/cvl/users/x_fahkh2/caches, appears to be on an NFS system. While this is generally acceptable, if you experience slowdowns or hanging when DeepSpeed exits, it is recommended to set the TRITON_CACHE_DIR environment variable to a non-NFS path. +/proj/cvl/users/x_fahkh2/envs/mobileov/lib/python3.10/site-packages/timm/models/layers/__init__.py:48: FutureWarning: Importing from timm.models.layers is deprecated, please import via timm.layers + warnings.warn(f"Importing from {__name__} is deprecated, please import via timm.layers", FutureWarning) +/proj/cvl/users/x_fahkh2/envs/mobileov/lib/python3.10/site-packages/timm/models/layers/__init__.py:48: FutureWarning: Importing from timm.models.layers is deprecated, please import via timm.layers + warnings.warn(f"Importing from {__name__} is deprecated, please import via timm.layers", FutureWarning) +/proj/cvl/users/x_fahkh2/envs/mobileov/lib/python3.10/site-packages/timm/models/layers/__init__.py:48: FutureWarning: Importing from timm.models.layers is deprecated, please import via timm.layers + warnings.warn(f"Importing from {__name__} is deprecated, please import via timm.layers", FutureWarning) +/proj/cvl/users/x_fahkh2/envs/mobileov/lib/python3.10/site-packages/timm/models/layers/__init__.py:48: FutureWarning: Importing from timm.models.layers is deprecated, please import via timm.layers + warnings.warn(f"Importing from {__name__} is deprecated, please import via timm.layers", FutureWarning) +/proj/cvl/users/x_fahkh2/envs/mobileov/lib/python3.10/site-packages/timm/models/layers/__init__.py:48: FutureWarning: Importing from timm.models.layers is deprecated, please import via timm.layers + warnings.warn(f"Importing from {__name__} is deprecated, please import via timm.layers", FutureWarning) +/proj/cvl/users/x_fahkh2/envs/mobileov/lib/python3.10/site-packages/timm/models/layers/__init__.py:48: FutureWarning: Importing from timm.models.layers is deprecated, please import via timm.layers + warnings.warn(f"Importing from {__name__} is deprecated, please import via timm.layers", FutureWarning) +/proj/cvl/users/x_fahkh2/envs/mobileov/lib/python3.10/site-packages/timm/models/layers/__init__.py:48: FutureWarning: Importing from timm.models.layers is deprecated, please import via timm.layers + warnings.warn(f"Importing from {__name__} is deprecated, please import via timm.layers", FutureWarning) +/proj/cvl/users/x_fahkh2/envs/mobileov/lib/python3.10/site-packages/timm/models/layers/__init__.py:48: FutureWarning: Importing from timm.models.layers is deprecated, please import via timm.layers + warnings.warn(f"Importing from {__name__} is deprecated, please import via timm.layers", FutureWarning) +/proj/cvl/users/x_fahkh2/Mobile-OV_Alpha/nets/third_party/sana/diffusion/model/wan/model.py:124: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. + @amp.autocast(enabled=False) +/proj/cvl/users/x_fahkh2/Mobile-OV_Alpha/nets/third_party/sana/diffusion/model/wan/model.py:134: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. + @amp.autocast(enabled=False) +/proj/cvl/users/x_fahkh2/Mobile-OV_Alpha/nets/third_party/sana/diffusion/model/wan/model.py:124: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. + @amp.autocast(enabled=False) +/proj/cvl/users/x_fahkh2/Mobile-OV_Alpha/nets/third_party/sana/diffusion/model/wan/model.py:134: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. + @amp.autocast(enabled=False) +/proj/cvl/users/x_fahkh2/Mobile-OV_Alpha/nets/third_party/sana/diffusion/model/wan/model.py:124: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. + @amp.autocast(enabled=False) +/proj/cvl/users/x_fahkh2/Mobile-OV_Alpha/nets/third_party/sana/diffusion/model/wan/model.py:134: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. + @amp.autocast(enabled=False) +/proj/cvl/users/x_fahkh2/Mobile-OV_Alpha/nets/third_party/sana/diffusion/model/wan/model.py:124: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. + @amp.autocast(enabled=False) +/proj/cvl/users/x_fahkh2/Mobile-OV_Alpha/nets/third_party/sana/diffusion/model/wan/model.py:134: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. + @amp.autocast(enabled=False) +/proj/cvl/users/x_fahkh2/Mobile-OV_Alpha/nets/third_party/sana/diffusion/model/nets/fastlinear/modules/nn/norm.py:186: UserWarning: Cannot import apex RMSNorm, switch to vanilla implementation + warnings.warn("Cannot import apex RMSNorm, switch to vanilla implementation") +/proj/cvl/users/x_fahkh2/Mobile-OV_Alpha/nets/third_party/sana/diffusion/model/wan/model.py:124: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. + @amp.autocast(enabled=False) +/proj/cvl/users/x_fahkh2/Mobile-OV_Alpha/nets/third_party/sana/diffusion/model/wan/model.py:134: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. + @amp.autocast(enabled=False) +/proj/cvl/users/x_fahkh2/Mobile-OV_Alpha/nets/third_party/sana/diffusion/model/nets/fastlinear/modules/nn/norm.py:186: UserWarning: Cannot import apex RMSNorm, switch to vanilla implementation + warnings.warn("Cannot import apex RMSNorm, switch to vanilla implementation") +/proj/cvl/users/x_fahkh2/Mobile-OV_Alpha/nets/third_party/sana/diffusion/model/nets/fastlinear/modules/nn/norm.py:186: UserWarning: Cannot import apex RMSNorm, switch to vanilla implementation + warnings.warn("Cannot import apex RMSNorm, switch to vanilla implementation") +/proj/cvl/users/x_fahkh2/Mobile-OV_Alpha/nets/third_party/sana/diffusion/model/wan/model.py:124: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. + @amp.autocast(enabled=False) +/proj/cvl/users/x_fahkh2/Mobile-OV_Alpha/nets/third_party/sana/diffusion/model/wan/model.py:134: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. + @amp.autocast(enabled=False) +/proj/cvl/users/x_fahkh2/Mobile-OV_Alpha/nets/third_party/sana/diffusion/model/wan/model.py:124: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. + @amp.autocast(enabled=False) +/proj/cvl/users/x_fahkh2/Mobile-OV_Alpha/nets/third_party/sana/diffusion/model/wan/model.py:134: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. + @amp.autocast(enabled=False) +/proj/cvl/users/x_fahkh2/Mobile-OV_Alpha/nets/third_party/sana/diffusion/model/nets/fastlinear/modules/nn/norm.py:186: UserWarning: Cannot import apex RMSNorm, switch to vanilla implementation + warnings.warn("Cannot import apex RMSNorm, switch to vanilla implementation") +/proj/cvl/users/x_fahkh2/Mobile-OV_Alpha/nets/third_party/sana/diffusion/model/nets/fastlinear/modules/nn/norm.py:186: UserWarning: Cannot import apex RMSNorm, switch to vanilla implementation + warnings.warn("Cannot import apex RMSNorm, switch to vanilla implementation") +/proj/cvl/users/x_fahkh2/Mobile-OV_Alpha/nets/third_party/sana/diffusion/model/nets/fastlinear/modules/nn/norm.py:186: UserWarning: Cannot import apex RMSNorm, switch to vanilla implementation + warnings.warn("Cannot import apex RMSNorm, switch to vanilla implementation") +/proj/cvl/users/x_fahkh2/Mobile-OV_Alpha/nets/third_party/sana/diffusion/model/wan/model.py:124: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. + @amp.autocast(enabled=False) +/proj/cvl/users/x_fahkh2/Mobile-OV_Alpha/nets/third_party/sana/diffusion/model/wan/model.py:134: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. + @amp.autocast(enabled=False) +/proj/cvl/users/x_fahkh2/Mobile-OV_Alpha/nets/third_party/sana/diffusion/model/nets/fastlinear/modules/nn/norm.py:186: UserWarning: Cannot import apex RMSNorm, switch to vanilla implementation + warnings.warn("Cannot import apex RMSNorm, switch to vanilla implementation") +/proj/cvl/users/x_fahkh2/Mobile-OV_Alpha/nets/third_party/sana/diffusion/model/nets/fastlinear/modules/nn/norm.py:186: UserWarning: Cannot import apex RMSNorm, switch to vanilla implementation + warnings.warn("Cannot import apex RMSNorm, switch to vanilla implementation") +2026-05-10 12:44:31,662 - INFO - Distributed init: world_size=8 local_rank=0 +2026-05-10 12:44:31,675 - INFO - Bootstrap stage: loaded config, auto_download_sana=True auto_download_smol=True +2026-05-10 12:44:31,675 - INFO - Bootstrap stage: ensure_sana_assets_available (start) +2026-05-10 12:44:31,675 - INFO - Bootstrap stage: ensure_sana_assets_available (done) +2026-05-10 12:44:31,675 - INFO - Bootstrap stage: ensure_smolvlm2_checkpoint_available (start) +2026-05-10 12:44:31,675 - INFO - Bootstrap stage: ensure_smolvlm2_checkpoint_available (done) +2026-05-10 12:44:31,676 - INFO - Init prewrap: torch.load checkpoint (start) path=output/stage1_bridge_fulldit_full_mobile_o_smolvlm2_500m_lexical_gated_k2_diffonly_init10k_bs64_v2_20260420_8gpu/20260425_135135/checkpoint_latest.pt +2026-05-10 12:44:39,977 - INFO - Building SANA model... +2026-05-10 12:44:39,977 - INFO - SANA build options: use_grad_checkpoint=True gc_step=1 use_fp32_attention=True +2026-05-10 12:44:39,977 - INFO - Building SANA model... +2026-05-10 12:44:39,977 - INFO - SANA build options: use_grad_checkpoint=True gc_step=1 use_fp32_attention=True +2026-05-10 12:44:39,977 - INFO - Building SANA model... +2026-05-10 12:44:39,977 - INFO - Init prewrap: torch.load checkpoint (done) step=17600 micro_step=17600 keys=['dit_train_modules', 'dit_trainable_state', 'infer_hints', 'micro_step', 'optimizer', 'scheduler', 'step', 'student_state'] +2026-05-10 12:44:39,977 - INFO - Building SANA model... +2026-05-10 12:44:39,977 - INFO - SANA build options: use_grad_checkpoint=True gc_step=1 use_fp32_attention=True +2026-05-10 12:44:39,977 - INFO - Strict SANA-parity text path enabled: model_max_length=300 fail_fast_mask=True full_text_window=True select_strategy=head_uniform_tail head_tokens=96 tail_tokens=96 +2026-05-10 12:44:39,977 - INFO - SANA build options: use_grad_checkpoint=True gc_step=1 use_fp32_attention=True +2026-05-10 12:44:39,977 - INFO - Building SANA model... +2026-05-10 12:44:39,977 - INFO - Building SANA model... +2026-05-10 12:44:39,977 - INFO - SANA build options: use_grad_checkpoint=True gc_step=1 use_fp32_attention=True +2026-05-10 12:44:39,977 - INFO - SANA build options: use_grad_checkpoint=True gc_step=1 use_fp32_attention=True +2026-05-10 12:44:39,977 - INFO - Building SANA model... +2026-05-10 12:44:39,977 - INFO - SANA build options: use_grad_checkpoint=True gc_step=1 use_fp32_attention=True +2026-05-10 12:44:39,980 - INFO - Building SANA model... +2026-05-10 12:44:39,981 - INFO - SANA build options: use_grad_checkpoint=True gc_step=1 use_fp32_attention=True +2026-05-10 18:44:45 - [Sana] - INFO - use pe: True, pos embed type: wan_rope, position embed interpolation: 1.0, base size: 60 +2026-05-10 18:44:45 - [Sana] - INFO - attention type: LiteLAReLURope; ffn type: GLUMBConvTemp; self-attn head dim: 112; self-attn qk norm: True; cross-attn type: flash; cross-attn qk norm: True; autocast linear attn: False +2026-05-10 18:44:52 - [Sana] - INFO - GLUMBConvTemp Temporal kernal: 3 +2026-05-10 12:44:58,045 - INFO - Built SANA model +2026-05-10 12:44:58,047 - INFO - Loading SANA checkpoint from /proj/cvl/users/x_fahkh2/Mobile-OV_Alpha/omni_ckpts/sana_video_2b_480p/checkpoints/SANA_Video_2B_480p.pth ... +2026-05-10 12:44:58,060 - INFO - Built SANA model +2026-05-10 12:44:58,062 - INFO - Loading SANA checkpoint from /proj/cvl/users/x_fahkh2/Mobile-OV_Alpha/omni_ckpts/sana_video_2b_480p/checkpoints/SANA_Video_2B_480p.pth ... +2026-05-10 12:44:58,188 - INFO - Built SANA model +2026-05-10 12:44:58,190 - INFO - Loading SANA checkpoint from /proj/cvl/users/x_fahkh2/Mobile-OV_Alpha/omni_ckpts/sana_video_2b_480p/checkpoints/SANA_Video_2B_480p.pth ... +2026-05-10 12:44:58,641 - INFO - Built SANA model +2026-05-10 12:44:58,642 - INFO - Loading SANA checkpoint from /proj/cvl/users/x_fahkh2/Mobile-OV_Alpha/omni_ckpts/sana_video_2b_480p/checkpoints/SANA_Video_2B_480p.pth ... +2026-05-10 12:44:58,667 - INFO - Built SANA model +2026-05-10 12:44:58,668 - INFO - Loading SANA checkpoint from /proj/cvl/users/x_fahkh2/Mobile-OV_Alpha/omni_ckpts/sana_video_2b_480p/checkpoints/SANA_Video_2B_480p.pth ... +2026-05-10 12:44:58,757 - INFO - Built SANA model +2026-05-10 12:44:58,759 - INFO - Loading SANA checkpoint from /proj/cvl/users/x_fahkh2/Mobile-OV_Alpha/omni_ckpts/sana_video_2b_480p/checkpoints/SANA_Video_2B_480p.pth ... +2026-05-10 12:44:59,196 - INFO - Built SANA model +2026-05-10 12:44:59,198 - INFO - Loading SANA checkpoint from /proj/cvl/users/x_fahkh2/Mobile-OV_Alpha/omni_ckpts/sana_video_2b_480p/checkpoints/SANA_Video_2B_480p.pth ... +2026-05-10 12:44:59,712 - INFO - Built SANA model +2026-05-10 12:44:59,714 - INFO - Loading SANA checkpoint from /proj/cvl/users/x_fahkh2/Mobile-OV_Alpha/omni_ckpts/sana_video_2b_480p/checkpoints/SANA_Video_2B_480p.pth ... +2026-05-10 12:45:08,533 - INFO - Loaded SANA checkpoint in 10.5s +2026-05-10 12:45:08,533 - INFO - Loaded SANA checkpoint in 9.9s +2026-05-10 12:45:08,533 - INFO - Loaded SANA checkpoint in 9.8s +2026-05-10 12:45:08,533 - INFO - Loaded SANA checkpoint in 10.5s +2026-05-10 12:45:08,533 - INFO - Loaded SANA checkpoint in 8.8s +2026-05-10 12:45:08,533 - INFO - Loaded SANA checkpoint in 9.9s +2026-05-10 12:45:08,533 - INFO - Loaded SANA checkpoint in 9.3s +2026-05-10 12:45:08,534 - INFO - Loaded SANA checkpoint in 10.3s +2026-05-10 12:45:15,454 - INFO - Loaded SANA DiT checkpoint +2026-05-10 12:45:15,454 - INFO - SANA caption dropout: y_embedder.uncond_prob=0.0000 +2026-05-10 12:45:15,456 - INFO - Configured DiT trainable modules: ['all'] (has_trainable=True, wrap_fsdp=True, wrap_ddp=False) +2026-05-10 12:45:15,456 - INFO - Init prewrap: loading dit_trainable_state into unwrapped DiT (start) keys=416 +2026-05-10 12:45:16,545 - INFO - Init prewrap: loading dit_trainable_state into unwrapped DiT (done) keys=416 missing=2 unexpected=0 +2026-05-10 12:45:16,545 - INFO - Wrapping DiT with FSDP (use_orig_params=True, sync_module_states=False, sharding=full_shard) +2026-05-10 12:45:16,571 - INFO - FSDP wrap done +2026-05-10 12:45:16,571 - INFO - Rank 0 entering build_student +2026-05-10 12:45:16,571 - INFO - build_student: backbone_type=smolvlm2 ckpt_path=omni_ckpts/smolvlm2_500m/smolvlm2_500m.pt projector_type=mcp_lexical_gated +2026-05-10 12:45:16,571 - INFO - Rank 5 entering build_student +2026-05-10 12:45:16,571 - INFO - Rank 6 entering build_student +2026-05-10 12:45:16,571 - INFO - build_student: backbone_type=smolvlm2 ckpt_path=omni_ckpts/smolvlm2_500m/smolvlm2_500m.pt projector_type=mcp_lexical_gated +2026-05-10 12:45:16,571 - INFO - build_student: backbone_type=smolvlm2 ckpt_path=omni_ckpts/smolvlm2_500m/smolvlm2_500m.pt projector_type=mcp_lexical_gated +2026-05-10 12:45:16,571 - INFO - Rank 4 entering build_student +2026-05-10 12:45:16,571 - INFO - Rank 2 entering build_student +2026-05-10 12:45:16,572 - INFO - build_student: backbone_type=smolvlm2 ckpt_path=omni_ckpts/smolvlm2_500m/smolvlm2_500m.pt projector_type=mcp_lexical_gated +2026-05-10 12:45:16,572 - INFO - build_student: backbone_type=smolvlm2 ckpt_path=omni_ckpts/smolvlm2_500m/smolvlm2_500m.pt projector_type=mcp_lexical_gated +2026-05-10 12:45:16,572 - INFO - Rank 7 entering build_student +2026-05-10 12:45:16,572 - INFO - build_student: backbone_type=smolvlm2 ckpt_path=omni_ckpts/smolvlm2_500m/smolvlm2_500m.pt projector_type=mcp_lexical_gated +2026-05-10 12:45:16,573 - INFO - Loading SmolVLM2 model from omni_ckpts/smolvlm2_500m/smolvlm2_500m.pt +2026-05-10 12:45:16,573 - INFO - Loading SmolVLM2 model from omni_ckpts/smolvlm2_500m/smolvlm2_500m.pt +2026-05-10 12:45:16,573 - INFO - Loading SmolVLM2 model from omni_ckpts/smolvlm2_500m/smolvlm2_500m.pt +2026-05-10 12:45:16,573 - INFO - Loading SmolVLM2 model from omni_ckpts/smolvlm2_500m/smolvlm2_500m.pt +2026-05-10 12:45:16,573 - INFO - Loading SmolVLM2 model from omni_ckpts/smolvlm2_500m/smolvlm2_500m.pt +2026-05-10 12:45:16,574 - INFO - Loading SmolVLM2 model from omni_ckpts/smolvlm2_500m/smolvlm2_500m.pt +2026-05-10 12:45:16,578 - INFO - Rank 3 entering build_student +2026-05-10 12:45:16,579 - INFO - build_student: backbone_type=smolvlm2 ckpt_path=omni_ckpts/smolvlm2_500m/smolvlm2_500m.pt projector_type=mcp_lexical_gated +2026-05-10 12:45:16,579 - INFO - Loading SmolVLM2 model from omni_ckpts/smolvlm2_500m/smolvlm2_500m.pt +2026-05-10 12:45:16,624 - INFO - Rank 1 entering build_student +2026-05-10 12:45:16,625 - INFO - build_student: backbone_type=smolvlm2 ckpt_path=omni_ckpts/smolvlm2_500m/smolvlm2_500m.pt projector_type=mcp_lexical_gated +2026-05-10 12:45:16,625 - INFO - Loading SmolVLM2 model from omni_ckpts/smolvlm2_500m/smolvlm2_500m.pt +2026-05-10 12:45:18,344 - INFO - ✓ Model loaded successfully from full object checkpoint +2026-05-10 12:45:18,344 - INFO - ✓ Model loaded successfully from full object checkpoint +2026-05-10 12:45:18,347 - INFO - Detected SmolVLM2 hidden_size=960 (overriding VisionHead input) +2026-05-10 12:45:18,348 - INFO - Detected SmolVLM2 hidden_size=960 (overriding VisionHead input) +2026-05-10 12:45:18,358 - INFO - Using MCP projector: type=mcp_lexical_gated d_vlm=960 no_compress=True hidden=960 d_cond=2304 K=2 refine=True lexical_mode=gated_add lexical_bottleneck_dim=256 lexical_gate_init=0.2000 +2026-05-10 12:45:18,358 - INFO - Using MCP projector: type=mcp_lexical_gated d_vlm=960 no_compress=True hidden=960 d_cond=2304 K=2 refine=True lexical_mode=gated_add lexical_bottleneck_dim=256 lexical_gate_init=0.2000 +2026-05-10 12:45:18,393 - INFO - Rank 4 using manual gradient all-reduce for student (DDP disabled) +2026-05-10 12:45:18,393 - INFO - Rank 4 building dataset +2026-05-10 12:45:18,393 - INFO - Rank 6 using manual gradient all-reduce for student (DDP disabled) +2026-05-10 12:45:18,393 - INFO - Rank 6 building dataset +2026-05-10 12:45:18,614 - INFO - ✓ Model loaded successfully from full object checkpoint +2026-05-10 12:45:18,617 - INFO - Detected SmolVLM2 hidden_size=960 (overriding VisionHead input) +2026-05-10 12:45:18,628 - INFO - Using MCP projector: type=mcp_lexical_gated d_vlm=960 no_compress=True hidden=960 d_cond=2304 K=2 refine=True lexical_mode=gated_add lexical_bottleneck_dim=256 lexical_gate_init=0.2000 +2026-05-10 12:45:18,645 - INFO - Rank 3 using manual gradient all-reduce for student (DDP disabled) +2026-05-10 12:45:18,645 - INFO - Rank 3 building dataset +2026-05-10 12:45:18,764 - INFO - ✓ Model loaded successfully from full object checkpoint +2026-05-10 12:45:18,768 - INFO - Detected SmolVLM2 hidden_size=960 (overriding VisionHead input) +2026-05-10 12:45:18,778 - INFO - Using MCP projector: type=mcp_lexical_gated d_vlm=960 no_compress=True hidden=960 d_cond=2304 K=2 refine=True lexical_mode=gated_add lexical_bottleneck_dim=256 lexical_gate_init=0.2000 +2026-05-10 12:45:18,797 - INFO - Rank 1 using manual gradient all-reduce for student (DDP disabled) +2026-05-10 12:45:18,797 - INFO - Rank 1 building dataset +2026-05-10 12:45:19,148 - INFO - ✓ Model loaded successfully from full object checkpoint +2026-05-10 12:45:19,151 - INFO - Detected SmolVLM2 hidden_size=960 (overriding VisionHead input) +2026-05-10 12:45:19,162 - INFO - Using MCP projector: type=mcp_lexical_gated d_vlm=960 no_compress=True hidden=960 d_cond=2304 K=2 refine=True lexical_mode=gated_add lexical_bottleneck_dim=256 lexical_gate_init=0.2000 +2026-05-10 12:45:19,180 - INFO - Rank 7 using manual gradient all-reduce for student (DDP disabled) +2026-05-10 12:45:19,180 - INFO - Rank 7 building dataset +2026-05-10 12:45:19,622 - INFO - ✓ Model loaded successfully from full object checkpoint +2026-05-10 12:45:19,625 - INFO - Detected SmolVLM2 hidden_size=960 (overriding VisionHead input) +2026-05-10 12:45:19,636 - INFO - Using MCP projector: type=mcp_lexical_gated d_vlm=960 no_compress=True hidden=960 d_cond=2304 K=2 refine=True lexical_mode=gated_add lexical_bottleneck_dim=256 lexical_gate_init=0.2000 +2026-05-10 12:45:19,653 - INFO - Rank 5 using manual gradient all-reduce for student (DDP disabled) +2026-05-10 12:45:19,653 - INFO - Rank 5 building dataset +2026-05-10 12:45:20,169 - INFO - ✓ Model loaded successfully from full object checkpoint +2026-05-10 12:45:20,172 - INFO - Detected SmolVLM2 hidden_size=960 (overriding VisionHead input) +2026-05-10 12:45:20,183 - INFO - Using MCP projector: type=mcp_lexical_gated d_vlm=960 no_compress=True hidden=960 d_cond=2304 K=2 refine=True lexical_mode=gated_add lexical_bottleneck_dim=256 lexical_gate_init=0.2000 +2026-05-10 12:45:20,201 - INFO - Rank 2 using manual gradient all-reduce for student (DDP disabled) +2026-05-10 12:45:20,201 - INFO - Rank 2 building dataset +2026-05-10 12:45:20,202 - INFO - Loaded dataset index cache from /home/x_fahkh/.cache/mobileov_dataset_cache/openvid_index_6d9c5a6a276769da.pkl (1019074 samples) +2026-05-10 12:45:20,202 - INFO - Loaded dataset index cache from /home/x_fahkh/.cache/mobileov_dataset_cache/openvid_index_6d9c5a6a276769da.pkl (1019074 samples) +2026-05-10 12:45:20,203 - INFO - Loaded 1019074 valid samples from OpenVid-style CSV (modality_filter=['video']) +2026-05-10 12:45:20,203 - INFO - Loaded 1019074 valid samples from OpenVid-style CSV (modality_filter=['video']) +2026-05-10 12:45:20,203 - INFO - Loaded dataset index cache from /home/x_fahkh/.cache/mobileov_dataset_cache/openvid_index_6d9c5a6a276769da.pkl (1019074 samples) +2026-05-10 12:45:20,203 - INFO - Loaded 1019074 valid samples from OpenVid-style CSV (modality_filter=['video']) +2026-05-10 12:45:20,203 - INFO - Loaded dataset index cache from /home/x_fahkh/.cache/mobileov_dataset_cache/openvid_index_6d9c5a6a276769da.pkl (1019074 samples) +2026-05-10 12:45:20,203 - INFO - Loaded 1019074 valid samples from OpenVid-style CSV (modality_filter=['video']) +2026-05-10 12:45:20,232 - INFO - Loaded dataset index cache from /home/x_fahkh/.cache/mobileov_dataset_cache/openvid_index_6d9c5a6a276769da.pkl (1019074 samples) +2026-05-10 12:45:20,233 - INFO - Loaded 1019074 valid samples from OpenVid-style CSV (modality_filter=['video']) +2026-05-10 12:45:20,360 - INFO - ✓ Model loaded successfully from full object checkpoint +2026-05-10 12:45:20,363 - INFO - Detected SmolVLM2 hidden_size=960 (overriding VisionHead input) +2026-05-10 12:45:20,374 - INFO - Using MCP projector: type=mcp_lexical_gated d_vlm=960 no_compress=True hidden=960 d_cond=2304 K=2 refine=True lexical_mode=gated_add lexical_bottleneck_dim=256 lexical_gate_init=0.2000 +2026-05-10 12:45:20,392 - INFO - Built student model +2026-05-10 12:45:20,392 - INFO - Rank 0 using manual gradient all-reduce for student (DDP disabled) +2026-05-10 12:45:20,392 - INFO - Bridge tokenizer: class=GPT2TokenizerFast module=transformers.models.gpt2.tokenization_gpt2_fast name_or_path=HuggingFaceTB/SmolVLM2-500M-Video-Instruct +2026-05-10 12:45:20,392 - INFO - Rank 0 building dataset +2026-05-10 12:45:20,701 - INFO - Loaded dataset index cache from /home/x_fahkh/.cache/mobileov_dataset_cache/openvid_index_6d9c5a6a276769da.pkl (1019074 samples) +2026-05-10 12:45:20,701 - INFO - Loaded 1019074 valid samples from OpenVid-style CSV (modality_filter=['video']) +2026-05-10 12:45:21,222 - INFO - Loaded dataset index cache from /home/x_fahkh/.cache/mobileov_dataset_cache/openvid_index_6d9c5a6a276769da.pkl (1019074 samples) +2026-05-10 12:45:21,223 - INFO - Loaded 1019074 valid samples from OpenVid-style CSV (modality_filter=['video']) +2026-05-10 12:45:21,472 - INFO - Loaded dataset index cache from /home/x_fahkh/.cache/mobileov_dataset_cache/openvid_index_6d9c5a6a276769da.pkl (1019074 samples) +2026-05-10 12:45:21,472 - INFO - Loaded 1019074 valid samples from OpenVid-style CSV (modality_filter=['video']) +2026-05-10 12:45:27,945 - INFO - Loaded dataset index cache from /home/x_fahkh/.cache/mobileov_dataset_cache/openvid_index_b81f63f227de2ac7.pkl (8996310 samples) +2026-05-10 12:45:27,945 - INFO - Loaded dataset index cache from /home/x_fahkh/.cache/mobileov_dataset_cache/openvid_index_b81f63f227de2ac7.pkl (8996310 samples) +2026-05-10 12:45:27,945 - INFO - Loaded dataset index cache from /home/x_fahkh/.cache/mobileov_dataset_cache/openvid_index_b81f63f227de2ac7.pkl (8996310 samples) +2026-05-10 12:45:27,945 - INFO - Loaded 8996310 valid samples from OpenVid-style CSV (modality_filter=['image']) +2026-05-10 12:45:27,945 - INFO - Loaded 8996310 valid samples from OpenVid-style CSV (modality_filter=['image']) +2026-05-10 12:45:27,945 - INFO - Rank 7 datasets initialized: video=1019074 image=8996310 +2026-05-10 12:45:27,945 - INFO - Loaded 8996310 valid samples from OpenVid-style CSV (modality_filter=['image']) +2026-05-10 12:45:27,945 - INFO - Rank 3 datasets initialized: video=1019074 image=8996310 +2026-05-10 12:45:27,945 - INFO - Rank 1 datasets initialized: video=1019074 image=8996310 +2026-05-10 12:45:27,945 - INFO - Loaded dataset index cache from /home/x_fahkh/.cache/mobileov_dataset_cache/openvid_index_b81f63f227de2ac7.pkl (8996310 samples) +2026-05-10 12:45:27,945 - INFO - Loaded 8996310 valid samples from OpenVid-style CSV (modality_filter=['image']) +2026-05-10 12:45:27,945 - INFO - Rank 6 datasets initialized: video=1019074 image=8996310 +2026-05-10 12:45:27,947 - INFO - Loaded dataset index cache from /home/x_fahkh/.cache/mobileov_dataset_cache/openvid_index_b81f63f227de2ac7.pkl (8996310 samples) +2026-05-10 12:45:27,947 - INFO - Loaded 8996310 valid samples from OpenVid-style CSV (modality_filter=['image']) +2026-05-10 12:45:27,947 - INFO - Rank 2 datasets initialized: video=1019074 image=8996310 +2026-05-10 12:45:27,947 - INFO - Rank 7 trainable bridge params: tensors=14 numel=3146694 +2026-05-10 12:45:27,947 - INFO - Rank 3 trainable bridge params: tensors=14 numel=3146694 +2026-05-10 12:45:27,948 - INFO - Rank 7 trainable DiT params: tensors=416 numel=257107768 +2026-05-10 12:45:27,948 - INFO - Rank 3 trainable DiT params: tensors=416 numel=257107768 +2026-05-10 12:45:27,948 - INFO - Loaded dataset index cache from /home/x_fahkh/.cache/mobileov_dataset_cache/openvid_index_b81f63f227de2ac7.pkl (8996310 samples) +2026-05-10 12:45:27,948 - INFO - Loaded 8996310 valid samples from OpenVid-style CSV (modality_filter=['image']) +2026-05-10 12:45:27,948 - INFO - Rank 5 datasets initialized: video=1019074 image=8996310 +2026-05-10 12:45:27,948 - INFO - Rank 6 trainable bridge params: tensors=14 numel=3146694 +2026-05-10 12:45:27,948 - INFO - Rank 6 trainable DiT params: tensors=416 numel=257107768 +2026-05-10 12:45:27,948 - INFO - Rank 1 trainable bridge params: tensors=14 numel=3146694 +2026-05-10 12:45:27,949 - INFO - Rank 1 trainable DiT params: tensors=416 numel=257107768 +2026-05-10 12:45:27,950 - INFO - Rank 5 trainable bridge params: tensors=14 numel=3146694 +2026-05-10 12:45:27,950 - INFO - Rank 5 trainable DiT params: tensors=416 numel=257107768 +2026-05-10 12:45:27,950 - INFO - Rank 2 trainable bridge params: tensors=14 numel=3146694 +2026-05-10 12:45:27,951 - INFO - Rank 2 trainable DiT params: tensors=416 numel=257107768 +2026-05-10 12:45:27,952 - INFO - Loaded dataset index cache from /home/x_fahkh/.cache/mobileov_dataset_cache/openvid_index_b81f63f227de2ac7.pkl (8996310 samples) +2026-05-10 12:45:27,952 - INFO - Loaded 8996310 valid samples from OpenVid-style CSV (modality_filter=['image']) +2026-05-10 12:45:27,952 - INFO - Datasets initialized: video=1019074 (video), image=8996310 (image), interval=5 +2026-05-10 12:45:27,952 - INFO - Dataloader ready (num_workers=0, joint=True, video_batches=127385, image_batches=281135) +2026-05-10 12:45:27,954 - INFO - Rank 0 trainable bridge params: tensors=14 numel=3146694 +2026-05-10 12:45:27,955 - INFO - Rank 0 trainable DiT params: tensors=416 numel=257107768 +2026-05-10 12:45:27,955 - INFO - LR schedule: mode=constant warmup=0 total_steps=500000 min_ratio=0.1000 cycles=0.500 decay_ratio=1.000 +/proj/cvl/users/x_fahkh2/Mobile-OV_Alpha/nets/third_party/sana/diffusion/model/gaussian_diffusion.py:224: RuntimeWarning: divide by zero encountered in divide + self.sqrt_recip_alphas_cumprod = np.sqrt(1.0 / self.alphas_cumprod) +/proj/cvl/users/x_fahkh2/Mobile-OV_Alpha/nets/third_party/sana/diffusion/model/gaussian_diffusion.py:225: RuntimeWarning: divide by zero encountered in divide + self.sqrt_recipm1_alphas_cumprod = np.sqrt(1.0 / self.alphas_cumprod - 1) +2026-05-10 12:45:27,960 - INFO - Loaded dataset index cache from /home/x_fahkh/.cache/mobileov_dataset_cache/openvid_index_b81f63f227de2ac7.pkl (8996310 samples) +2026-05-10 12:45:27,960 - INFO - Loaded 8996310 valid samples from OpenVid-style CSV (modality_filter=['image']) +2026-05-10 12:45:27,960 - INFO - Rank 4 datasets initialized: video=1019074 image=8996310 +/proj/cvl/users/x_fahkh2/Mobile-OV_Alpha/nets/third_party/sana/diffusion/model/gaussian_diffusion.py:224: RuntimeWarning: divide by zero encountered in divide + self.sqrt_recip_alphas_cumprod = np.sqrt(1.0 / self.alphas_cumprod) +/proj/cvl/users/x_fahkh2/Mobile-OV_Alpha/nets/third_party/sana/diffusion/model/gaussian_diffusion.py:225: RuntimeWarning: divide by zero encountered in divide + self.sqrt_recipm1_alphas_cumprod = np.sqrt(1.0 / self.alphas_cumprod - 1) +2026-05-10 12:45:27,963 - INFO - Rank 4 trainable bridge params: tensors=14 numel=3146694 +2026-05-10 12:45:27,963 - INFO - Rank 4 trainable DiT params: tensors=416 numel=257107768 +/proj/cvl/users/x_fahkh2/Mobile-OV_Alpha/nets/third_party/sana/diffusion/model/gaussian_diffusion.py:224: RuntimeWarning: divide by zero encountered in divide + self.sqrt_recip_alphas_cumprod = np.sqrt(1.0 / self.alphas_cumprod) +/proj/cvl/users/x_fahkh2/Mobile-OV_Alpha/nets/third_party/sana/diffusion/model/gaussian_diffusion.py:225: RuntimeWarning: divide by zero encountered in divide + self.sqrt_recipm1_alphas_cumprod = np.sqrt(1.0 / self.alphas_cumprod - 1) +/proj/cvl/users/x_fahkh2/Mobile-OV_Alpha/nets/third_party/sana/diffusion/model/gaussian_diffusion.py:224: RuntimeWarning: divide by zero encountered in divide + self.sqrt_recip_alphas_cumprod = np.sqrt(1.0 / self.alphas_cumprod) +/proj/cvl/users/x_fahkh2/Mobile-OV_Alpha/nets/third_party/sana/diffusion/model/gaussian_diffusion.py:225: RuntimeWarning: divide by zero encountered in divide + self.sqrt_recipm1_alphas_cumprod = np.sqrt(1.0 / self.alphas_cumprod - 1) +/proj/cvl/users/x_fahkh2/Mobile-OV_Alpha/nets/third_party/sana/diffusion/model/gaussian_diffusion.py:224: RuntimeWarning: divide by zero encountered in divide + self.sqrt_recip_alphas_cumprod = np.sqrt(1.0 / self.alphas_cumprod) +/proj/cvl/users/x_fahkh2/Mobile-OV_Alpha/nets/third_party/sana/diffusion/model/gaussian_diffusion.py:224: RuntimeWarning: divide by zero encountered in divide + self.sqrt_recip_alphas_cumprod = np.sqrt(1.0 / self.alphas_cumprod) +/proj/cvl/users/x_fahkh2/Mobile-OV_Alpha/nets/third_party/sana/diffusion/model/gaussian_diffusion.py:225: RuntimeWarning: divide by zero encountered in divide + self.sqrt_recipm1_alphas_cumprod = np.sqrt(1.0 / self.alphas_cumprod - 1) +/proj/cvl/users/x_fahkh2/Mobile-OV_Alpha/nets/third_party/sana/diffusion/model/gaussian_diffusion.py:224: RuntimeWarning: divide by zero encountered in divide + self.sqrt_recip_alphas_cumprod = np.sqrt(1.0 / self.alphas_cumprod) +/proj/cvl/users/x_fahkh2/Mobile-OV_Alpha/nets/third_party/sana/diffusion/model/gaussian_diffusion.py:225: RuntimeWarning: divide by zero encountered in divide + self.sqrt_recipm1_alphas_cumprod = np.sqrt(1.0 / self.alphas_cumprod - 1) +/proj/cvl/users/x_fahkh2/Mobile-OV_Alpha/nets/third_party/sana/diffusion/model/gaussian_diffusion.py:224: RuntimeWarning: divide by zero encountered in divide + self.sqrt_recip_alphas_cumprod = np.sqrt(1.0 / self.alphas_cumprod) +/proj/cvl/users/x_fahkh2/Mobile-OV_Alpha/nets/third_party/sana/diffusion/model/gaussian_diffusion.py:225: RuntimeWarning: divide by zero encountered in divide + self.sqrt_recipm1_alphas_cumprod = np.sqrt(1.0 / self.alphas_cumprod - 1) +/proj/cvl/users/x_fahkh2/Mobile-OV_Alpha/nets/third_party/sana/diffusion/model/gaussian_diffusion.py:225: RuntimeWarning: divide by zero encountered in divide + self.sqrt_recipm1_alphas_cumprod = np.sqrt(1.0 / self.alphas_cumprod - 1) +2026-05-10 12:45:28,241 - INFO - SANA objective: train_sampling_steps=1000 noise_schedule=linear_flow predict_flow_v=True flow_shift=3.000 weighting_scheme=logit_normal mode_scale=1.2900 use_process_timesteps=False force_ivjoint_timestep=True chunk_index=None chunk_sampling=uniform time_sampler=False timestep_weight=False +2026-05-10 12:45:28,241 - INFO - Aux losses will be enabled from update_step >= 1 +2026-05-10 12:45:29,472 - INFO - CFG uncond bridge bypass enabled: fixed_eval=True prompt='' fixed_len=300 fixed_tok=1.0 +2026-05-10 12:45:29,473 - INFO - Manual gradient sync backend: nccl +2026-05-10 12:45:29,473 - INFO - Trainable params bridge: 3.15M +2026-05-10 12:45:29,474 - INFO - Trainable params DiT: 257.11M +2026-05-10 12:45:29,474 - INFO - CFG settings: dropout_prob=0.000 delta_every_steps=0 uncond_prompt='' +2026-05-10 12:45:29,474 - INFO - CFG uncond detach settings: enabled=True fixed_eval=True +2026-05-10 12:45:29,474 - INFO - Conditioning diagnostics: every=200 shuffle=True uncond=True grad=True +2026-05-10 12:45:29,474 - INFO - Init stage: loading student_state (start) +2026-05-10 12:45:29,476 - INFO - Init stage: loading student_state (done) +2026-05-10 12:45:29,476 - INFO - Initialized model weights from output/stage1_bridge_fulldit_full_mobile_o_smolvlm2_500m_lexical_gated_k2_diffonly_init10k_bs64_v2_20260420_8gpu/20260425_135135/checkpoint_latest.pt (starting fresh at step=0) +2026-05-10 12:45:29,477 - INFO - Entering training loop; waiting for first batch... +2026-05-10 12:45:29,505 - INFO - Rank 5 before student forward +2026-05-10 12:45:29,506 - INFO - Rank 4 before student forward +2026-05-10 12:45:29,506 - INFO - Rank 1 before student forward +2026-05-10 12:45:29,507 - INFO - Rank 3 before student forward +2026-05-10 12:45:29,508 - INFO - Rank 2 before student forward +2026-05-10 12:45:29,515 - INFO - Rank 7 before student forward +2026-05-10 12:45:29,520 - INFO - Rank 6 before student forward +2026-05-10 12:45:29,525 - INFO - Rank 2 reached student forward +2026-05-10 12:45:29,525 - INFO - Rank 3 reached student forward +2026-05-10 12:45:29,525 - INFO - Rank 1 reached student forward +2026-05-10 12:45:29,525 - INFO - Rank 4 reached student forward +2026-05-10 12:45:29,525 - INFO - Rank 5 reached student forward +2026-05-10 12:45:29,533 - INFO - Rank 7 reached student forward +2026-05-10 12:45:29,534 - INFO - First batch fetched in 0.06s (mode=video) +2026-05-10 12:45:29,534 - INFO - Rank 0 before student forward +2026-05-10 12:45:29,538 - INFO - Rank 6 reached student forward +2026-05-10 12:45:29,552 - INFO - Rank 0 reached student forward +2026-05-10 12:45:29,553 - INFO - Observed train latent_t=21 (raw dataset) +2026-05-10 12:45:29,553 - INFO - Observed train frame_num=81 +2026-05-10 12:45:29,553 - INFO - Latent shape (1, 16, 21, 60, 104) dtype=torch.bfloat16 +2026-05-10 12:45:29,553 - INFO - Observed train effective latent_t=21 (post-window) +2026-05-10 12:45:29,615 - INFO - Rank 7 before SANA training_losses forward +2026-05-10 12:45:29,616 - INFO - Rank 3 before SANA training_losses forward +2026-05-10 12:45:29,616 - INFO - Rank 2 before SANA training_losses forward +2026-05-10 12:45:29,616 - INFO - Rank 0 before SANA training_losses forward +2026-05-10 12:45:29,616 - INFO - Rank 6 before SANA training_losses forward +2026-05-10 12:45:29,616 - INFO - Rank 1 before SANA training_losses forward +2026-05-10 12:45:29,616 - INFO - Rank 5 before SANA training_losses forward +2026-05-10 12:45:29,616 - INFO - Rank 4 before SANA training_losses forward +2026-05-10 12:45:38,774 - INFO - Rank 3 after SANA training_losses forward +2026-05-10 12:45:38,774 - INFO - Rank 6 after SANA training_losses forward +2026-05-10 12:45:39,252 - INFO - Rank 7 after SANA training_losses forward +2026-05-10 12:45:39,252 - INFO - Rank 0 after SANA training_losses forward +2026-05-10 12:45:39,252 - INFO - Rank 1 after SANA training_losses forward +2026-05-10 12:45:39,252 - INFO - Rank 4 after SANA training_losses forward +2026-05-10 12:45:39,252 - INFO - Rank 2 after SANA training_losses forward +2026-05-10 12:45:39,252 - INFO - Rank 5 after SANA training_losses forward +2026-05-10 12:45:42,048 - INFO - Rank 4 after backward +2026-05-10 12:45:42,048 - INFO - Rank 0 after backward +2026-05-10 12:45:42,048 - INFO - Rank 6 after backward +2026-05-10 12:45:42,048 - INFO - Rank 1 after backward +2026-05-10 12:45:42,048 - INFO - Rank 2 after backward +2026-05-10 12:45:42,048 - INFO - Rank 7 after backward +2026-05-10 12:45:42,048 - INFO - Rank 3 after backward +2026-05-10 12:45:42,048 - INFO - Rank 5 after backward +2026-05-10 12:45:42,082 - INFO - Rank 2 before optimizer.step +2026-05-10 12:45:42,082 - INFO - Rank 6 before optimizer.step +2026-05-10 12:45:42,082 - INFO - Rank 4 before optimizer.step +2026-05-10 12:45:42,082 - INFO - Rank 5 before optimizer.step +2026-05-10 12:45:42,083 - INFO - Rank 7 before optimizer.step +2026-05-10 12:45:42,083 - INFO - Rank 0 before optimizer.step +2026-05-10 12:45:42,083 - INFO - Rank 1 before optimizer.step +2026-05-10 12:45:42,083 - INFO - Rank 3 before optimizer.step +2026-05-10 12:45:42,178 - INFO - Rank 3 after optimizer.step +2026-05-10 12:45:42,178 - INFO - Rank 7 after optimizer.step +2026-05-10 12:45:42,178 - INFO - Rank 3 after scheduler.step +2026-05-10 12:45:42,178 - INFO - Rank 7 after scheduler.step +2026-05-10 12:45:42,178 - INFO - Rank 2 after optimizer.step +2026-05-10 12:45:42,178 - INFO - Rank 5 after optimizer.step +2026-05-10 12:45:42,178 - INFO - Rank 1 after optimizer.step +2026-05-10 12:45:42,178 - INFO - Rank 2 after scheduler.step +2026-05-10 12:45:42,178 - INFO - Rank 4 after optimizer.step +2026-05-10 12:45:42,178 - INFO - Rank 5 after scheduler.step +2026-05-10 12:45:42,178 - INFO - Rank 1 after scheduler.step +2026-05-10 12:45:42,178 - INFO - Rank 4 after scheduler.step +2026-05-10 12:45:42,178 - INFO - Rank 0 after optimizer.step +2026-05-10 12:45:42,178 - INFO - Rank 6 after optimizer.step +2026-05-10 12:45:42,179 - INFO - Rank 0 after scheduler.step +2026-05-10 12:45:42,179 - INFO - Rank 6 after scheduler.step +2026-05-10 12:45:59,950 - INFO - Step 10 | mode=video loss=0.156250 diff=0.156250 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1368 emb_mean=0.0007 emb_std=1.0018 tok_norm=48.0875 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=9 i_micro=1 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3733 mask_tok=112.00 y_norm=48.0875±0.0062 +2026-05-10 12:46:17,951 - INFO - Step 20 | mode=video loss=0.090820 diff=0.090820 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1187 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0855 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=17 i_micro=3 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3733 mask_tok=112.00 y_norm=48.0855±0.0062 +2026-05-10 12:46:37,504 - INFO - Step 30 | mode=video loss=0.140625 diff=0.140625 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0664 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0898 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=25 i_micro=5 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3567 mask_tok=107.00 y_norm=48.0898±0.0060 +2026-05-10 12:46:55,457 - INFO - Step 40 | mode=video loss=0.092773 diff=0.092773 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1315 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0880 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=33 i_micro=7 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3900 mask_tok=117.00 y_norm=48.0880±0.0074 +2026-05-10 12:47:14,079 - INFO - Step 50 | mode=video loss=0.090820 diff=0.090820 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0558 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0922 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=41 i_micro=9 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3933 mask_tok=118.00 y_norm=48.0922±0.0054 +2026-05-10 12:47:32,960 - INFO - Step 60 | mode=video loss=0.186523 diff=0.186523 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1066 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0922 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=49 i_micro=11 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3467 mask_tok=104.00 y_norm=48.0922±0.0060 +2026-05-10 12:47:50,928 - INFO - Step 70 | mode=video loss=0.114746 diff=0.114746 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1068 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0910 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=57 i_micro=13 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.5033 mask_tok=151.00 y_norm=48.0910±0.0065 +2026-05-10 12:48:10,447 - INFO - Step 80 | mode=video loss=0.114258 diff=0.114258 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0983 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0921 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=65 i_micro=15 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3800 mask_tok=114.00 y_norm=48.0921±0.0056 +2026-05-10 12:48:28,400 - INFO - Step 90 | mode=video loss=0.088379 diff=0.088379 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0555 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0896 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=73 i_micro=17 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4433 mask_tok=133.00 y_norm=48.0896±0.0066 +2026-05-10 12:48:46,359 - INFO - Step 100 | mode=video loss=0.083008 diff=0.083008 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0858 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0916 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=81 i_micro=19 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3333 mask_tok=100.00 y_norm=48.0916±0.0053 +/proj/cvl/users/x_fahkh2/envs/mobileov/lib/python3.10/site-packages/torch/distributed/fsdp/fully_sharded_data_parallel.py:690: FutureWarning: FSDP.state_dict_type() and FSDP.set_state_dict_type() are being deprecated. Please use APIs, get_state_dict() and set_state_dict(), which can support different parallelisms, FSDP1, FSDP2, DDP. API doc: https://pytorch.org/docs/stable/distributed.checkpoint.html#torch.distributed.checkpoint.state_dict.get_state_dict .Tutorial: https://pytorch.org/tutorials/recipes/distributed_checkpoint_recipe.html . + warnings.warn( +/proj/cvl/users/x_fahkh2/envs/mobileov/lib/python3.10/site-packages/torch/distributed/fsdp/fully_sharded_data_parallel.py:690: FutureWarning: FSDP.state_dict_type() and FSDP.set_state_dict_type() are being deprecated. Please use APIs, get_state_dict() and set_state_dict(), which can support different parallelisms, FSDP1, FSDP2, DDP. API doc: https://pytorch.org/docs/stable/distributed.checkpoint.html#torch.distributed.checkpoint.state_dict.get_state_dict .Tutorial: https://pytorch.org/tutorials/recipes/distributed_checkpoint_recipe.html . + warnings.warn( +/proj/cvl/users/x_fahkh2/envs/mobileov/lib/python3.10/site-packages/torch/distributed/fsdp/fully_sharded_data_parallel.py:690: FutureWarning: FSDP.state_dict_type() and FSDP.set_state_dict_type() are being deprecated. Please use APIs, get_state_dict() and set_state_dict(), which can support different parallelisms, FSDP1, FSDP2, DDP. API doc: https://pytorch.org/docs/stable/distributed.checkpoint.html#torch.distributed.checkpoint.state_dict.get_state_dict .Tutorial: https://pytorch.org/tutorials/recipes/distributed_checkpoint_recipe.html . + warnings.warn( +/proj/cvl/users/x_fahkh2/envs/mobileov/lib/python3.10/site-packages/torch/distributed/fsdp/fully_sharded_data_parallel.py:690: FutureWarning: FSDP.state_dict_type() and FSDP.set_state_dict_type() are being deprecated. Please use APIs, get_state_dict() and set_state_dict(), which can support different parallelisms, FSDP1, FSDP2, DDP. API doc: https://pytorch.org/docs/stable/distributed.checkpoint.html#torch.distributed.checkpoint.state_dict.get_state_dict .Tutorial: https://pytorch.org/tutorials/recipes/distributed_checkpoint_recipe.html . + warnings.warn( +/proj/cvl/users/x_fahkh2/envs/mobileov/lib/python3.10/site-packages/torch/distributed/fsdp/fully_sharded_data_parallel.py:690: FutureWarning: FSDP.state_dict_type() and FSDP.set_state_dict_type() are being deprecated. Please use APIs, get_state_dict() and set_state_dict(), which can support different parallelisms, FSDP1, FSDP2, DDP. API doc: https://pytorch.org/docs/stable/distributed.checkpoint.html#torch.distributed.checkpoint.state_dict.get_state_dict .Tutorial: https://pytorch.org/tutorials/recipes/distributed_checkpoint_recipe.html . + warnings.warn( +/proj/cvl/users/x_fahkh2/envs/mobileov/lib/python3.10/site-packages/torch/distributed/fsdp/fully_sharded_data_parallel.py:690: FutureWarning: FSDP.state_dict_type() and FSDP.set_state_dict_type() are being deprecated. Please use APIs, get_state_dict() and set_state_dict(), which can support different parallelisms, FSDP1, FSDP2, DDP. API doc: https://pytorch.org/docs/stable/distributed.checkpoint.html#torch.distributed.checkpoint.state_dict.get_state_dict .Tutorial: https://pytorch.org/tutorials/recipes/distributed_checkpoint_recipe.html . + warnings.warn( +/proj/cvl/users/x_fahkh2/envs/mobileov/lib/python3.10/site-packages/torch/distributed/fsdp/fully_sharded_data_parallel.py:690: FutureWarning: FSDP.state_dict_type() and FSDP.set_state_dict_type() are being deprecated. Please use APIs, get_state_dict() and set_state_dict(), which can support different parallelisms, FSDP1, FSDP2, DDP. API doc: https://pytorch.org/docs/stable/distributed.checkpoint.html#torch.distributed.checkpoint.state_dict.get_state_dict .Tutorial: https://pytorch.org/tutorials/recipes/distributed_checkpoint_recipe.html . + warnings.warn( +/proj/cvl/users/x_fahkh2/envs/mobileov/lib/python3.10/site-packages/torch/distributed/fsdp/fully_sharded_data_parallel.py:690: FutureWarning: FSDP.state_dict_type() and FSDP.set_state_dict_type() are being deprecated. Please use APIs, get_state_dict() and set_state_dict(), which can support different parallelisms, FSDP1, FSDP2, DDP. API doc: https://pytorch.org/docs/stable/distributed.checkpoint.html#torch.distributed.checkpoint.state_dict.get_state_dict .Tutorial: https://pytorch.org/tutorials/recipes/distributed_checkpoint_recipe.html . + warnings.warn( +2026-05-10 12:48:53,945 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 12:49:13,476 - INFO - Step 110 | mode=video loss=0.097656 diff=0.097656 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0688 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0874 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=89 i_micro=21 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3967 mask_tok=119.00 y_norm=48.0874±0.0098 +2026-05-10 12:49:31,421 - INFO - Step 120 | mode=video loss=0.090332 diff=0.090332 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1862 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0888 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=97 i_micro=23 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3700 mask_tok=111.00 y_norm=48.0888±0.0060 +2026-05-10 12:49:50,886 - INFO - Step 130 | mode=video loss=0.108887 diff=0.108887 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0702 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0907 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=105 i_micro=25 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.5000 mask_tok=150.00 y_norm=48.0907±0.0068 +2026-05-10 12:50:08,839 - INFO - Step 140 | mode=video loss=0.202148 diff=0.202148 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.3454 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0913 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=113 i_micro=27 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3933 mask_tok=118.00 y_norm=48.0913±0.0061 +2026-05-10 12:50:26,807 - INFO - Step 150 | mode=video loss=0.168945 diff=0.168945 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0617 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0877 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=121 i_micro=29 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4267 mask_tok=128.00 y_norm=48.0877±0.0061 +2026-05-10 12:50:46,208 - INFO - Step 160 | mode=video loss=0.175781 diff=0.175781 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0758 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0898 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=129 i_micro=31 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3667 mask_tok=110.00 y_norm=48.0898±0.0060 +2026-05-10 12:51:04,170 - INFO - Step 170 | mode=video loss=0.075195 diff=0.075195 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0670 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0914 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=137 i_micro=33 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3700 mask_tok=111.00 y_norm=48.0914±0.0071 +2026-05-10 12:51:22,878 - INFO - Step 180 | mode=video loss=0.134766 diff=0.134766 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0346 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0890 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=145 i_micro=35 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4133 mask_tok=124.00 y_norm=48.0890±0.0061 +2026-05-10 12:51:41,487 - INFO - Step 190 | mode=video loss=0.082031 diff=0.082031 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0481 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0911 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=153 i_micro=37 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.2833 mask_tok=85.00 y_norm=48.0911±0.0053 +2026-05-10 12:52:01,075 - INFO - Step 200 | mode=video loss=0.081055 diff=0.081055 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0519 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0886 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=161 i_micro=39 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3600 mask_tok=108.00 y_norm=48.0886±0.0059 cond_uncond_dloss=0.312500 cond_grad=0.000116 cond_pred_l2=35.554745 cond_pred_ratio=0.022403 +2026-05-10 12:52:01,172 - INFO - Step 200 | probe_semantic mcp_offdiag(mean/min/max)=0.646128/0.532540/0.711444 smol_offdiag=0.660557 prompts=4 mcp_tok=17.50 smol_tok=17.50 +2026-05-10 12:52:08,659 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 12:52:28,125 - INFO - Step 210 | mode=video loss=0.192383 diff=0.192383 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1160 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0911 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=169 i_micro=41 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3600 mask_tok=108.00 y_norm=48.0911±0.0057 +2026-05-10 12:52:46,089 - INFO - Step 220 | mode=video loss=0.083008 diff=0.083008 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0479 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0910 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=177 i_micro=43 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3667 mask_tok=110.00 y_norm=48.0910±0.0058 +2026-05-10 12:53:04,064 - INFO - Step 230 | mode=video loss=0.048096 diff=0.048096 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0464 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0906 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=185 i_micro=45 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3400 mask_tok=102.00 y_norm=48.0906±0.0060 +2026-05-10 12:53:23,391 - INFO - Step 240 | mode=video loss=0.143555 diff=0.143555 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0728 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0928 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=193 i_micro=47 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3733 mask_tok=112.00 y_norm=48.0928±0.0054 +2026-05-10 12:53:41,353 - INFO - Step 250 | mode=video loss=0.068848 diff=0.068848 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0843 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0881 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=201 i_micro=49 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4200 mask_tok=126.00 y_norm=48.0881±0.0061 +2026-05-10 12:54:00,800 - INFO - Step 260 | mode=video loss=0.104004 diff=0.104004 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1222 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0894 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=209 i_micro=51 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4500 mask_tok=135.00 y_norm=48.0894±0.0061 +2026-05-10 12:54:18,758 - INFO - Step 270 | mode=video loss=0.121094 diff=0.121094 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0292 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0890 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=217 i_micro=53 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4800 mask_tok=144.00 y_norm=48.0890±0.0069 +2026-05-10 12:54:36,716 - INFO - Step 280 | mode=video loss=0.081055 diff=0.081055 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0793 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0931 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=225 i_micro=55 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3500 mask_tok=105.00 y_norm=48.0931±0.0053 +2026-05-10 12:54:56,024 - INFO - Step 290 | mode=video loss=0.091309 diff=0.091309 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0789 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0936 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=233 i_micro=57 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3667 mask_tok=110.00 y_norm=48.0936±0.0059 +2026-05-10 12:55:13,985 - INFO - Step 300 | mode=video loss=0.097656 diff=0.097656 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0561 emb_mean=0.0008 emb_std=1.0020 tok_norm=48.0938 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=241 i_micro=59 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3400 mask_tok=102.00 y_norm=48.0938±0.0059 +2026-05-10 12:55:21,142 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 12:55:39,858 - INFO - Step 310 | mode=video loss=0.077148 diff=0.077148 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1233 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0924 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=249 i_micro=61 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3567 mask_tok=107.00 y_norm=48.0924±0.0059 +2026-05-10 12:55:58,608 - INFO - Step 320 | mode=video loss=0.125977 diff=0.125977 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0972 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0893 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=257 i_micro=63 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4100 mask_tok=123.00 y_norm=48.0893±0.0058 +2026-05-10 12:56:16,573 - INFO - Step 330 | mode=video loss=0.111328 diff=0.111328 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0322 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0907 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=265 i_micro=65 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4200 mask_tok=126.00 y_norm=48.0907±0.0061 +2026-05-10 12:56:35,851 - INFO - Step 340 | mode=video loss=0.104492 diff=0.104492 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0641 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0934 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=273 i_micro=67 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.1900 mask_tok=57.00 y_norm=48.0934±0.0046 +2026-05-10 12:56:53,822 - INFO - Step 350 | mode=video loss=0.171875 diff=0.171875 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0747 emb_mean=0.0008 emb_std=1.0020 tok_norm=48.0947 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=281 i_micro=69 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3667 mask_tok=110.00 y_norm=48.0947±0.0056 +2026-05-10 12:57:11,787 - INFO - Step 360 | mode=video loss=0.155273 diff=0.155273 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0371 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0909 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=289 i_micro=71 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4000 mask_tok=120.00 y_norm=48.0909±0.0061 +2026-05-10 12:57:31,263 - INFO - Step 370 | mode=video loss=0.076172 diff=0.076172 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0863 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0898 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=297 i_micro=73 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4100 mask_tok=123.00 y_norm=48.0898±0.0058 +2026-05-10 12:57:49,224 - INFO - Step 380 | mode=video loss=0.095215 diff=0.095215 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0560 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0915 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=305 i_micro=75 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3400 mask_tok=102.00 y_norm=48.0915±0.0059 +2026-05-10 12:58:08,325 - INFO - Step 390 | mode=video loss=0.064941 diff=0.064941 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0845 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0893 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=313 i_micro=77 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4433 mask_tok=133.00 y_norm=48.0893±0.0065 +2026-05-10 12:58:27,917 - INFO - Step 400 | mode=video loss=0.205078 diff=0.205078 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0473 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0886 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=321 i_micro=79 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4100 mask_tok=123.00 y_norm=48.0886±0.0060 cond_uncond_dloss=0.306641 cond_grad=0.000247 cond_pred_l2=36.795620 cond_pred_ratio=0.020859 +2026-05-10 12:58:27,955 - INFO - Step 400 | probe_semantic mcp_offdiag(mean/min/max)=0.650840/0.539349/0.713905 smol_offdiag=0.660557 prompts=4 mcp_tok=17.50 smol_tok=17.50 +2026-05-10 12:58:34,827 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 12:58:52,783 - INFO - Step 410 | mode=video loss=0.094727 diff=0.094727 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0453 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0907 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=329 i_micro=81 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3400 mask_tok=102.00 y_norm=48.0907±0.0066 +2026-05-10 12:59:12,302 - INFO - Step 420 | mode=video loss=0.074707 diff=0.074707 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0352 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0930 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=337 i_micro=83 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3800 mask_tok=114.00 y_norm=48.0930±0.0059 +2026-05-10 12:59:30,267 - INFO - Step 430 | mode=video loss=0.105957 diff=0.105957 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0359 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0925 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=345 i_micro=85 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3567 mask_tok=107.00 y_norm=48.0925±0.0055 +2026-05-10 12:59:48,665 - INFO - Step 440 | mode=video loss=0.122559 diff=0.122559 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0824 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0923 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=353 i_micro=87 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3533 mask_tok=106.00 y_norm=48.0923±0.0053 +2026-05-10 13:00:07,818 - INFO - Step 450 | mode=video loss=0.083008 diff=0.083008 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0914 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0908 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=361 i_micro=89 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3700 mask_tok=111.00 y_norm=48.0908±0.0054 +2026-05-10 13:00:25,821 - INFO - Step 460 | mode=video loss=0.123047 diff=0.123047 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1300 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0903 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=369 i_micro=91 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3333 mask_tok=100.00 y_norm=48.0903±0.0055 +2026-05-10 13:00:44,984 - INFO - Step 470 | mode=video loss=0.090820 diff=0.090820 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0452 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0932 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=377 i_micro=93 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.2467 mask_tok=74.00 y_norm=48.0932±0.0052 +2026-05-10 13:01:02,972 - INFO - Step 480 | mode=video loss=0.060547 diff=0.060547 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0296 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0928 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=385 i_micro=95 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3567 mask_tok=107.00 y_norm=48.0928±0.0056 +2026-05-10 13:01:20,951 - INFO - Step 490 | mode=video loss=0.106934 diff=0.106934 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0468 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0921 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=393 i_micro=97 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3467 mask_tok=104.00 y_norm=48.0921±0.0058 +2026-05-10 13:01:40,279 - INFO - Step 500 | mode=video loss=0.065430 diff=0.065430 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0389 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0912 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=401 i_micro=99 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3500 mask_tok=105.00 y_norm=48.0912±0.0067 +2026-05-10 13:01:47,319 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 13:02:05,302 - INFO - Step 510 | mode=video loss=0.144531 diff=0.144531 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0704 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0882 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=409 i_micro=101 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4433 mask_tok=133.00 y_norm=48.0882±0.0065 +2026-05-10 13:02:24,573 - INFO - Step 520 | mode=video loss=0.079102 diff=0.079102 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0718 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0921 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=417 i_micro=103 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3733 mask_tok=112.00 y_norm=48.0921±0.0067 +2026-05-10 13:02:42,576 - INFO - Step 530 | mode=video loss=0.161133 diff=0.161133 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0915 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0898 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=425 i_micro=105 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3767 mask_tok=113.00 y_norm=48.0898±0.0064 +2026-05-10 13:03:00,586 - INFO - Step 540 | mode=video loss=0.086426 diff=0.086426 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0381 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0903 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=433 i_micro=107 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3767 mask_tok=113.00 y_norm=48.0903±0.0072 +2026-05-10 13:03:20,015 - INFO - Step 550 | mode=video loss=0.091309 diff=0.091309 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0585 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0915 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=441 i_micro=109 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3333 mask_tok=100.00 y_norm=48.0915±0.0059 +2026-05-10 13:03:38,011 - INFO - Step 560 | mode=video loss=0.125000 diff=0.125000 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0422 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0895 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=449 i_micro=111 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4333 mask_tok=130.00 y_norm=48.0895±0.0056 +2026-05-10 13:03:57,481 - INFO - Step 570 | mode=video loss=0.063477 diff=0.063477 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0383 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0920 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=457 i_micro=113 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.8167 mask_tok=245.00 y_norm=48.0920±0.0065 +2026-05-10 13:04:15,420 - INFO - Step 580 | mode=video loss=0.104004 diff=0.104004 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0512 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0912 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=465 i_micro=115 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4300 mask_tok=129.00 y_norm=48.0912±0.0069 +2026-05-10 13:04:33,374 - INFO - Step 590 | mode=video loss=0.111328 diff=0.111328 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0567 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0897 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=473 i_micro=117 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3933 mask_tok=118.00 y_norm=48.0897±0.0055 +2026-05-10 13:04:54,336 - INFO - Step 600 | mode=video loss=0.058105 diff=0.058105 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0763 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0890 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=481 i_micro=119 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.5100 mask_tok=153.00 y_norm=48.0890±0.0061 cond_uncond_dloss=1.414062 cond_grad=0.000067 cond_pred_l2=37.635181 cond_pred_ratio=0.021345 +2026-05-10 13:04:54,374 - INFO - Step 600 | probe_semantic mcp_offdiag(mean/min/max)=0.654644/0.544468/0.718473 smol_offdiag=0.660557 prompts=4 mcp_tok=17.50 smol_tok=17.50 +2026-05-10 13:05:01,257 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 13:05:19,206 - INFO - Step 610 | mode=video loss=0.097168 diff=0.097168 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0435 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0892 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=489 i_micro=121 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3533 mask_tok=106.00 y_norm=48.0892±0.0056 +2026-05-10 13:05:37,169 - INFO - Step 620 | mode=video loss=0.122559 diff=0.122559 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0447 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0883 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=497 i_micro=123 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4033 mask_tok=121.00 y_norm=48.0883±0.0055 +2026-05-10 13:05:56,189 - INFO - Step 630 | mode=video loss=0.116699 diff=0.116699 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0789 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0898 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=505 i_micro=125 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3633 mask_tok=109.00 y_norm=48.0898±0.0058 +2026-05-10 13:06:14,157 - INFO - Step 640 | mode=video loss=0.078613 diff=0.078613 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0767 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0895 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=513 i_micro=127 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3500 mask_tok=105.00 y_norm=48.0895±0.0060 +2026-05-10 13:06:33,273 - INFO - Step 650 | mode=video loss=0.086914 diff=0.086914 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0578 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0911 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=521 i_micro=129 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3700 mask_tok=111.00 y_norm=48.0911±0.0068 +2026-05-10 13:06:51,226 - INFO - Step 660 | mode=video loss=0.235352 diff=0.235352 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0629 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0908 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=529 i_micro=131 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3300 mask_tok=99.00 y_norm=48.0908±0.0056 +2026-05-10 13:07:09,184 - INFO - Step 670 | mode=video loss=0.145508 diff=0.145508 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0403 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0884 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=537 i_micro=133 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4533 mask_tok=136.00 y_norm=48.0884±0.0064 +2026-05-10 13:07:28,322 - INFO - Step 680 | mode=video loss=0.122559 diff=0.122559 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1161 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0902 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=545 i_micro=135 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4733 mask_tok=142.00 y_norm=48.0902±0.0063 +2026-05-10 13:07:46,270 - INFO - Step 690 | mode=video loss=0.160156 diff=0.160156 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0582 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0905 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=553 i_micro=137 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3900 mask_tok=117.00 y_norm=48.0905±0.0058 +2026-05-10 13:08:04,929 - INFO - Step 700 | mode=video loss=0.125000 diff=0.125000 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0694 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0900 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=561 i_micro=139 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4500 mask_tok=135.00 y_norm=48.0900±0.0063 +2026-05-10 13:08:11,962 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 13:08:29,902 - INFO - Step 710 | mode=video loss=0.066895 diff=0.066895 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0602 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0924 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=569 i_micro=141 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3367 mask_tok=101.00 y_norm=48.0924±0.0057 +2026-05-10 13:08:47,864 - INFO - Step 720 | mode=video loss=0.080078 diff=0.080078 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0622 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0923 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=577 i_micro=143 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.2667 mask_tok=80.00 y_norm=48.0923±0.0049 +2026-05-10 13:09:06,982 - INFO - Step 730 | mode=video loss=0.107422 diff=0.107422 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0636 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0883 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=585 i_micro=145 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4867 mask_tok=146.00 y_norm=48.0883±0.0067 +2026-05-10 13:09:24,948 - INFO - Step 740 | mode=video loss=0.119141 diff=0.119141 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0983 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0920 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=593 i_micro=147 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3433 mask_tok=103.00 y_norm=48.0920±0.0049 +2026-05-10 13:09:43,623 - INFO - Step 750 | mode=video loss=0.079102 diff=0.079102 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0482 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0907 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=601 i_micro=149 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4900 mask_tok=147.00 y_norm=48.0907±0.0070 +2026-05-10 13:10:02,347 - INFO - Step 760 | mode=video loss=0.049561 diff=0.049561 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0296 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0904 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=609 i_micro=151 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4033 mask_tok=121.00 y_norm=48.0904±0.0064 +2026-05-10 13:10:20,307 - INFO - Step 770 | mode=video loss=0.100586 diff=0.100586 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0672 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0895 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=617 i_micro=153 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.5200 mask_tok=156.00 y_norm=48.0895±0.0072 +2026-05-10 13:10:39,743 - INFO - Step 780 | mode=video loss=0.093750 diff=0.093750 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0396 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0891 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=625 i_micro=155 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4267 mask_tok=128.00 y_norm=48.0891±0.0054 +2026-05-10 13:10:57,727 - INFO - Step 790 | mode=video loss=0.139648 diff=0.139648 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1738 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0915 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=633 i_micro=157 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3600 mask_tok=108.00 y_norm=48.0915±0.0052 +2026-05-10 13:11:17,338 - INFO - Step 800 | mode=video loss=0.044434 diff=0.044434 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0316 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0887 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=641 i_micro=159 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4067 mask_tok=122.00 y_norm=48.0887±0.0056 cond_uncond_dloss=0.140625 cond_grad=0.000050 cond_pred_l2=31.283968 cond_pred_ratio=0.017478 +2026-05-10 13:11:17,375 - INFO - Step 800 | probe_semantic mcp_offdiag(mean/min/max)=0.653730/0.544577/0.716086 smol_offdiag=0.660557 prompts=4 mcp_tok=17.50 smol_tok=17.50 +2026-05-10 13:11:24,505 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 13:11:44,039 - INFO - Step 810 | mode=video loss=0.105469 diff=0.105469 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0444 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0923 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=649 i_micro=161 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3067 mask_tok=92.00 y_norm=48.0923±0.0051 +2026-05-10 13:12:02,005 - INFO - Step 820 | mode=video loss=0.220703 diff=0.220703 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0574 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0891 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=657 i_micro=163 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4367 mask_tok=131.00 y_norm=48.0891±0.0065 +2026-05-10 13:12:21,468 - INFO - Step 830 | mode=video loss=0.167969 diff=0.167969 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0803 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0914 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=665 i_micro=165 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.2867 mask_tok=86.00 y_norm=48.0914±0.0050 +2026-05-10 13:12:39,447 - INFO - Step 840 | mode=video loss=0.086914 diff=0.086914 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0579 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0895 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=673 i_micro=167 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4400 mask_tok=132.00 y_norm=48.0895±0.0063 +2026-05-10 13:12:57,431 - INFO - Step 850 | mode=video loss=0.125000 diff=0.125000 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0754 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0901 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=681 i_micro=169 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3500 mask_tok=105.00 y_norm=48.0901±0.0057 +2026-05-10 13:13:16,680 - INFO - Step 860 | mode=video loss=0.122070 diff=0.122070 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0417 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0898 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=689 i_micro=171 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4767 mask_tok=143.00 y_norm=48.0898±0.0066 +2026-05-10 13:13:34,655 - INFO - Step 870 | mode=video loss=0.257812 diff=0.257812 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.3066 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0913 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=697 i_micro=173 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3367 mask_tok=101.00 y_norm=48.0913±0.0055 +2026-05-10 13:13:53,449 - INFO - Step 880 | mode=video loss=0.267578 diff=0.267578 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0421 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0880 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=705 i_micro=175 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3667 mask_tok=110.00 y_norm=48.0880±0.0061 +2026-05-10 13:14:12,008 - INFO - Step 890 | mode=video loss=0.048096 diff=0.048096 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0340 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0934 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=713 i_micro=177 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3500 mask_tok=105.00 y_norm=48.0934±0.0051 +2026-05-10 13:14:29,993 - INFO - Step 900 | mode=video loss=0.154297 diff=0.154297 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0622 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0917 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=721 i_micro=179 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3433 mask_tok=103.00 y_norm=48.0917±0.0053 +2026-05-10 13:14:36,788 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 13:14:55,991 - INFO - Step 910 | mode=video loss=0.078125 diff=0.078125 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0328 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0907 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=729 i_micro=181 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3500 mask_tok=105.00 y_norm=48.0907±0.0056 +2026-05-10 13:15:13,948 - INFO - Step 920 | mode=video loss=0.106445 diff=0.106445 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0382 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0912 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=737 i_micro=183 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3600 mask_tok=108.00 y_norm=48.0912±0.0062 +2026-05-10 13:15:32,511 - INFO - Step 930 | mode=video loss=0.099121 diff=0.099121 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0474 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0884 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=745 i_micro=185 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4667 mask_tok=140.00 y_norm=48.0884±0.0058 +2026-05-10 13:15:51,105 - INFO - Step 940 | mode=video loss=0.097168 diff=0.097168 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0328 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0889 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=753 i_micro=187 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3767 mask_tok=113.00 y_norm=48.0889±0.0058 +2026-05-10 13:16:09,067 - INFO - Step 950 | mode=video loss=0.113770 diff=0.113770 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0659 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0907 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=761 i_micro=189 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3533 mask_tok=106.00 y_norm=48.0907±0.0064 +2026-05-10 13:16:28,448 - INFO - Step 960 | mode=video loss=0.108887 diff=0.108887 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.3339 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0921 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=769 i_micro=191 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4100 mask_tok=123.00 y_norm=48.0921±0.0056 +2026-05-10 13:16:46,406 - INFO - Step 970 | mode=video loss=0.070801 diff=0.070801 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0550 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0903 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=777 i_micro=193 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4900 mask_tok=147.00 y_norm=48.0903±0.0069 +2026-05-10 13:17:04,365 - INFO - Step 980 | mode=video loss=0.104492 diff=0.104492 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0440 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0913 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=785 i_micro=195 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3233 mask_tok=97.00 y_norm=48.0913±0.0050 +2026-05-10 13:17:23,674 - INFO - Step 990 | mode=video loss=0.074219 diff=0.074219 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0491 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0903 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=793 i_micro=197 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4167 mask_tok=125.00 y_norm=48.0903±0.0062 +2026-05-10 13:17:43,245 - INFO - Step 1000 | mode=video loss=0.112305 diff=0.112305 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0653 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0903 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=801 i_micro=199 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3467 mask_tok=104.00 y_norm=48.0903±0.0062 cond_uncond_dloss=0.386719 cond_grad=0.000171 cond_pred_l2=41.111095 cond_pred_ratio=0.026038 +2026-05-10 13:17:43,283 - INFO - Step 1000 | probe_semantic mcp_offdiag(mean/min/max)=0.655720/0.545695/0.717087 smol_offdiag=0.660557 prompts=4 mcp_tok=17.50 smol_tok=17.50 +2026-05-10 13:17:50,170 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 13:18:08,948 - INFO - Step 1010 | mode=video loss=0.077148 diff=0.077148 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0488 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0913 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=809 i_micro=201 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3433 mask_tok=103.00 y_norm=48.0913±0.0052 +2026-05-10 13:18:27,644 - INFO - Step 1020 | mode=video loss=0.084473 diff=0.084473 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0544 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0901 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=817 i_micro=203 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3733 mask_tok=112.00 y_norm=48.0901±0.0056 +2026-05-10 13:18:45,621 - INFO - Step 1030 | mode=video loss=0.139648 diff=0.139648 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0632 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0901 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=825 i_micro=205 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3700 mask_tok=111.00 y_norm=48.0901±0.0054 +2026-05-10 13:19:04,787 - INFO - Step 1040 | mode=video loss=0.079590 diff=0.079590 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0295 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0910 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=833 i_micro=207 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3000 mask_tok=90.00 y_norm=48.0910±0.0044 +2026-05-10 13:19:22,736 - INFO - Step 1050 | mode=video loss=0.229492 diff=0.229492 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.4080 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0908 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=841 i_micro=209 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3567 mask_tok=107.00 y_norm=48.0908±0.0065 +2026-05-10 13:19:41,550 - INFO - Step 1060 | mode=video loss=0.119141 diff=0.119141 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0947 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0887 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=849 i_micro=211 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3933 mask_tok=118.00 y_norm=48.0887±0.0066 +2026-05-10 13:19:59,922 - INFO - Step 1070 | mode=video loss=0.069336 diff=0.069336 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0398 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0899 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=857 i_micro=213 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3500 mask_tok=105.00 y_norm=48.0899±0.0064 +2026-05-10 13:20:17,864 - INFO - Step 1080 | mode=video loss=0.161133 diff=0.161133 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1787 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0903 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=865 i_micro=215 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3367 mask_tok=101.00 y_norm=48.0903±0.0063 +2026-05-10 13:20:37,008 - INFO - Step 1090 | mode=video loss=0.092773 diff=0.092773 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0379 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0894 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=873 i_micro=217 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4433 mask_tok=133.00 y_norm=48.0894±0.0060 +2026-05-10 13:20:54,964 - INFO - Step 1100 | mode=video loss=0.084961 diff=0.084961 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0621 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0871 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=881 i_micro=219 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4867 mask_tok=146.00 y_norm=48.0871±0.0057 +2026-05-10 13:21:01,880 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 13:21:20,605 - INFO - Step 1110 | mode=video loss=0.072754 diff=0.072754 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0451 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0903 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=889 i_micro=221 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3333 mask_tok=100.00 y_norm=48.0903±0.0061 +2026-05-10 13:21:39,135 - INFO - Step 1120 | mode=video loss=0.096680 diff=0.096680 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1309 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0892 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=897 i_micro=223 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4933 mask_tok=148.00 y_norm=48.0892±0.0064 +2026-05-10 13:21:57,103 - INFO - Step 1130 | mode=video loss=0.077637 diff=0.077637 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0309 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0913 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=905 i_micro=225 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3500 mask_tok=105.00 y_norm=48.0913±0.0053 +2026-05-10 13:22:15,905 - INFO - Step 1140 | mode=video loss=0.100586 diff=0.100586 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0545 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0897 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=913 i_micro=227 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3533 mask_tok=106.00 y_norm=48.0897±0.0059 +2026-05-10 13:22:34,607 - INFO - Step 1150 | mode=video loss=0.099121 diff=0.099121 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0504 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0899 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=921 i_micro=229 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3467 mask_tok=104.00 y_norm=48.0899±0.0053 +2026-05-10 13:22:52,577 - INFO - Step 1160 | mode=video loss=0.058594 diff=0.058594 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0736 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0880 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=929 i_micro=231 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.5067 mask_tok=152.00 y_norm=48.0880±0.0059 +2026-05-10 13:23:12,077 - INFO - Step 1170 | mode=video loss=0.062500 diff=0.062500 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0465 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0880 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=937 i_micro=233 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4667 mask_tok=140.00 y_norm=48.0880±0.0057 +2026-05-10 13:23:30,049 - INFO - Step 1180 | mode=video loss=0.086914 diff=0.086914 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0218 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0877 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=945 i_micro=235 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4800 mask_tok=144.00 y_norm=48.0877±0.0056 +2026-05-10 13:23:48,847 - INFO - Step 1190 | mode=video loss=0.158203 diff=0.158203 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0626 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0898 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=953 i_micro=237 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3233 mask_tok=97.00 y_norm=48.0898±0.0056 +2026-05-10 13:24:09,095 - INFO - Step 1200 | mode=video loss=0.076172 diff=0.076172 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0297 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0877 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=961 i_micro=239 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4467 mask_tok=134.00 y_norm=48.0877±0.0069 cond_uncond_dloss=0.250000 cond_grad=0.000031 cond_pred_l2=28.794571 cond_pred_ratio=0.016103 +2026-05-10 13:24:09,132 - INFO - Step 1200 | probe_semantic mcp_offdiag(mean/min/max)=0.651338/0.541407/0.713142 smol_offdiag=0.660557 prompts=4 mcp_tok=17.50 smol_tok=17.50 +2026-05-10 13:24:16,002 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 13:24:33,962 - INFO - Step 1210 | mode=video loss=0.093262 diff=0.093262 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0585 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0893 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=969 i_micro=241 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3767 mask_tok=113.00 y_norm=48.0893±0.0057 +2026-05-10 13:24:53,447 - INFO - Step 1220 | mode=video loss=0.164062 diff=0.164062 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0740 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0886 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=977 i_micro=243 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3433 mask_tok=103.00 y_norm=48.0886±0.0054 +2026-05-10 13:25:11,407 - INFO - Step 1230 | mode=video loss=0.086426 diff=0.086426 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0813 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0889 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=985 i_micro=245 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3600 mask_tok=108.00 y_norm=48.0889±0.0057 +2026-05-10 13:25:30,115 - INFO - Step 1240 | mode=video loss=0.092285 diff=0.092285 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0383 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0894 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=993 i_micro=247 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3867 mask_tok=116.00 y_norm=48.0894±0.0051 +2026-05-10 13:25:48,633 - INFO - Step 1250 | mode=video loss=0.111816 diff=0.111816 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0377 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0897 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=1001 i_micro=249 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3400 mask_tok=102.00 y_norm=48.0897±0.0059 +2026-05-10 13:26:06,603 - INFO - Step 1260 | mode=video loss=0.080078 diff=0.080078 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0677 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0892 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=1009 i_micro=251 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3600 mask_tok=108.00 y_norm=48.0892±0.0062 +2026-05-10 13:26:25,249 - INFO - Step 1270 | mode=video loss=0.141602 diff=0.141602 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0686 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0907 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=1017 i_micro=253 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3500 mask_tok=105.00 y_norm=48.0907±0.0056 +2026-05-10 13:26:43,755 - INFO - Step 1280 | mode=video loss=0.087891 diff=0.087891 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0559 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0893 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=1025 i_micro=255 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3367 mask_tok=101.00 y_norm=48.0893±0.0055 +2026-05-10 13:27:02,312 - INFO - Step 1290 | mode=video loss=0.109863 diff=0.109863 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0519 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0890 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=1033 i_micro=257 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3900 mask_tok=117.00 y_norm=48.0890±0.0056 +2026-05-10 13:27:20,693 - INFO - Step 1300 | mode=video loss=0.087402 diff=0.087402 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0414 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0856 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=1041 i_micro=259 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4100 mask_tok=123.00 y_norm=48.0856±0.0057 +2026-05-10 13:27:27,559 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 13:27:45,503 - INFO - Step 1310 | mode=video loss=0.077637 diff=0.077637 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0694 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0870 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=1049 i_micro=261 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3600 mask_tok=108.00 y_norm=48.0870±0.0062 +2026-05-10 13:28:04,175 - INFO - Step 1320 | mode=video loss=0.091797 diff=0.091797 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0491 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0883 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=1057 i_micro=263 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3500 mask_tok=105.00 y_norm=48.0883±0.0058 +2026-05-10 13:28:22,565 - INFO - Step 1330 | mode=video loss=0.058594 diff=0.058594 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0266 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0885 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=1065 i_micro=265 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3967 mask_tok=119.00 y_norm=48.0885±0.0059 +2026-05-10 13:28:40,518 - INFO - Step 1340 | mode=video loss=0.069824 diff=0.069824 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0832 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0893 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=1073 i_micro=267 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3467 mask_tok=104.00 y_norm=48.0893±0.0052 +2026-05-10 13:28:59,620 - INFO - Step 1350 | mode=video loss=0.072266 diff=0.072266 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0276 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0910 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=1081 i_micro=269 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3933 mask_tok=118.00 y_norm=48.0910±0.0059 +2026-05-10 13:29:17,573 - INFO - Step 1360 | mode=video loss=0.059082 diff=0.059082 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0294 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0903 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=1089 i_micro=271 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3433 mask_tok=103.00 y_norm=48.0903±0.0053 +2026-05-10 13:29:36,254 - INFO - Step 1370 | mode=video loss=0.083008 diff=0.083008 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0310 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0892 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=1097 i_micro=273 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3167 mask_tok=95.00 y_norm=48.0892±0.0052 +2026-05-10 13:29:54,911 - INFO - Step 1380 | mode=video loss=0.079102 diff=0.079102 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0388 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0884 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=1105 i_micro=275 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3233 mask_tok=97.00 y_norm=48.0884±0.0054 +2026-05-10 13:30:12,867 - INFO - Step 1390 | mode=video loss=0.086426 diff=0.086426 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0290 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0885 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=1113 i_micro=277 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.5200 mask_tok=156.00 y_norm=48.0885±0.0066 +2026-05-10 13:30:33,209 - INFO - Step 1400 | mode=video loss=0.080566 diff=0.080566 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0374 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0870 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=1121 i_micro=279 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4533 mask_tok=136.00 y_norm=48.0870±0.0059 cond_uncond_dloss=0.183594 cond_grad=0.000078 cond_pred_l2=32.951401 cond_pred_ratio=0.018539 +2026-05-10 13:30:33,246 - INFO - Step 1400 | probe_semantic mcp_offdiag(mean/min/max)=0.650026/0.539708/0.712027 smol_offdiag=0.660557 prompts=4 mcp_tok=17.50 smol_tok=17.50 +2026-05-10 13:30:40,110 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 13:30:58,625 - INFO - Step 1410 | mode=video loss=0.121094 diff=0.121094 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0260 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0897 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=1129 i_micro=281 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3833 mask_tok=115.00 y_norm=48.0897±0.0057 +2026-05-10 13:31:17,460 - INFO - Step 1420 | mode=video loss=0.085938 diff=0.085938 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0308 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0877 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=1137 i_micro=283 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4467 mask_tok=134.00 y_norm=48.0877±0.0065 +2026-05-10 13:31:35,885 - INFO - Step 1430 | mode=video loss=0.229492 diff=0.229492 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1195 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0903 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=1145 i_micro=285 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3533 mask_tok=106.00 y_norm=48.0903±0.0058 +2026-05-10 13:31:53,848 - INFO - Step 1440 | mode=video loss=0.063965 diff=0.063965 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0783 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0904 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=1153 i_micro=287 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3467 mask_tok=104.00 y_norm=48.0904±0.0058 +2026-05-10 13:32:12,605 - INFO - Step 1450 | mode=video loss=0.061768 diff=0.061768 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0557 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0890 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=1161 i_micro=289 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3667 mask_tok=110.00 y_norm=48.0890±0.0057 +2026-05-10 13:32:30,985 - INFO - Step 1460 | mode=video loss=0.080078 diff=0.080078 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0384 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0913 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=1169 i_micro=291 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.8433 mask_tok=253.00 y_norm=48.0913±0.0076 +2026-05-10 13:32:48,951 - INFO - Step 1470 | mode=video loss=0.103027 diff=0.103027 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0260 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0900 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=1177 i_micro=293 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3567 mask_tok=107.00 y_norm=48.0900±0.0069 +2026-05-10 13:33:08,007 - INFO - Step 1480 | mode=video loss=0.078125 diff=0.078125 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0620 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0893 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=1185 i_micro=295 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3667 mask_tok=110.00 y_norm=48.0893±0.0055 +2026-05-10 13:33:25,981 - INFO - Step 1490 | mode=video loss=0.086914 diff=0.086914 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0405 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0922 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=1193 i_micro=297 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3967 mask_tok=119.00 y_norm=48.0922±0.0058 +2026-05-10 13:33:44,681 - INFO - Step 1500 | mode=video loss=0.179688 diff=0.179688 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1176 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0903 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=1201 i_micro=299 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.2733 mask_tok=82.00 y_norm=48.0903±0.0057 +2026-05-10 13:33:51,717 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 13:34:10,182 - INFO - Step 1510 | mode=video loss=0.083496 diff=0.083496 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0995 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0902 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=1209 i_micro=301 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3567 mask_tok=107.00 y_norm=48.0902±0.0055 +2026-05-10 13:34:28,141 - INFO - Step 1520 | mode=video loss=0.077148 diff=0.077148 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0279 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0891 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=1217 i_micro=303 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3267 mask_tok=98.00 y_norm=48.0891±0.0058 +2026-05-10 13:34:46,896 - INFO - Step 1530 | mode=video loss=0.069824 diff=0.069824 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0715 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0899 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=1225 i_micro=305 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3333 mask_tok=100.00 y_norm=48.0899±0.0052 +2026-05-10 13:35:05,505 - INFO - Step 1540 | mode=video loss=0.056641 diff=0.056641 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0696 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0882 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=1233 i_micro=307 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3267 mask_tok=98.00 y_norm=48.0882±0.0052 +2026-05-10 13:35:24,284 - INFO - Step 1550 | mode=video loss=0.089844 diff=0.089844 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0539 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0889 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=1241 i_micro=309 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3567 mask_tok=107.00 y_norm=48.0889±0.0056 +2026-05-10 13:35:42,943 - INFO - Step 1560 | mode=video loss=0.114746 diff=0.114746 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0398 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0885 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=1249 i_micro=311 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4000 mask_tok=120.00 y_norm=48.0885±0.0063 +2026-05-10 13:36:00,906 - INFO - Step 1570 | mode=video loss=0.100586 diff=0.100586 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0710 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0888 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=1257 i_micro=313 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3000 mask_tok=90.00 y_norm=48.0888±0.0053 +2026-05-10 13:36:19,673 - INFO - Step 1580 | mode=video loss=0.080566 diff=0.080566 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0543 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0892 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=1265 i_micro=315 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.2733 mask_tok=82.00 y_norm=48.0892±0.0053 +2026-05-10 13:36:38,274 - INFO - Step 1590 | mode=video loss=0.086914 diff=0.086914 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0579 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0864 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=1273 i_micro=317 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4600 mask_tok=138.00 y_norm=48.0864±0.0058 +2026-05-10 13:36:57,861 - INFO - Step 1600 | mode=video loss=0.087891 diff=0.087891 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0432 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0870 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=1281 i_micro=319 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4700 mask_tok=141.00 y_norm=48.0870±0.0065 cond_uncond_dloss=0.257812 cond_grad=0.000092 cond_pred_l2=39.192776 cond_pred_ratio=0.023605 +2026-05-10 13:36:57,899 - INFO - Step 1600 | probe_semantic mcp_offdiag(mean/min/max)=0.648131/0.536814/0.711246 smol_offdiag=0.660557 prompts=4 mcp_tok=17.50 smol_tok=17.50 +2026-05-10 13:37:04,835 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 13:37:24,338 - INFO - Step 1610 | mode=video loss=0.123535 diff=0.123535 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0353 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0866 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=1289 i_micro=321 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3567 mask_tok=107.00 y_norm=48.0866±0.0058 +2026-05-10 13:37:42,299 - INFO - Step 1620 | mode=video loss=0.159180 diff=0.159180 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0999 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0884 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=1297 i_micro=323 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3567 mask_tok=107.00 y_norm=48.0884±0.0055 +2026-05-10 13:38:00,997 - INFO - Step 1630 | mode=video loss=0.077148 diff=0.077148 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0712 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0852 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=1305 i_micro=325 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3467 mask_tok=104.00 y_norm=48.0852±0.0060 +2026-05-10 13:38:19,439 - INFO - Step 1640 | mode=video loss=0.123047 diff=0.123047 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0656 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0886 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=1313 i_micro=327 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3333 mask_tok=100.00 y_norm=48.0886±0.0056 +2026-05-10 13:38:37,399 - INFO - Step 1650 | mode=video loss=0.147461 diff=0.147461 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0612 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0882 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=1321 i_micro=329 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3333 mask_tok=100.00 y_norm=48.0882±0.0050 +2026-05-10 13:38:56,718 - INFO - Step 1660 | mode=video loss=0.070801 diff=0.070801 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0284 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0891 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=1329 i_micro=331 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3667 mask_tok=110.00 y_norm=48.0891±0.0059 +2026-05-10 13:39:14,674 - INFO - Step 1670 | mode=video loss=0.096191 diff=0.096191 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0474 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0864 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=1337 i_micro=333 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4400 mask_tok=132.00 y_norm=48.0864±0.0061 +2026-05-10 13:39:33,293 - INFO - Step 1680 | mode=video loss=0.126953 diff=0.126953 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0703 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0858 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=1345 i_micro=335 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4367 mask_tok=131.00 y_norm=48.0858±0.0058 +2026-05-10 13:39:51,801 - INFO - Step 1690 | mode=video loss=0.096680 diff=0.096680 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0418 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0894 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=1353 i_micro=337 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3600 mask_tok=108.00 y_norm=48.0894±0.0057 +2026-05-10 13:40:09,769 - INFO - Step 1700 | mode=video loss=0.118652 diff=0.118652 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0331 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0902 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=1361 i_micro=339 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3467 mask_tok=104.00 y_norm=48.0902±0.0064 +2026-05-10 13:40:16,740 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 13:40:35,380 - INFO - Step 1710 | mode=video loss=0.082520 diff=0.082520 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1093 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0879 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=1369 i_micro=341 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3667 mask_tok=110.00 y_norm=48.0879±0.0057 +2026-05-10 13:40:54,065 - INFO - Step 1720 | mode=video loss=0.104492 diff=0.104492 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0555 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0877 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=1377 i_micro=343 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.2633 mask_tok=79.00 y_norm=48.0877±0.0053 +2026-05-10 13:41:12,025 - INFO - Step 1730 | mode=video loss=0.155273 diff=0.155273 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0638 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0889 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=1385 i_micro=345 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3267 mask_tok=98.00 y_norm=48.0889±0.0053 +2026-05-10 13:41:31,473 - INFO - Step 1740 | mode=video loss=0.128906 diff=0.128906 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0966 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0876 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=1393 i_micro=347 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3400 mask_tok=102.00 y_norm=48.0876±0.0051 +2026-05-10 13:41:49,434 - INFO - Step 1750 | mode=video loss=0.074219 diff=0.074219 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0637 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0888 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=1401 i_micro=349 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3233 mask_tok=97.00 y_norm=48.0888±0.0058 +2026-05-10 13:42:08,007 - INFO - Step 1760 | mode=video loss=0.102539 diff=0.102539 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0286 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0869 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=1409 i_micro=351 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3433 mask_tok=103.00 y_norm=48.0869±0.0057 +2026-05-10 13:42:26,606 - INFO - Step 1770 | mode=video loss=0.156250 diff=0.156250 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0566 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0883 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=1417 i_micro=353 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3300 mask_tok=99.00 y_norm=48.0883±0.0061 +2026-05-10 13:42:44,554 - INFO - Step 1780 | mode=video loss=0.134766 diff=0.134766 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0383 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0863 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=1425 i_micro=355 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3900 mask_tok=117.00 y_norm=48.0863±0.0057 +2026-05-10 13:43:03,227 - INFO - Step 1790 | mode=video loss=0.070801 diff=0.070801 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0354 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0870 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=1433 i_micro=357 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3600 mask_tok=108.00 y_norm=48.0870±0.0058 +2026-05-10 13:43:23,288 - INFO - Step 1800 | mode=video loss=0.079590 diff=0.079590 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0392 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0883 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=1441 i_micro=359 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4133 mask_tok=124.00 y_norm=48.0883±0.0056 cond_uncond_dloss=0.330078 cond_grad=0.000101 cond_pred_l2=44.003990 cond_pred_ratio=0.026679 +2026-05-10 13:43:23,326 - INFO - Step 1800 | probe_semantic mcp_offdiag(mean/min/max)=0.646379/0.533452/0.710518 smol_offdiag=0.660557 prompts=4 mcp_tok=17.50 smol_tok=17.50 +2026-05-10 13:43:30,325 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 13:43:48,987 - INFO - Step 1810 | mode=video loss=0.081543 diff=0.081543 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0369 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0876 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=1449 i_micro=361 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3767 mask_tok=113.00 y_norm=48.0876±0.0057 +2026-05-10 13:44:07,494 - INFO - Step 1820 | mode=video loss=0.088379 diff=0.088379 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0373 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0875 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=1457 i_micro=363 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4100 mask_tok=123.00 y_norm=48.0875±0.0061 +2026-05-10 13:44:25,462 - INFO - Step 1830 | mode=video loss=0.086914 diff=0.086914 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0283 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0880 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=1465 i_micro=365 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3467 mask_tok=104.00 y_norm=48.0880±0.0057 +2026-05-10 13:44:44,121 - INFO - Step 1840 | mode=video loss=0.074219 diff=0.074219 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0563 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0874 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=1473 i_micro=367 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4733 mask_tok=142.00 y_norm=48.0874±0.0059 +2026-05-10 13:45:02,513 - INFO - Step 1850 | mode=video loss=0.117188 diff=0.117188 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0825 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0858 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=1481 i_micro=369 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4000 mask_tok=120.00 y_norm=48.0858±0.0063 +2026-05-10 13:45:20,476 - INFO - Step 1860 | mode=video loss=0.103027 diff=0.103027 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0321 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0872 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=1489 i_micro=371 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3633 mask_tok=109.00 y_norm=48.0872±0.0061 +2026-05-10 13:45:39,840 - INFO - Step 1870 | mode=video loss=0.078125 diff=0.078125 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0423 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0868 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=1497 i_micro=373 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3733 mask_tok=112.00 y_norm=48.0868±0.0052 +2026-05-10 13:45:57,797 - INFO - Step 1880 | mode=video loss=0.157227 diff=0.157227 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0628 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0862 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=1505 i_micro=375 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3833 mask_tok=115.00 y_norm=48.0862±0.0060 +2026-05-10 13:46:16,359 - INFO - Step 1890 | mode=video loss=0.092285 diff=0.092285 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0407 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0888 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=1513 i_micro=377 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3833 mask_tok=115.00 y_norm=48.0888±0.0063 +2026-05-10 13:46:35,088 - INFO - Step 1900 | mode=video loss=0.170898 diff=0.170898 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0461 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0891 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=1521 i_micro=379 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3400 mask_tok=102.00 y_norm=48.0891±0.0056 +2026-05-10 13:46:44,305 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 13:47:02,251 - INFO - Step 1910 | mode=video loss=0.139648 diff=0.139648 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1392 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0873 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=1529 i_micro=381 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3433 mask_tok=103.00 y_norm=48.0873±0.0057 +2026-05-10 13:47:21,002 - INFO - Step 1920 | mode=video loss=0.058350 diff=0.058350 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1181 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0855 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=1537 i_micro=383 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3533 mask_tok=106.00 y_norm=48.0855±0.0052 +2026-05-10 13:47:39,678 - INFO - Step 1930 | mode=video loss=0.283203 diff=0.283203 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0333 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0883 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=1545 i_micro=385 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3333 mask_tok=100.00 y_norm=48.0883±0.0053 +2026-05-10 13:47:58,344 - INFO - Step 1940 | mode=video loss=0.119629 diff=0.119629 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0699 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0884 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=1553 i_micro=387 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3600 mask_tok=108.00 y_norm=48.0884±0.0060 +2026-05-10 13:48:16,787 - INFO - Step 1950 | mode=video loss=0.138672 diff=0.138672 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0456 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0845 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=1561 i_micro=389 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4133 mask_tok=124.00 y_norm=48.0845±0.0055 +2026-05-10 13:48:34,744 - INFO - Step 1960 | mode=video loss=0.095703 diff=0.095703 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0440 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0898 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=1569 i_micro=391 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3400 mask_tok=102.00 y_norm=48.0898±0.0059 +2026-05-10 13:48:53,502 - INFO - Step 1970 | mode=video loss=0.088379 diff=0.088379 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0437 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0885 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=1577 i_micro=393 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3667 mask_tok=110.00 y_norm=48.0885±0.0060 +2026-05-10 13:49:11,879 - INFO - Step 1980 | mode=video loss=0.142578 diff=0.142578 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0732 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0892 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=1585 i_micro=395 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3933 mask_tok=118.00 y_norm=48.0892±0.0061 +2026-05-10 13:49:29,834 - INFO - Step 1990 | mode=video loss=0.132812 diff=0.132812 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0888 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0906 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=1593 i_micro=397 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.5500 mask_tok=165.00 y_norm=48.0906±0.0066 +2026-05-10 13:49:50,551 - INFO - Step 2000 | mode=video loss=0.092773 diff=0.092773 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0430 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0880 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=1601 i_micro=399 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3833 mask_tok=115.00 y_norm=48.0880±0.0063 cond_uncond_dloss=0.300781 cond_grad=0.000228 cond_pred_l2=37.447960 cond_pred_ratio=0.023445 +2026-05-10 13:49:50,589 - INFO - Step 2000 | probe_semantic mcp_offdiag(mean/min/max)=0.647626/0.535539/0.710514 smol_offdiag=0.660557 prompts=4 mcp_tok=17.50 smol_tok=17.50 +2026-05-10 13:49:57,511 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 13:50:15,459 - INFO - Step 2010 | mode=video loss=0.065430 diff=0.065430 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1091 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0873 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=1609 i_micro=401 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4433 mask_tok=133.00 y_norm=48.0873±0.0064 +2026-05-10 13:50:34,239 - INFO - Step 2020 | mode=video loss=0.131836 diff=0.131836 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0451 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0887 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=1617 i_micro=403 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.5500 mask_tok=165.00 y_norm=48.0887±0.0067 +2026-05-10 13:50:52,853 - INFO - Step 2030 | mode=video loss=0.078125 diff=0.078125 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0644 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0900 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=1625 i_micro=405 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3800 mask_tok=114.00 y_norm=48.0900±0.0072 +2026-05-10 13:51:10,814 - INFO - Step 2040 | mode=video loss=0.081543 diff=0.081543 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0534 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0888 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=1633 i_micro=407 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4667 mask_tok=140.00 y_norm=48.0888±0.0057 +2026-05-10 13:51:29,585 - INFO - Step 2050 | mode=video loss=0.098145 diff=0.098145 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0414 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0871 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=1641 i_micro=409 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4733 mask_tok=142.00 y_norm=48.0871±0.0060 +2026-05-10 13:51:48,072 - INFO - Step 2060 | mode=video loss=0.078613 diff=0.078613 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0439 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0896 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=1649 i_micro=411 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3400 mask_tok=102.00 y_norm=48.0896±0.0056 +2026-05-10 13:52:06,781 - INFO - Step 2070 | mode=video loss=0.076172 diff=0.076172 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0478 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0887 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=1657 i_micro=413 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3667 mask_tok=110.00 y_norm=48.0887±0.0058 +2026-05-10 13:52:25,206 - INFO - Step 2080 | mode=video loss=0.075684 diff=0.075684 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0361 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0891 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=1665 i_micro=415 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3600 mask_tok=108.00 y_norm=48.0891±0.0054 +2026-05-10 13:52:43,175 - INFO - Step 2090 | mode=video loss=0.102539 diff=0.102539 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0453 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0884 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=1673 i_micro=417 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3933 mask_tok=118.00 y_norm=48.0884±0.0061 +2026-05-10 13:53:01,839 - INFO - Step 2100 | mode=video loss=0.094238 diff=0.094238 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1000 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0890 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=1681 i_micro=419 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3600 mask_tok=108.00 y_norm=48.0890±0.0060 +2026-05-10 13:53:08,709 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 13:53:27,088 - INFO - Step 2110 | mode=video loss=0.095703 diff=0.095703 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0402 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0892 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=1689 i_micro=421 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3667 mask_tok=110.00 y_norm=48.0892±0.0062 +2026-05-10 13:53:45,785 - INFO - Step 2120 | mode=video loss=0.108887 diff=0.108887 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0357 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0897 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=1697 i_micro=423 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3267 mask_tok=98.00 y_norm=48.0897±0.0054 +2026-05-10 13:54:04,254 - INFO - Step 2130 | mode=video loss=0.135742 diff=0.135742 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0285 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0871 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=1705 i_micro=425 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3833 mask_tok=115.00 y_norm=48.0871±0.0062 +2026-05-10 13:54:22,211 - INFO - Step 2140 | mode=video loss=0.092285 diff=0.092285 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0298 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0907 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=1713 i_micro=427 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.2500 mask_tok=75.00 y_norm=48.0907±0.0047 +2026-05-10 13:54:40,874 - INFO - Step 2150 | mode=video loss=0.097168 diff=0.097168 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0606 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0905 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=1721 i_micro=429 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3600 mask_tok=108.00 y_norm=48.0905±0.0060 +2026-05-10 13:54:59,488 - INFO - Step 2160 | mode=video loss=0.082031 diff=0.082031 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0195 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0894 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=1729 i_micro=431 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3933 mask_tok=118.00 y_norm=48.0894±0.0061 +2026-05-10 13:55:17,447 - INFO - Step 2170 | mode=video loss=0.081543 diff=0.081543 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0570 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0885 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=1737 i_micro=433 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.5100 mask_tok=153.00 y_norm=48.0885±0.0066 +2026-05-10 13:55:35,932 - INFO - Step 2180 | mode=video loss=0.186523 diff=0.186523 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0437 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0896 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=1745 i_micro=435 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3667 mask_tok=110.00 y_norm=48.0896±0.0054 +2026-05-10 13:55:54,527 - INFO - Step 2190 | mode=video loss=0.123535 diff=0.123535 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0461 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0910 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=1753 i_micro=437 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3433 mask_tok=103.00 y_norm=48.0910±0.0055 +2026-05-10 13:56:14,838 - INFO - Step 2200 | mode=video loss=0.105469 diff=0.105469 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0708 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0886 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=1761 i_micro=439 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3500 mask_tok=105.00 y_norm=48.0886±0.0062 cond_uncond_dloss=0.292969 cond_grad=0.000043 cond_pred_l2=48.797344 cond_pred_ratio=0.029157 +2026-05-10 13:56:14,875 - INFO - Step 2200 | probe_semantic mcp_offdiag(mean/min/max)=0.649348/0.538018/0.711767 smol_offdiag=0.660557 prompts=4 mcp_tok=17.50 smol_tok=17.50 +2026-05-10 13:56:21,924 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 13:56:40,527 - INFO - Step 2210 | mode=video loss=0.083496 diff=0.083496 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0363 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0899 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=1769 i_micro=441 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.2900 mask_tok=87.00 y_norm=48.0899±0.0049 +2026-05-10 13:56:58,494 - INFO - Step 2220 | mode=video loss=0.074707 diff=0.074707 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0790 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0882 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=1777 i_micro=443 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.5133 mask_tok=154.00 y_norm=48.0882±0.0073 +2026-05-10 13:57:17,197 - INFO - Step 2230 | mode=video loss=0.169922 diff=0.169922 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0772 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0904 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=1785 i_micro=445 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3467 mask_tok=104.00 y_norm=48.0904±0.0055 +2026-05-10 13:57:35,734 - INFO - Step 2240 | mode=video loss=0.076660 diff=0.076660 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1101 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0896 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=1793 i_micro=447 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3700 mask_tok=111.00 y_norm=48.0896±0.0060 +2026-05-10 13:57:54,438 - INFO - Step 2250 | mode=video loss=0.076660 diff=0.076660 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1474 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0887 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=1801 i_micro=449 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3967 mask_tok=119.00 y_norm=48.0887±0.0062 +2026-05-10 13:58:13,005 - INFO - Step 2260 | mode=video loss=0.092285 diff=0.092285 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0602 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0885 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=1809 i_micro=451 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3633 mask_tok=109.00 y_norm=48.0885±0.0057 +2026-05-10 13:58:30,969 - INFO - Step 2270 | mode=video loss=0.048340 diff=0.048340 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0280 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0875 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=1817 i_micro=453 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3700 mask_tok=111.00 y_norm=48.0875±0.0059 +2026-05-10 13:58:49,703 - INFO - Step 2280 | mode=video loss=0.109863 diff=0.109863 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0243 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0891 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=1825 i_micro=455 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3767 mask_tok=113.00 y_norm=48.0891±0.0057 +2026-05-10 13:59:08,129 - INFO - Step 2290 | mode=video loss=0.093262 diff=0.093262 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0693 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0897 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=1833 i_micro=457 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3033 mask_tok=91.00 y_norm=48.0897±0.0059 +2026-05-10 13:59:26,134 - INFO - Step 2300 | mode=video loss=0.125000 diff=0.125000 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0678 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0875 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=1841 i_micro=459 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3767 mask_tok=113.00 y_norm=48.0875±0.0058 +2026-05-10 13:59:32,989 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 13:59:51,397 - INFO - Step 2310 | mode=video loss=0.121094 diff=0.121094 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1186 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0869 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=1849 i_micro=461 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4133 mask_tok=124.00 y_norm=48.0869±0.0057 +2026-05-10 14:00:09,371 - INFO - Step 2320 | mode=video loss=0.062256 diff=0.062256 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0755 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0906 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=1857 i_micro=463 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3667 mask_tok=110.00 y_norm=48.0906±0.0055 +2026-05-10 14:00:28,159 - INFO - Step 2330 | mode=video loss=0.094238 diff=0.094238 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0467 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0906 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=1865 i_micro=465 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3467 mask_tok=104.00 y_norm=48.0906±0.0057 +2026-05-10 14:00:46,639 - INFO - Step 2340 | mode=video loss=0.086914 diff=0.086914 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0731 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0903 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=1873 i_micro=467 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3067 mask_tok=92.00 y_norm=48.0903±0.0058 +2026-05-10 14:01:04,609 - INFO - Step 2350 | mode=video loss=0.098145 diff=0.098145 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0459 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0896 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=1881 i_micro=469 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3367 mask_tok=101.00 y_norm=48.0896±0.0056 +2026-05-10 14:01:23,369 - INFO - Step 2360 | mode=video loss=0.076172 diff=0.076172 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0384 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0899 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=1889 i_micro=471 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3767 mask_tok=113.00 y_norm=48.0899±0.0053 +2026-05-10 14:01:41,896 - INFO - Step 2370 | mode=video loss=0.067871 diff=0.067871 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0347 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0896 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=1897 i_micro=473 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.2800 mask_tok=84.00 y_norm=48.0896±0.0050 +2026-05-10 14:02:00,631 - INFO - Step 2380 | mode=video loss=0.094727 diff=0.094727 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0262 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0887 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=1905 i_micro=475 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3367 mask_tok=101.00 y_norm=48.0887±0.0059 +2026-05-10 14:02:19,011 - INFO - Step 2390 | mode=video loss=0.076172 diff=0.076172 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0584 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0879 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=1913 i_micro=477 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4500 mask_tok=135.00 y_norm=48.0879±0.0066 +2026-05-10 14:02:38,630 - INFO - Step 2400 | mode=video loss=0.070801 diff=0.070801 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0256 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0860 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=1921 i_micro=479 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4967 mask_tok=149.00 y_norm=48.0860±0.0059 cond_uncond_dloss=0.148438 cond_grad=0.000104 cond_pred_l2=40.233215 cond_pred_ratio=0.023536 +2026-05-10 14:02:38,669 - INFO - Step 2400 | probe_semantic mcp_offdiag(mean/min/max)=0.646573/0.533850/0.709683 smol_offdiag=0.660557 prompts=4 mcp_tok=17.50 smol_tok=17.50 +2026-05-10 14:02:45,457 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 14:03:04,159 - INFO - Step 2410 | mode=video loss=0.135742 diff=0.135742 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0466 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0893 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=1929 i_micro=481 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3667 mask_tok=110.00 y_norm=48.0893±0.0060 +2026-05-10 14:03:22,844 - INFO - Step 2420 | mode=video loss=0.128906 diff=0.128906 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.2008 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0892 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=1937 i_micro=483 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3333 mask_tok=100.00 y_norm=48.0892±0.0053 +2026-05-10 14:03:41,576 - INFO - Step 2430 | mode=video loss=0.093262 diff=0.093262 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0406 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0889 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=1945 i_micro=485 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3267 mask_tok=98.00 y_norm=48.0889±0.0059 +2026-05-10 14:04:00,011 - INFO - Step 2440 | mode=video loss=0.083496 diff=0.083496 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0524 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0891 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=1953 i_micro=487 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3367 mask_tok=101.00 y_norm=48.0891±0.0054 +2026-05-10 14:04:17,966 - INFO - Step 2450 | mode=video loss=0.087402 diff=0.087402 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0303 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0894 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=1961 i_micro=489 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3567 mask_tok=107.00 y_norm=48.0894±0.0059 +2026-05-10 14:04:36,647 - INFO - Step 2460 | mode=video loss=0.089355 diff=0.089355 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0213 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0898 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=1969 i_micro=491 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3600 mask_tok=108.00 y_norm=48.0898±0.0053 +2026-05-10 14:04:55,062 - INFO - Step 2470 | mode=video loss=0.123047 diff=0.123047 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1094 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0865 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=1977 i_micro=493 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4733 mask_tok=142.00 y_norm=48.0865±0.0075 +2026-05-10 14:05:13,015 - INFO - Step 2480 | mode=video loss=0.092773 diff=0.092773 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1435 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0873 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=1985 i_micro=495 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4367 mask_tok=131.00 y_norm=48.0873±0.0068 +2026-05-10 14:05:31,950 - INFO - Step 2490 | mode=video loss=0.103516 diff=0.103516 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0568 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0882 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=1993 i_micro=497 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3600 mask_tok=108.00 y_norm=48.0882±0.0058 +2026-05-10 14:05:49,910 - INFO - Step 2500 | mode=video loss=0.099609 diff=0.099609 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0394 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0898 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=2001 i_micro=499 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3800 mask_tok=114.00 y_norm=48.0898±0.0060 +2026-05-10 14:05:56,848 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 14:06:15,574 - INFO - Step 2510 | mode=video loss=0.087402 diff=0.087402 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0351 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0881 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=2009 i_micro=501 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3400 mask_tok=102.00 y_norm=48.0881±0.0051 +2026-05-10 14:06:34,329 - INFO - Step 2520 | mode=video loss=0.162109 diff=0.162109 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0599 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0880 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=2017 i_micro=503 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3300 mask_tok=99.00 y_norm=48.0880±0.0054 +2026-05-10 14:06:52,294 - INFO - Step 2530 | mode=video loss=0.065430 diff=0.065430 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0434 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0901 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=2025 i_micro=505 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.2900 mask_tok=87.00 y_norm=48.0901±0.0055 +2026-05-10 14:07:11,037 - INFO - Step 2540 | mode=video loss=0.110840 diff=0.110840 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0312 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0872 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=2033 i_micro=507 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4467 mask_tok=134.00 y_norm=48.0872±0.0062 +2026-05-10 14:07:29,743 - INFO - Step 2550 | mode=video loss=0.224609 diff=0.224609 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1061 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0895 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=2041 i_micro=509 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3500 mask_tok=105.00 y_norm=48.0895±0.0054 +2026-05-10 14:07:48,386 - INFO - Step 2560 | mode=video loss=0.176758 diff=0.176758 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0761 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0891 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=2049 i_micro=511 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4000 mask_tok=120.00 y_norm=48.0891±0.0054 +2026-05-10 14:08:06,939 - INFO - Step 2570 | mode=video loss=0.087891 diff=0.087891 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0426 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0882 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=2057 i_micro=513 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3800 mask_tok=114.00 y_norm=48.0882±0.0056 +2026-05-10 14:08:24,899 - INFO - Step 2580 | mode=video loss=0.096680 diff=0.096680 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0364 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0903 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=2065 i_micro=515 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3767 mask_tok=113.00 y_norm=48.0903±0.0059 +2026-05-10 14:08:43,501 - INFO - Step 2590 | mode=video loss=0.116211 diff=0.116211 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0695 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0900 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=2073 i_micro=517 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3233 mask_tok=97.00 y_norm=48.0900±0.0054 +2026-05-10 14:09:03,652 - INFO - Step 2600 | mode=video loss=0.084473 diff=0.084473 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0345 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0869 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=2081 i_micro=519 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3700 mask_tok=111.00 y_norm=48.0869±0.0056 cond_uncond_dloss=0.328125 cond_grad=0.000063 cond_pred_l2=44.359280 cond_pred_ratio=0.026300 +2026-05-10 14:09:03,690 - INFO - Step 2600 | probe_semantic mcp_offdiag(mean/min/max)=0.646461/0.533924/0.708522 smol_offdiag=0.660557 prompts=4 mcp_tok=17.50 smol_tok=17.50 +2026-05-10 14:09:10,517 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 14:09:29,297 - INFO - Step 2610 | mode=video loss=0.136719 diff=0.136719 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1345 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0888 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=2089 i_micro=521 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3800 mask_tok=114.00 y_norm=48.0888±0.0053 +2026-05-10 14:09:47,736 - INFO - Step 2620 | mode=video loss=0.101074 diff=0.101074 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0634 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0868 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=2097 i_micro=523 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4233 mask_tok=127.00 y_norm=48.0868±0.0056 +2026-05-10 14:10:05,703 - INFO - Step 2630 | mode=video loss=0.030396 diff=0.030396 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0294 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0864 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=2105 i_micro=525 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3900 mask_tok=117.00 y_norm=48.0864±0.0057 +2026-05-10 14:10:24,440 - INFO - Step 2640 | mode=video loss=0.081543 diff=0.081543 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0423 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0872 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=2113 i_micro=527 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3500 mask_tok=105.00 y_norm=48.0872±0.0060 +2026-05-10 14:10:42,890 - INFO - Step 2650 | mode=video loss=0.083008 diff=0.083008 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0251 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0882 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=2121 i_micro=529 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3233 mask_tok=97.00 y_norm=48.0882±0.0057 +2026-05-10 14:11:00,859 - INFO - Step 2660 | mode=video loss=0.064453 diff=0.064453 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0261 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0861 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=2129 i_micro=531 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4300 mask_tok=129.00 y_norm=48.0861±0.0060 +2026-05-10 14:11:19,543 - INFO - Step 2670 | mode=video loss=0.085938 diff=0.085938 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0267 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0883 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=2137 i_micro=533 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3533 mask_tok=106.00 y_norm=48.0883±0.0058 +2026-05-10 14:11:38,082 - INFO - Step 2680 | mode=video loss=0.056396 diff=0.056396 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0396 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0876 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=2145 i_micro=535 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4067 mask_tok=122.00 y_norm=48.0876±0.0065 +2026-05-10 14:11:56,713 - INFO - Step 2690 | mode=video loss=0.110840 diff=0.110840 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1790 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0885 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=2153 i_micro=537 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3767 mask_tok=113.00 y_norm=48.0885±0.0055 +2026-05-10 14:12:15,170 - INFO - Step 2700 | mode=video loss=0.082031 diff=0.082031 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0626 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0851 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=2161 i_micro=539 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3700 mask_tok=111.00 y_norm=48.0851±0.0057 +2026-05-10 14:12:21,951 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 14:12:39,884 - INFO - Step 2710 | mode=video loss=0.187500 diff=0.187500 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.7376 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0896 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=2169 i_micro=541 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.2833 mask_tok=85.00 y_norm=48.0896±0.0057 +2026-05-10 14:12:58,373 - INFO - Step 2720 | mode=video loss=0.074707 diff=0.074707 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0551 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0891 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=2177 i_micro=543 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3267 mask_tok=98.00 y_norm=48.0891±0.0054 +2026-05-10 14:13:16,861 - INFO - Step 2730 | mode=video loss=0.100586 diff=0.100586 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0630 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0877 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=2185 i_micro=545 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3533 mask_tok=106.00 y_norm=48.0877±0.0059 +2026-05-10 14:13:35,355 - INFO - Step 2740 | mode=video loss=0.131836 diff=0.131836 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0987 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0850 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=2193 i_micro=547 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4433 mask_tok=133.00 y_norm=48.0850±0.0062 +2026-05-10 14:13:53,854 - INFO - Step 2750 | mode=video loss=0.101562 diff=0.101562 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0409 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0883 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=2201 i_micro=549 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3533 mask_tok=106.00 y_norm=48.0883±0.0057 +2026-05-10 14:14:11,797 - INFO - Step 2760 | mode=video loss=0.140625 diff=0.140625 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1336 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0886 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=2209 i_micro=551 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3633 mask_tok=109.00 y_norm=48.0886±0.0063 +2026-05-10 14:14:30,488 - INFO - Step 2770 | mode=video loss=0.086426 diff=0.086426 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0403 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0871 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=2217 i_micro=553 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3800 mask_tok=114.00 y_norm=48.0871±0.0063 +2026-05-10 14:14:49,160 - INFO - Step 2780 | mode=video loss=0.102539 diff=0.102539 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0677 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0892 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=2225 i_micro=555 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3267 mask_tok=98.00 y_norm=48.0892±0.0053 +2026-05-10 14:15:07,876 - INFO - Step 2790 | mode=video loss=0.072266 diff=0.072266 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0388 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0877 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=2233 i_micro=557 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3733 mask_tok=112.00 y_norm=48.0877±0.0055 +2026-05-10 14:15:27,463 - INFO - Step 2800 | mode=video loss=0.229492 diff=0.229492 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.2628 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0893 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=2241 i_micro=559 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3033 mask_tok=91.00 y_norm=48.0893±0.0050 cond_uncond_dloss=0.972656 cond_grad=0.002896 cond_pred_l2=81.304337 cond_pred_ratio=0.053249 +2026-05-10 14:15:27,500 - INFO - Step 2800 | probe_semantic mcp_offdiag(mean/min/max)=0.645392/0.532038/0.708098 smol_offdiag=0.660557 prompts=4 mcp_tok=17.50 smol_tok=17.50 +2026-05-10 14:15:34,532 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 14:15:53,238 - INFO - Step 2810 | mode=video loss=0.060791 diff=0.060791 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0826 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0872 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=2249 i_micro=561 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3300 mask_tok=99.00 y_norm=48.0872±0.0058 +2026-05-10 14:16:11,962 - INFO - Step 2820 | mode=video loss=0.092285 diff=0.092285 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0383 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0886 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=2257 i_micro=563 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3500 mask_tok=105.00 y_norm=48.0886±0.0061 +2026-05-10 14:16:30,533 - INFO - Step 2830 | mode=video loss=0.086914 diff=0.086914 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0284 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0866 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=2265 i_micro=565 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4333 mask_tok=130.00 y_norm=48.0866±0.0060 +2026-05-10 14:16:48,499 - INFO - Step 2840 | mode=video loss=0.073242 diff=0.073242 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0363 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0898 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=2273 i_micro=567 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3300 mask_tok=99.00 y_norm=48.0898±0.0051 +2026-05-10 14:17:07,057 - INFO - Step 2850 | mode=video loss=0.082031 diff=0.082031 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0377 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0875 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=2281 i_micro=569 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3600 mask_tok=108.00 y_norm=48.0875±0.0050 +2026-05-10 14:17:25,634 - INFO - Step 2860 | mode=video loss=0.072266 diff=0.072266 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0303 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0895 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=2289 i_micro=571 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3333 mask_tok=100.00 y_norm=48.0895±0.0053 +2026-05-10 14:17:44,368 - INFO - Step 2870 | mode=video loss=0.087891 diff=0.087891 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0448 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0863 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=2297 i_micro=573 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3500 mask_tok=105.00 y_norm=48.0863±0.0063 +2026-05-10 14:18:02,892 - INFO - Step 2880 | mode=video loss=0.073242 diff=0.073242 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0676 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0869 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=2305 i_micro=575 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4200 mask_tok=126.00 y_norm=48.0869±0.0061 +2026-05-10 14:18:20,843 - INFO - Step 2890 | mode=video loss=0.050293 diff=0.050293 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0357 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0855 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=2313 i_micro=577 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4333 mask_tok=130.00 y_norm=48.0855±0.0056 +2026-05-10 14:18:39,577 - INFO - Step 2900 | mode=video loss=0.103027 diff=0.103027 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1026 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0905 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=2321 i_micro=579 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3200 mask_tok=96.00 y_norm=48.0905±0.0057 +2026-05-10 14:18:46,391 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 14:19:04,808 - INFO - Step 2910 | mode=video loss=0.098145 diff=0.098145 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0428 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0885 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=2329 i_micro=581 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4067 mask_tok=122.00 y_norm=48.0885±0.0062 +2026-05-10 14:19:23,573 - INFO - Step 2920 | mode=video loss=0.114746 diff=0.114746 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0453 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0856 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=2337 i_micro=583 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4800 mask_tok=144.00 y_norm=48.0856±0.0064 +2026-05-10 14:19:41,532 - INFO - Step 2930 | mode=video loss=0.102051 diff=0.102051 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0684 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0881 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=2345 i_micro=585 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3533 mask_tok=106.00 y_norm=48.0881±0.0050 +2026-05-10 14:19:59,928 - INFO - Step 2940 | mode=video loss=0.165039 diff=0.165039 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0988 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0895 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=2353 i_micro=587 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3300 mask_tok=99.00 y_norm=48.0895±0.0066 +2026-05-10 14:20:18,616 - INFO - Step 2950 | mode=video loss=0.075684 diff=0.075684 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0412 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0871 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=2361 i_micro=589 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.5133 mask_tok=154.00 y_norm=48.0871±0.0067 +2026-05-10 14:20:37,015 - INFO - Step 2960 | mode=video loss=0.084473 diff=0.084473 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0373 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0892 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=2369 i_micro=591 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3967 mask_tok=119.00 y_norm=48.0892±0.0062 +2026-05-10 14:20:54,977 - INFO - Step 2970 | mode=video loss=0.066406 diff=0.066406 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0404 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0891 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=2377 i_micro=593 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3900 mask_tok=117.00 y_norm=48.0891±0.0060 +2026-05-10 14:21:13,613 - INFO - Step 2980 | mode=video loss=0.094727 diff=0.094727 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0321 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0906 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=2385 i_micro=595 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.5467 mask_tok=164.00 y_norm=48.0906±0.0065 +2026-05-10 14:21:32,014 - INFO - Step 2990 | mode=video loss=0.082031 diff=0.082031 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0264 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0893 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=2393 i_micro=597 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3833 mask_tok=115.00 y_norm=48.0893±0.0063 +2026-05-10 14:21:52,369 - INFO - Step 3000 | mode=video loss=0.087891 diff=0.087891 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0377 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0876 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=2401 i_micro=599 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3433 mask_tok=103.00 y_norm=48.0876±0.0059 cond_uncond_dloss=0.287109 cond_grad=0.000049 cond_pred_l2=35.166573 cond_pred_ratio=0.020823 +2026-05-10 14:21:52,407 - INFO - Step 3000 | probe_semantic mcp_offdiag(mean/min/max)=0.647535/0.534644/0.710498 smol_offdiag=0.660557 prompts=4 mcp_tok=17.50 smol_tok=17.50 +2026-05-10 14:21:59,436 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 14:22:17,858 - INFO - Step 3010 | mode=video loss=0.143555 diff=0.143555 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0479 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0866 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=2409 i_micro=601 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3767 mask_tok=113.00 y_norm=48.0866±0.0059 +2026-05-10 14:22:35,827 - INFO - Step 3020 | mode=video loss=0.070801 diff=0.070801 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0426 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0875 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=2417 i_micro=603 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3900 mask_tok=117.00 y_norm=48.0875±0.0064 +2026-05-10 14:22:54,565 - INFO - Step 3030 | mode=video loss=0.071289 diff=0.071289 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0483 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0892 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=2425 i_micro=605 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3600 mask_tok=108.00 y_norm=48.0892±0.0057 +2026-05-10 14:23:13,159 - INFO - Step 3040 | mode=video loss=0.062988 diff=0.062988 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0356 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0898 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=2433 i_micro=607 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3467 mask_tok=104.00 y_norm=48.0898±0.0049 +2026-05-10 14:23:31,941 - INFO - Step 3050 | mode=video loss=0.080078 diff=0.080078 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0336 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0895 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=2441 i_micro=609 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3400 mask_tok=102.00 y_norm=48.0895±0.0057 +2026-05-10 14:23:49,906 - INFO - Step 3060 | mode=video loss=0.087891 diff=0.087891 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0515 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0865 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=2449 i_micro=611 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4033 mask_tok=121.00 y_norm=48.0865±0.0059 +2026-05-10 14:24:08,295 - INFO - Step 3070 | mode=video loss=0.090820 diff=0.090820 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0222 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0894 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=2457 i_micro=613 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3900 mask_tok=117.00 y_norm=48.0894±0.0058 +2026-05-10 14:24:26,972 - INFO - Step 3080 | mode=video loss=0.055908 diff=0.055908 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1751 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0869 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=2465 i_micro=615 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4533 mask_tok=136.00 y_norm=48.0869±0.0061 +2026-05-10 14:24:45,365 - INFO - Step 3090 | mode=video loss=0.081543 diff=0.081543 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0624 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0902 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=2473 i_micro=617 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3533 mask_tok=106.00 y_norm=48.0902±0.0060 +2026-05-10 14:25:03,339 - INFO - Step 3100 | mode=video loss=0.115234 diff=0.115234 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.6378 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0880 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=2481 i_micro=619 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3967 mask_tok=119.00 y_norm=48.0880±0.0065 +2026-05-10 14:25:10,227 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 14:25:28,960 - INFO - Step 3110 | mode=video loss=0.098145 diff=0.098145 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0314 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0873 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=2489 i_micro=621 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4067 mask_tok=122.00 y_norm=48.0873±0.0065 +2026-05-10 14:25:47,389 - INFO - Step 3120 | mode=video loss=0.142578 diff=0.142578 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0823 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0899 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=2497 i_micro=623 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3400 mask_tok=102.00 y_norm=48.0899±0.0055 +2026-05-10 14:26:06,017 - INFO - Step 3130 | mode=video loss=0.223633 diff=0.223633 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0951 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0898 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=2505 i_micro=625 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3400 mask_tok=102.00 y_norm=48.0898±0.0053 +2026-05-10 14:26:23,972 - INFO - Step 3140 | mode=video loss=0.097656 diff=0.097656 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0600 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0871 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=2513 i_micro=627 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4233 mask_tok=127.00 y_norm=48.0871±0.0061 +2026-05-10 14:26:42,497 - INFO - Step 3150 | mode=video loss=0.094727 diff=0.094727 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1442 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0899 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=2521 i_micro=629 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3733 mask_tok=112.00 y_norm=48.0899±0.0054 +2026-05-10 14:27:01,196 - INFO - Step 3160 | mode=video loss=0.088379 diff=0.088379 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0693 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0876 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=2529 i_micro=631 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.2833 mask_tok=85.00 y_norm=48.0876±0.0057 +2026-05-10 14:27:19,688 - INFO - Step 3170 | mode=video loss=0.092285 diff=0.092285 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0488 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0896 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=2537 i_micro=633 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4833 mask_tok=145.00 y_norm=48.0896±0.0066 +2026-05-10 14:27:38,205 - INFO - Step 3180 | mode=video loss=0.112793 diff=0.112793 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0711 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0909 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=2545 i_micro=635 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3000 mask_tok=90.00 y_norm=48.0909±0.0059 +2026-05-10 14:27:56,167 - INFO - Step 3190 | mode=video loss=0.109375 diff=0.109375 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0607 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0902 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=2553 i_micro=637 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3400 mask_tok=102.00 y_norm=48.0902±0.0061 +2026-05-10 14:28:16,184 - INFO - Step 3200 | mode=video loss=0.063477 diff=0.063477 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0299 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0884 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=2561 i_micro=639 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4567 mask_tok=137.00 y_norm=48.0884±0.0073 cond_uncond_dloss=0.628906 cond_grad=0.000067 cond_pred_l2=42.648476 cond_pred_ratio=0.024630 +2026-05-10 14:28:16,222 - INFO - Step 3200 | probe_semantic mcp_offdiag(mean/min/max)=0.649371/0.537765/0.711772 smol_offdiag=0.660557 prompts=4 mcp_tok=17.50 smol_tok=17.50 +2026-05-10 14:28:23,017 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 14:28:41,738 - INFO - Step 3210 | mode=video loss=0.098633 diff=0.098633 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0582 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0924 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=2569 i_micro=641 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.5500 mask_tok=165.00 y_norm=48.0924±0.0068 +2026-05-10 14:29:00,135 - INFO - Step 3220 | mode=video loss=0.114258 diff=0.114258 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0493 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0880 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=2577 i_micro=643 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3900 mask_tok=117.00 y_norm=48.0880±0.0055 +2026-05-10 14:29:18,906 - INFO - Step 3230 | mode=video loss=0.069336 diff=0.069336 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0292 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0895 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=2585 i_micro=645 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3333 mask_tok=100.00 y_norm=48.0895±0.0069 +2026-05-10 14:29:36,863 - INFO - Step 3240 | mode=video loss=0.089355 diff=0.089355 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0508 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0899 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=2593 i_micro=647 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3500 mask_tok=105.00 y_norm=48.0899±0.0060 +2026-05-10 14:29:55,257 - INFO - Step 3250 | mode=video loss=0.107422 diff=0.107422 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0659 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0902 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=2601 i_micro=649 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3267 mask_tok=98.00 y_norm=48.0902±0.0054 +2026-05-10 14:30:13,927 - INFO - Step 3260 | mode=video loss=0.092285 diff=0.092285 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0418 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0879 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=2609 i_micro=651 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3500 mask_tok=105.00 y_norm=48.0879±0.0054 +2026-05-10 14:30:31,879 - INFO - Step 3270 | mode=video loss=0.073730 diff=0.073730 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0540 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0886 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=2617 i_micro=653 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.2400 mask_tok=72.00 y_norm=48.0886±0.0047 +2026-05-10 14:30:50,299 - INFO - Step 3280 | mode=video loss=0.109375 diff=0.109375 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0646 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0884 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=2625 i_micro=655 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3633 mask_tok=109.00 y_norm=48.0884±0.0061 +2026-05-10 14:31:08,972 - INFO - Step 3290 | mode=video loss=0.144531 diff=0.144531 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0706 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0872 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=2633 i_micro=657 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3533 mask_tok=106.00 y_norm=48.0872±0.0058 +2026-05-10 14:31:27,398 - INFO - Step 3300 | mode=video loss=0.102051 diff=0.102051 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0501 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0857 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=2641 i_micro=659 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.5100 mask_tok=153.00 y_norm=48.0857±0.0064 +2026-05-10 14:31:34,296 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 14:31:52,732 - INFO - Step 3310 | mode=video loss=0.068848 diff=0.068848 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0398 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0916 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=2649 i_micro=661 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.5500 mask_tok=165.00 y_norm=48.0916±0.0077 +2026-05-10 14:32:10,702 - INFO - Step 3320 | mode=video loss=0.088379 diff=0.088379 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0316 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0888 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=2657 i_micro=663 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.2867 mask_tok=86.00 y_norm=48.0888±0.0050 +2026-05-10 14:32:29,123 - INFO - Step 3330 | mode=video loss=0.094727 diff=0.094727 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0487 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0900 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=2665 i_micro=665 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3267 mask_tok=98.00 y_norm=48.0900±0.0059 +2026-05-10 14:32:47,771 - INFO - Step 3340 | mode=video loss=0.120605 diff=0.120605 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0532 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0885 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=2673 i_micro=667 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4167 mask_tok=125.00 y_norm=48.0885±0.0056 +2026-05-10 14:33:06,187 - INFO - Step 3350 | mode=video loss=0.081055 diff=0.081055 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0477 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0892 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=2681 i_micro=669 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3567 mask_tok=107.00 y_norm=48.0892±0.0057 +2026-05-10 14:33:24,925 - INFO - Step 3360 | mode=video loss=0.089844 diff=0.089844 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0421 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0859 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=2689 i_micro=671 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3600 mask_tok=108.00 y_norm=48.0859±0.0060 +2026-05-10 14:33:42,902 - INFO - Step 3370 | mode=video loss=0.061279 diff=0.061279 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0325 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0884 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=2697 i_micro=673 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.5100 mask_tok=153.00 y_norm=48.0884±0.0075 +2026-05-10 14:34:01,291 - INFO - Step 3380 | mode=video loss=0.169922 diff=0.169922 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0386 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0883 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=2705 i_micro=675 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3467 mask_tok=104.00 y_norm=48.0883±0.0061 +2026-05-10 14:34:20,013 - INFO - Step 3390 | mode=video loss=0.287109 diff=0.287109 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0791 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0894 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=2713 i_micro=677 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.5000 mask_tok=150.00 y_norm=48.0894±0.0072 +2026-05-10 14:34:40,153 - INFO - Step 3400 | mode=video loss=0.151367 diff=0.151367 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1442 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0899 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=2721 i_micro=679 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3467 mask_tok=104.00 y_norm=48.0899±0.0056 cond_uncond_dloss=0.593750 cond_grad=0.001144 cond_pred_l2=104.998825 cond_pred_ratio=0.068151 +2026-05-10 14:34:40,190 - INFO - Step 3400 | probe_semantic mcp_offdiag(mean/min/max)=0.646153/0.533084/0.709507 smol_offdiag=0.660557 prompts=4 mcp_tok=17.50 smol_tok=17.50 +2026-05-10 14:34:46,969 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 14:35:04,924 - INFO - Step 3410 | mode=video loss=0.235352 diff=0.235352 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0417 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0873 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=2729 i_micro=681 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4433 mask_tok=133.00 y_norm=48.0873±0.0063 +2026-05-10 14:35:23,699 - INFO - Step 3420 | mode=video loss=0.101074 diff=0.101074 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0567 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0867 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=2737 i_micro=683 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3933 mask_tok=118.00 y_norm=48.0867±0.0060 +2026-05-10 14:35:42,309 - INFO - Step 3430 | mode=video loss=0.073242 diff=0.073242 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0365 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0892 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=2745 i_micro=685 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3533 mask_tok=106.00 y_norm=48.0892±0.0059 +2026-05-10 14:36:01,087 - INFO - Step 3440 | mode=video loss=0.065918 diff=0.065918 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0578 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0861 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=2753 i_micro=687 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4233 mask_tok=127.00 y_norm=48.0861±0.0062 +2026-05-10 14:36:19,058 - INFO - Step 3450 | mode=video loss=0.084961 diff=0.084961 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0280 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0887 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=2761 i_micro=689 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4000 mask_tok=120.00 y_norm=48.0887±0.0054 +2026-05-10 14:36:37,459 - INFO - Step 3460 | mode=video loss=0.089844 diff=0.089844 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0545 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0871 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=2769 i_micro=691 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4333 mask_tok=130.00 y_norm=48.0871±0.0063 +2026-05-10 14:36:56,127 - INFO - Step 3470 | mode=video loss=0.085449 diff=0.085449 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0401 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0896 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=2777 i_micro=693 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3333 mask_tok=100.00 y_norm=48.0896±0.0053 +2026-05-10 14:37:14,530 - INFO - Step 3480 | mode=video loss=0.094238 diff=0.094238 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0271 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0877 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=2785 i_micro=695 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3567 mask_tok=107.00 y_norm=48.0877±0.0056 +2026-05-10 14:37:33,064 - INFO - Step 3490 | mode=video loss=0.084961 diff=0.084961 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0268 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0899 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=2793 i_micro=697 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3467 mask_tok=104.00 y_norm=48.0899±0.0057 +2026-05-10 14:37:51,024 - INFO - Step 3500 | mode=video loss=0.359375 diff=0.359375 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.4272 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0885 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=2801 i_micro=699 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3133 mask_tok=94.00 y_norm=48.0885±0.0058 +2026-05-10 14:37:58,145 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 14:38:16,537 - INFO - Step 3510 | mode=video loss=0.078125 diff=0.078125 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0343 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0885 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=2809 i_micro=701 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3467 mask_tok=104.00 y_norm=48.0885±0.0055 +2026-05-10 14:38:35,199 - INFO - Step 3520 | mode=video loss=0.098145 diff=0.098145 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0707 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0901 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=2817 i_micro=703 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3333 mask_tok=100.00 y_norm=48.0901±0.0055 +2026-05-10 14:38:53,619 - INFO - Step 3530 | mode=video loss=0.106445 diff=0.106445 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0442 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0876 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=2825 i_micro=705 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4733 mask_tok=142.00 y_norm=48.0876±0.0066 +2026-05-10 14:39:11,573 - INFO - Step 3540 | mode=video loss=0.145508 diff=0.145508 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0201 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0878 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=2833 i_micro=707 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3767 mask_tok=113.00 y_norm=48.0878±0.0055 +2026-05-10 14:39:30,002 - INFO - Step 3550 | mode=video loss=0.115234 diff=0.115234 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0400 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0887 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=2841 i_micro=709 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4000 mask_tok=120.00 y_norm=48.0887±0.0068 +2026-05-10 14:39:48,540 - INFO - Step 3560 | mode=video loss=0.083984 diff=0.083984 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0323 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0891 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=2849 i_micro=711 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3367 mask_tok=101.00 y_norm=48.0891±0.0058 +2026-05-10 14:40:06,998 - INFO - Step 3570 | mode=video loss=0.111816 diff=0.111816 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0407 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0865 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=2857 i_micro=713 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4300 mask_tok=129.00 y_norm=48.0865±0.0067 +2026-05-10 14:40:24,946 - INFO - Step 3580 | mode=video loss=0.052734 diff=0.052734 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0315 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0884 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=2865 i_micro=715 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3733 mask_tok=112.00 y_norm=48.0884±0.0050 +2026-05-10 14:40:43,362 - INFO - Step 3590 | mode=video loss=0.119629 diff=0.119629 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0431 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0881 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=2873 i_micro=717 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3267 mask_tok=98.00 y_norm=48.0881±0.0056 +2026-05-10 14:41:03,653 - INFO - Step 3600 | mode=video loss=0.096191 diff=0.096191 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0488 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0893 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=2881 i_micro=719 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3500 mask_tok=105.00 y_norm=48.0893±0.0060 cond_uncond_dloss=0.478516 cond_grad=0.000254 cond_pred_l2=49.572796 cond_pred_ratio=0.030383 +2026-05-10 14:41:03,690 - INFO - Step 3600 | probe_semantic mcp_offdiag(mean/min/max)=0.645785/0.533029/0.708804 smol_offdiag=0.660557 prompts=4 mcp_tok=17.50 smol_tok=17.50 +2026-05-10 14:41:10,791 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 14:41:29,302 - INFO - Step 3610 | mode=video loss=0.062500 diff=0.062500 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0518 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0879 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=2889 i_micro=721 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3900 mask_tok=117.00 y_norm=48.0879±0.0065 +2026-05-10 14:41:48,020 - INFO - Step 3620 | mode=video loss=0.224609 diff=0.224609 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.2735 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0890 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=2897 i_micro=723 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3367 mask_tok=101.00 y_norm=48.0890±0.0054 +2026-05-10 14:42:05,985 - INFO - Step 3630 | mode=video loss=0.098145 diff=0.098145 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0481 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0884 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=2905 i_micro=725 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3267 mask_tok=98.00 y_norm=48.0884±0.0055 +2026-05-10 14:42:24,602 - INFO - Step 3640 | mode=video loss=0.070801 diff=0.070801 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1014 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0884 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=2913 i_micro=727 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3567 mask_tok=107.00 y_norm=48.0884±0.0051 +2026-05-10 14:42:43,263 - INFO - Step 3650 | mode=video loss=0.079590 diff=0.079590 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0436 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0880 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=2921 i_micro=729 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3333 mask_tok=100.00 y_norm=48.0880±0.0061 +2026-05-10 14:43:01,741 - INFO - Step 3660 | mode=video loss=0.067383 diff=0.067383 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0412 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0866 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=2929 i_micro=731 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4867 mask_tok=146.00 y_norm=48.0866±0.0064 +2026-05-10 14:43:19,706 - INFO - Step 3670 | mode=video loss=0.061523 diff=0.061523 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0278 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0881 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=2937 i_micro=733 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3433 mask_tok=103.00 y_norm=48.0881±0.0052 +2026-05-10 14:43:38,432 - INFO - Step 3680 | mode=video loss=0.091309 diff=0.091309 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0412 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0891 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=2945 i_micro=735 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3600 mask_tok=108.00 y_norm=48.0891±0.0053 +2026-05-10 14:43:56,881 - INFO - Step 3690 | mode=video loss=0.120605 diff=0.120605 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0492 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0885 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=2953 i_micro=737 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3533 mask_tok=106.00 y_norm=48.0885±0.0052 +2026-05-10 14:44:15,549 - INFO - Step 3700 | mode=video loss=0.189453 diff=0.189453 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0901 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0886 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=2961 i_micro=739 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3167 mask_tok=95.00 y_norm=48.0886±0.0050 +2026-05-10 14:44:22,413 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 14:44:40,406 - INFO - Step 3710 | mode=video loss=0.086914 diff=0.086914 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0336 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0877 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=2969 i_micro=741 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3833 mask_tok=115.00 y_norm=48.0877±0.0052 +2026-05-10 14:44:59,087 - INFO - Step 3720 | mode=video loss=0.093262 diff=0.093262 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0298 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0860 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=2977 i_micro=743 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4400 mask_tok=132.00 y_norm=48.0860±0.0062 +2026-05-10 14:45:17,845 - INFO - Step 3730 | mode=video loss=0.076660 diff=0.076660 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0466 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0875 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=2985 i_micro=745 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3167 mask_tok=95.00 y_norm=48.0875±0.0053 +2026-05-10 14:45:36,270 - INFO - Step 3740 | mode=video loss=0.098145 diff=0.098145 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0828 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0893 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=2993 i_micro=747 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3567 mask_tok=107.00 y_norm=48.0893±0.0061 +2026-05-10 14:45:55,020 - INFO - Step 3750 | mode=video loss=0.121582 diff=0.121582 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0322 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0864 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=3001 i_micro=749 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3400 mask_tok=102.00 y_norm=48.0864±0.0069 +2026-05-10 14:46:12,982 - INFO - Step 3760 | mode=video loss=0.063965 diff=0.063965 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1035 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0861 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=3009 i_micro=751 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3733 mask_tok=112.00 y_norm=48.0861±0.0065 +2026-05-10 14:46:31,580 - INFO - Step 3770 | mode=video loss=0.087891 diff=0.087891 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0517 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0875 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=3017 i_micro=753 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3267 mask_tok=98.00 y_norm=48.0875±0.0046 +2026-05-10 14:46:50,314 - INFO - Step 3780 | mode=video loss=0.085938 diff=0.085938 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0566 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0876 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=3025 i_micro=755 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3333 mask_tok=100.00 y_norm=48.0876±0.0055 +2026-05-10 14:47:08,838 - INFO - Step 3790 | mode=video loss=0.117188 diff=0.117188 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1067 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0873 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=3033 i_micro=757 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3333 mask_tok=100.00 y_norm=48.0873±0.0052 +2026-05-10 14:47:28,419 - INFO - Step 3800 | mode=video loss=0.075684 diff=0.075684 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0290 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0883 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=3041 i_micro=759 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3600 mask_tok=108.00 y_norm=48.0883±0.0068 cond_uncond_dloss=0.302734 cond_grad=0.000077 cond_pred_l2=35.267273 cond_pred_ratio=0.021705 +2026-05-10 14:47:28,456 - INFO - Step 3800 | probe_semantic mcp_offdiag(mean/min/max)=0.645288/0.533035/0.708631 smol_offdiag=0.660557 prompts=4 mcp_tok=17.50 smol_tok=17.50 +2026-05-10 14:47:35,345 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 14:47:54,094 - INFO - Step 3810 | mode=video loss=0.072754 diff=0.072754 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0541 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0855 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=3049 i_micro=761 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4833 mask_tok=145.00 y_norm=48.0855±0.0062 +2026-05-10 14:48:12,705 - INFO - Step 3820 | mode=video loss=0.071777 diff=0.071777 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0276 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0862 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=3057 i_micro=763 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3233 mask_tok=97.00 y_norm=48.0862±0.0060 +2026-05-10 14:48:31,334 - INFO - Step 3830 | mode=video loss=0.129883 diff=0.129883 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0220 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0873 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=3065 i_micro=765 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.9300 mask_tok=279.00 y_norm=48.0873±0.0085 +2026-05-10 14:48:49,303 - INFO - Step 3840 | mode=video loss=0.159180 diff=0.159180 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0547 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0877 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=3073 i_micro=767 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3267 mask_tok=98.00 y_norm=48.0877±0.0054 +2026-05-10 14:49:07,693 - INFO - Step 3850 | mode=video loss=0.097656 diff=0.097656 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0469 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0886 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=3081 i_micro=769 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3400 mask_tok=102.00 y_norm=48.0886±0.0061 +2026-05-10 14:49:26,380 - INFO - Step 3860 | mode=video loss=0.104004 diff=0.104004 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0596 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0883 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=3089 i_micro=771 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3267 mask_tok=98.00 y_norm=48.0883±0.0050 +2026-05-10 14:49:44,750 - INFO - Step 3870 | mode=video loss=0.068848 diff=0.068848 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1131 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0877 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=3097 i_micro=773 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3900 mask_tok=117.00 y_norm=48.0877±0.0055 +2026-05-10 14:50:03,415 - INFO - Step 3880 | mode=video loss=0.069336 diff=0.069336 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0413 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0895 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=3105 i_micro=775 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3333 mask_tok=100.00 y_norm=48.0895±0.0057 +2026-05-10 14:50:21,374 - INFO - Step 3890 | mode=video loss=0.057617 diff=0.057617 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0616 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0856 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=3113 i_micro=777 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4400 mask_tok=132.00 y_norm=48.0856±0.0059 +2026-05-10 14:50:39,921 - INFO - Step 3900 | mode=video loss=0.164062 diff=0.164062 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1167 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0877 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=3121 i_micro=779 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3200 mask_tok=96.00 y_norm=48.0877±0.0054 +2026-05-10 14:50:46,718 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 14:51:05,443 - INFO - Step 3910 | mode=video loss=0.091309 diff=0.091309 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0474 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0851 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=3129 i_micro=781 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4133 mask_tok=124.00 y_norm=48.0851±0.0064 +2026-05-10 14:51:24,121 - INFO - Step 3920 | mode=video loss=0.091797 diff=0.091797 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0550 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0878 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=3137 i_micro=783 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.2667 mask_tok=80.00 y_norm=48.0878±0.0054 +2026-05-10 14:51:42,082 - INFO - Step 3930 | mode=video loss=0.114746 diff=0.114746 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1101 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0849 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=3145 i_micro=785 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3867 mask_tok=116.00 y_norm=48.0849±0.0064 +2026-05-10 14:52:00,714 - INFO - Step 3940 | mode=video loss=0.128906 diff=0.128906 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0604 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0849 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=3153 i_micro=787 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3633 mask_tok=109.00 y_norm=48.0849±0.0056 +2026-05-10 14:52:19,343 - INFO - Step 3950 | mode=video loss=0.073242 diff=0.073242 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0418 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0863 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=3161 i_micro=789 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.2767 mask_tok=83.00 y_norm=48.0863±0.0058 +2026-05-10 14:52:38,100 - INFO - Step 3960 | mode=video loss=0.118652 diff=0.118652 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0484 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0902 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=3169 i_micro=791 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4433 mask_tok=133.00 y_norm=48.0902±0.0058 +2026-05-10 14:52:56,056 - INFO - Step 3970 | mode=video loss=0.337891 diff=0.337891 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.5378 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0880 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=3177 i_micro=793 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3633 mask_tok=109.00 y_norm=48.0880±0.0052 +2026-05-10 14:53:14,413 - INFO - Step 3980 | mode=video loss=0.142578 diff=0.142578 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0711 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0866 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=3185 i_micro=795 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3567 mask_tok=107.00 y_norm=48.0866±0.0060 +2026-05-10 14:53:33,111 - INFO - Step 3990 | mode=video loss=0.086426 diff=0.086426 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0406 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0879 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=3193 i_micro=797 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3600 mask_tok=108.00 y_norm=48.0879±0.0057 +2026-05-10 14:53:53,105 - INFO - Step 4000 | mode=video loss=0.076660 diff=0.076660 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0530 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0881 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=3201 i_micro=799 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3700 mask_tok=111.00 y_norm=48.0881±0.0065 cond_uncond_dloss=0.427734 cond_grad=0.000082 cond_pred_l2=39.102684 cond_pred_ratio=0.023736 +2026-05-10 14:53:53,143 - INFO - Step 4000 | probe_semantic mcp_offdiag(mean/min/max)=0.644730/0.532322/0.708461 smol_offdiag=0.660557 prompts=4 mcp_tok=17.50 smol_tok=17.50 +2026-05-10 14:53:59,891 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 14:54:18,647 - INFO - Step 4010 | mode=video loss=0.128906 diff=0.128906 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1600 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0873 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=3209 i_micro=801 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3933 mask_tok=118.00 y_norm=48.0873±0.0058 +2026-05-10 14:54:36,605 - INFO - Step 4020 | mode=video loss=0.078125 diff=0.078125 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0423 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0882 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=3217 i_micro=803 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.2767 mask_tok=83.00 y_norm=48.0882±0.0054 +2026-05-10 14:54:55,187 - INFO - Step 4030 | mode=video loss=0.214844 diff=0.214844 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1833 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0879 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=3225 i_micro=805 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3033 mask_tok=91.00 y_norm=48.0879±0.0051 +2026-05-10 14:55:13,790 - INFO - Step 4040 | mode=video loss=0.071289 diff=0.071289 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0643 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0860 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=3233 i_micro=807 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3533 mask_tok=106.00 y_norm=48.0860±0.0054 +2026-05-10 14:55:32,164 - INFO - Step 4050 | mode=video loss=0.136719 diff=0.136719 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.4978 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0868 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=3241 i_micro=809 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4467 mask_tok=134.00 y_norm=48.0868±0.0065 +2026-05-10 14:55:50,109 - INFO - Step 4060 | mode=video loss=0.096191 diff=0.096191 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0890 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0857 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=3249 i_micro=811 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3467 mask_tok=104.00 y_norm=48.0857±0.0059 +2026-05-10 14:56:08,858 - INFO - Step 4070 | mode=video loss=0.075195 diff=0.075195 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0647 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0876 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=3257 i_micro=813 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3200 mask_tok=96.00 y_norm=48.0876±0.0069 +2026-05-10 14:56:27,264 - INFO - Step 4080 | mode=video loss=0.083496 diff=0.083496 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0722 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0881 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=3265 i_micro=815 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3433 mask_tok=103.00 y_norm=48.0881±0.0054 +2026-05-10 14:56:45,969 - INFO - Step 4090 | mode=video loss=0.090820 diff=0.090820 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0442 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0855 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=3273 i_micro=817 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3467 mask_tok=104.00 y_norm=48.0855±0.0059 +2026-05-10 14:57:03,925 - INFO - Step 4100 | mode=video loss=0.086914 diff=0.086914 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0503 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0880 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=3281 i_micro=819 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3233 mask_tok=97.00 y_norm=48.0880±0.0049 +2026-05-10 14:57:10,826 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 14:57:29,247 - INFO - Step 4110 | mode=video loss=0.060547 diff=0.060547 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0552 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0872 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=3289 i_micro=821 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.2900 mask_tok=87.00 y_norm=48.0872±0.0057 +2026-05-10 14:57:48,000 - INFO - Step 4120 | mode=video loss=0.128906 diff=0.128906 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0585 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0865 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=3297 i_micro=823 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3433 mask_tok=103.00 y_norm=48.0865±0.0057 +2026-05-10 14:58:06,382 - INFO - Step 4130 | mode=video loss=0.136719 diff=0.136719 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0381 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0880 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=3305 i_micro=825 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4000 mask_tok=120.00 y_norm=48.0880±0.0057 +2026-05-10 14:58:24,989 - INFO - Step 4140 | mode=video loss=0.216797 diff=0.216797 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0559 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0869 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=3313 i_micro=827 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3833 mask_tok=115.00 y_norm=48.0869±0.0053 +2026-05-10 14:58:42,935 - INFO - Step 4150 | mode=video loss=0.076660 diff=0.076660 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0266 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0872 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=3321 i_micro=829 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3267 mask_tok=98.00 y_norm=48.0872±0.0051 +2026-05-10 14:59:01,349 - INFO - Step 4160 | mode=video loss=0.085449 diff=0.085449 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0392 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0860 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=3329 i_micro=831 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3333 mask_tok=100.00 y_norm=48.0860±0.0049 +2026-05-10 14:59:19,979 - INFO - Step 4170 | mode=video loss=0.097656 diff=0.097656 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0961 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0869 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=3337 i_micro=833 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3767 mask_tok=113.00 y_norm=48.0869±0.0054 +2026-05-10 14:59:38,654 - INFO - Step 4180 | mode=video loss=0.061279 diff=0.061279 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0501 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0869 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=3345 i_micro=835 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4433 mask_tok=133.00 y_norm=48.0869±0.0064 +2026-05-10 14:59:56,605 - INFO - Step 4190 | mode=video loss=0.116211 diff=0.116211 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0293 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0856 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=3353 i_micro=837 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3633 mask_tok=109.00 y_norm=48.0856±0.0057 +2026-05-10 15:00:16,707 - INFO - Step 4200 | mode=video loss=0.174805 diff=0.174805 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1809 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0861 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=3361 i_micro=839 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3700 mask_tok=111.00 y_norm=48.0861±0.0060 cond_uncond_dloss=1.062500 cond_grad=0.002103 cond_pred_l2=46.300522 cond_pred_ratio=0.029789 +2026-05-10 15:00:16,744 - INFO - Step 4200 | probe_semantic mcp_offdiag(mean/min/max)=0.644474/0.531646/0.707823 smol_offdiag=0.660557 prompts=4 mcp_tok=17.50 smol_tok=17.50 +2026-05-10 15:00:23,372 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 15:00:42,053 - INFO - Step 4210 | mode=video loss=0.108887 diff=0.108887 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0333 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0887 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=3369 i_micro=841 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3467 mask_tok=104.00 y_norm=48.0887±0.0055 +2026-05-10 15:01:00,821 - INFO - Step 4220 | mode=video loss=0.093262 diff=0.093262 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0309 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0865 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=3377 i_micro=843 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.2867 mask_tok=86.00 y_norm=48.0865±0.0060 +2026-05-10 15:01:18,776 - INFO - Step 4230 | mode=video loss=0.085938 diff=0.085938 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0918 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0881 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=3385 i_micro=845 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4167 mask_tok=125.00 y_norm=48.0881±0.0057 +2026-05-10 15:01:37,428 - INFO - Step 4240 | mode=video loss=0.123535 diff=0.123535 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0758 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0861 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=3393 i_micro=847 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3467 mask_tok=104.00 y_norm=48.0861±0.0059 +2026-05-10 15:01:56,101 - INFO - Step 4250 | mode=video loss=0.097656 diff=0.097656 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0501 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0870 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=3401 i_micro=849 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4600 mask_tok=138.00 y_norm=48.0870±0.0060 +2026-05-10 15:02:14,665 - INFO - Step 4260 | mode=video loss=0.121582 diff=0.121582 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0663 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0879 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=3409 i_micro=851 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3700 mask_tok=111.00 y_norm=48.0879±0.0060 +2026-05-10 15:02:33,408 - INFO - Step 4270 | mode=video loss=0.087891 diff=0.087891 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0277 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0879 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=3417 i_micro=853 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.2500 mask_tok=75.00 y_norm=48.0879±0.0048 +2026-05-10 15:02:51,368 - INFO - Step 4280 | mode=video loss=0.096191 diff=0.096191 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0434 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0888 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=3425 i_micro=855 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.2967 mask_tok=89.00 y_norm=48.0888±0.0056 +2026-05-10 15:03:09,758 - INFO - Step 4290 | mode=video loss=0.077148 diff=0.077148 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0281 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0882 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=3433 i_micro=857 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3267 mask_tok=98.00 y_norm=48.0882±0.0053 +2026-05-10 15:03:28,479 - INFO - Step 4300 | mode=video loss=0.052246 diff=0.052246 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0473 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0877 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=3441 i_micro=859 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4233 mask_tok=127.00 y_norm=48.0877±0.0062 +2026-05-10 15:03:35,627 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 15:03:54,000 - INFO - Step 4310 | mode=video loss=0.095215 diff=0.095215 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0532 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0853 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=3449 i_micro=861 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4200 mask_tok=126.00 y_norm=48.0853±0.0059 +2026-05-10 15:04:12,742 - INFO - Step 4320 | mode=video loss=0.064941 diff=0.064941 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0391 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0881 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=3457 i_micro=863 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3400 mask_tok=102.00 y_norm=48.0881±0.0057 +2026-05-10 15:04:30,737 - INFO - Step 4330 | mode=video loss=0.082520 diff=0.082520 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0283 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0874 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=3465 i_micro=865 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3267 mask_tok=98.00 y_norm=48.0874±0.0054 +2026-05-10 15:04:49,271 - INFO - Step 4340 | mode=video loss=0.131836 diff=0.131836 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0434 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0860 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=3473 i_micro=867 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4667 mask_tok=140.00 y_norm=48.0860±0.0061 +2026-05-10 15:05:07,920 - INFO - Step 4350 | mode=video loss=0.099609 diff=0.099609 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0617 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0873 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=3481 i_micro=869 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4100 mask_tok=123.00 y_norm=48.0873±0.0062 +2026-05-10 15:05:25,888 - INFO - Step 4360 | mode=video loss=0.083008 diff=0.083008 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0270 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0869 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=3489 i_micro=871 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4200 mask_tok=126.00 y_norm=48.0869±0.0061 +2026-05-10 15:05:44,342 - INFO - Step 4370 | mode=video loss=0.105469 diff=0.105469 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0791 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0869 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=3497 i_micro=873 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3033 mask_tok=91.00 y_norm=48.0869±0.0049 +2026-05-10 15:06:02,868 - INFO - Step 4380 | mode=video loss=0.112305 diff=0.112305 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0645 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0872 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=3505 i_micro=875 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4167 mask_tok=125.00 y_norm=48.0872±0.0052 +2026-05-10 15:06:21,435 - INFO - Step 4390 | mode=video loss=0.114746 diff=0.114746 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0458 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0884 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=3513 i_micro=877 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3267 mask_tok=98.00 y_norm=48.0884±0.0055 +2026-05-10 15:06:41,543 - INFO - Step 4400 | mode=video loss=0.062500 diff=0.062500 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0346 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0891 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=3521 i_micro=879 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3333 mask_tok=100.00 y_norm=48.0891±0.0052 cond_uncond_dloss=0.212891 cond_grad=0.000135 cond_pred_l2=35.555058 cond_pred_ratio=0.021197 +2026-05-10 15:06:41,580 - INFO - Step 4400 | probe_semantic mcp_offdiag(mean/min/max)=0.645513/0.533117/0.708654 smol_offdiag=0.660557 prompts=4 mcp_tok=17.50 smol_tok=17.50 +2026-05-10 15:06:49,127 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 15:07:07,103 - INFO - Step 4410 | mode=video loss=0.071289 diff=0.071289 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1062 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0871 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=3529 i_micro=881 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4400 mask_tok=132.00 y_norm=48.0871±0.0062 +2026-05-10 15:07:25,643 - INFO - Step 4420 | mode=video loss=0.141602 diff=0.141602 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0596 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0879 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=3537 i_micro=883 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3767 mask_tok=113.00 y_norm=48.0879±0.0052 +2026-05-10 15:07:44,386 - INFO - Step 4430 | mode=video loss=0.102051 diff=0.102051 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0601 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0878 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=3545 i_micro=885 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3533 mask_tok=106.00 y_norm=48.0878±0.0052 +2026-05-10 15:08:02,811 - INFO - Step 4440 | mode=video loss=0.051270 diff=0.051270 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0377 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0859 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=3553 i_micro=887 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4733 mask_tok=142.00 y_norm=48.0859±0.0062 +2026-05-10 15:08:21,481 - INFO - Step 4450 | mode=video loss=0.059082 diff=0.059082 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0775 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0867 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=3561 i_micro=889 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3467 mask_tok=104.00 y_norm=48.0867±0.0049 +2026-05-10 15:08:39,440 - INFO - Step 4460 | mode=video loss=0.125977 diff=0.125977 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1172 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0879 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=3569 i_micro=891 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3600 mask_tok=108.00 y_norm=48.0879±0.0063 +2026-05-10 15:08:57,837 - INFO - Step 4470 | mode=video loss=0.073730 diff=0.073730 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0455 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0879 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=3577 i_micro=893 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3667 mask_tok=110.00 y_norm=48.0879±0.0059 +2026-05-10 15:09:16,493 - INFO - Step 4480 | mode=video loss=0.061523 diff=0.061523 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0419 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0868 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=3585 i_micro=895 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3733 mask_tok=112.00 y_norm=48.0868±0.0057 +2026-05-10 15:09:34,441 - INFO - Step 4490 | mode=video loss=0.155273 diff=0.155273 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1303 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0874 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=3593 i_micro=897 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4367 mask_tok=131.00 y_norm=48.0874±0.0065 +2026-05-10 15:09:52,820 - INFO - Step 4500 | mode=video loss=0.298828 diff=0.298828 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.3128 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0877 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=3601 i_micro=899 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3600 mask_tok=108.00 y_norm=48.0877±0.0060 +2026-05-10 15:09:59,693 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 15:10:17,632 - INFO - Step 4510 | mode=video loss=0.103027 diff=0.103027 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0770 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0864 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=3609 i_micro=901 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4300 mask_tok=129.00 y_norm=48.0864±0.0066 +2026-05-10 15:10:36,015 - INFO - Step 4520 | mode=video loss=0.074707 diff=0.074707 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0581 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0881 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=3617 i_micro=903 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3267 mask_tok=98.00 y_norm=48.0881±0.0058 +2026-05-10 15:10:54,795 - INFO - Step 4530 | mode=video loss=0.110840 diff=0.110840 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0284 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0852 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=3625 i_micro=905 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4167 mask_tok=125.00 y_norm=48.0852±0.0061 +2026-05-10 15:11:12,766 - INFO - Step 4540 | mode=video loss=0.061035 diff=0.061035 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0461 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0867 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=3633 i_micro=907 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3667 mask_tok=110.00 y_norm=48.0867±0.0057 +2026-05-10 15:11:31,242 - INFO - Step 4550 | mode=video loss=0.092285 diff=0.092285 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.2265 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0871 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=3641 i_micro=909 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3600 mask_tok=108.00 y_norm=48.0871±0.0062 +2026-05-10 15:11:49,920 - INFO - Step 4560 | mode=video loss=0.105469 diff=0.105469 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0375 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0856 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=3649 i_micro=911 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4933 mask_tok=148.00 y_norm=48.0856±0.0064 +2026-05-10 15:12:07,874 - INFO - Step 4570 | mode=video loss=0.125977 diff=0.125977 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0477 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0874 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=3657 i_micro=913 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3600 mask_tok=108.00 y_norm=48.0874±0.0055 +2026-05-10 15:12:26,899 - INFO - Step 4580 | mode=video loss=0.076660 diff=0.076660 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0504 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0876 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=3665 i_micro=915 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.2833 mask_tok=85.00 y_norm=48.0876±0.0054 +2026-05-10 15:12:44,857 - INFO - Step 4590 | mode=video loss=0.064453 diff=0.064453 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0368 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0841 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=3673 i_micro=917 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3500 mask_tok=105.00 y_norm=48.0841±0.0063 +2026-05-10 15:13:05,076 - INFO - Step 4600 | mode=video loss=0.054932 diff=0.054932 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0711 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0866 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=3681 i_micro=919 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3767 mask_tok=113.00 y_norm=48.0866±0.0054 cond_uncond_dloss=0.832031 cond_grad=0.000179 cond_pred_l2=44.563931 cond_pred_ratio=0.026671 +2026-05-10 15:13:05,113 - INFO - Step 4600 | probe_semantic mcp_offdiag(mean/min/max)=0.644932/0.532460/0.707883 smol_offdiag=0.660557 prompts=4 mcp_tok=17.50 smol_tok=17.50 +2026-05-10 15:13:11,972 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 15:13:30,687 - INFO - Step 4610 | mode=video loss=0.066406 diff=0.066406 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0409 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0862 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=3689 i_micro=921 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3600 mask_tok=108.00 y_norm=48.0862±0.0054 +2026-05-10 15:13:48,646 - INFO - Step 4620 | mode=video loss=0.089355 diff=0.089355 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0638 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0878 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=3697 i_micro=923 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3467 mask_tok=104.00 y_norm=48.0878±0.0050 +2026-05-10 15:14:07,830 - INFO - Step 4630 | mode=video loss=0.103027 diff=0.103027 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0359 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0887 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=3705 i_micro=925 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3300 mask_tok=99.00 y_norm=48.0887±0.0053 +2026-05-10 15:14:25,793 - INFO - Step 4640 | mode=video loss=0.084961 diff=0.084961 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0266 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0872 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=3713 i_micro=927 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3567 mask_tok=107.00 y_norm=48.0872±0.0062 +2026-05-10 15:14:44,298 - INFO - Step 4650 | mode=video loss=0.094238 diff=0.094238 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0524 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0879 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=3721 i_micro=929 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3400 mask_tok=102.00 y_norm=48.0879±0.0053 +2026-05-10 15:15:02,943 - INFO - Step 4660 | mode=video loss=0.082031 diff=0.082031 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0460 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0882 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=3729 i_micro=931 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3500 mask_tok=105.00 y_norm=48.0882±0.0052 +2026-05-10 15:15:20,902 - INFO - Step 4670 | mode=video loss=0.070801 diff=0.070801 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0448 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0862 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=3737 i_micro=933 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3967 mask_tok=119.00 y_norm=48.0862±0.0059 +2026-05-10 15:15:39,267 - INFO - Step 4680 | mode=video loss=0.082031 diff=0.082031 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0688 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0855 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=3745 i_micro=935 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3600 mask_tok=108.00 y_norm=48.0855±0.0056 +2026-05-10 15:15:57,916 - INFO - Step 4690 | mode=video loss=0.071777 diff=0.071777 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0348 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0863 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=3753 i_micro=937 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3633 mask_tok=109.00 y_norm=48.0863±0.0051 +2026-05-10 15:16:15,861 - INFO - Step 4700 | mode=video loss=0.083496 diff=0.083496 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0534 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0876 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=3761 i_micro=939 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.8500 mask_tok=255.00 y_norm=48.0876±0.0077 +2026-05-10 15:16:22,801 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 15:16:41,891 - INFO - Step 4710 | mode=video loss=0.100586 diff=0.100586 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0531 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0880 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=3769 i_micro=941 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3900 mask_tok=117.00 y_norm=48.0880±0.0057 +2026-05-10 15:16:59,848 - INFO - Step 4720 | mode=video loss=0.144531 diff=0.144531 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1354 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0894 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=3777 i_micro=943 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.2900 mask_tok=87.00 y_norm=48.0894±0.0050 +2026-05-10 15:17:18,244 - INFO - Step 4730 | mode=video loss=0.099121 diff=0.099121 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0546 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0883 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=3785 i_micro=945 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.2800 mask_tok=84.00 y_norm=48.0883±0.0052 +2026-05-10 15:17:36,742 - INFO - Step 4740 | mode=video loss=0.069824 diff=0.069824 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0575 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0877 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=3793 i_micro=947 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4067 mask_tok=122.00 y_norm=48.0877±0.0065 +2026-05-10 15:17:54,735 - INFO - Step 4750 | mode=video loss=0.065918 diff=0.065918 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0487 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0876 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=3801 i_micro=949 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3700 mask_tok=111.00 y_norm=48.0876±0.0055 +2026-05-10 15:18:13,760 - INFO - Step 4760 | mode=video loss=0.139648 diff=0.139648 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0971 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0878 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=3809 i_micro=951 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3633 mask_tok=109.00 y_norm=48.0878±0.0054 +2026-05-10 15:18:31,712 - INFO - Step 4770 | mode=video loss=0.113770 diff=0.113770 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0675 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0866 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=3817 i_micro=953 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4200 mask_tok=126.00 y_norm=48.0866±0.0056 +2026-05-10 15:18:50,082 - INFO - Step 4780 | mode=video loss=0.109863 diff=0.109863 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0470 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0877 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=3825 i_micro=955 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3733 mask_tok=112.00 y_norm=48.0877±0.0054 +2026-05-10 15:19:08,666 - INFO - Step 4790 | mode=video loss=0.083496 diff=0.083496 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0516 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0887 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=3833 i_micro=957 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3533 mask_tok=106.00 y_norm=48.0887±0.0056 +2026-05-10 15:19:28,248 - INFO - Step 4800 | mode=video loss=0.059326 diff=0.059326 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0353 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0856 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=3841 i_micro=959 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4067 mask_tok=122.00 y_norm=48.0856±0.0061 cond_uncond_dloss=0.941406 cond_grad=0.000069 cond_pred_l2=33.888435 cond_pred_ratio=0.019221 +2026-05-10 15:19:28,285 - INFO - Step 4800 | probe_semantic mcp_offdiag(mean/min/max)=0.646616/0.534660/0.708811 smol_offdiag=0.660557 prompts=4 mcp_tok=17.50 smol_tok=17.50 +2026-05-10 15:19:35,057 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 15:19:54,251 - INFO - Step 4810 | mode=video loss=0.098633 diff=0.098633 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0488 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0859 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=3849 i_micro=961 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4833 mask_tok=145.00 y_norm=48.0859±0.0064 +2026-05-10 15:20:12,209 - INFO - Step 4820 | mode=video loss=0.068359 diff=0.068359 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0326 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0876 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=3857 i_micro=963 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3467 mask_tok=104.00 y_norm=48.0876±0.0056 +2026-05-10 15:20:30,177 - INFO - Step 4830 | mode=video loss=0.089844 diff=0.089844 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0840 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0887 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=3865 i_micro=965 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3167 mask_tok=95.00 y_norm=48.0887±0.0049 +2026-05-10 15:20:49,362 - INFO - Step 4840 | mode=video loss=0.111816 diff=0.111816 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0526 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0868 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=3873 i_micro=967 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4833 mask_tok=145.00 y_norm=48.0868±0.0062 +2026-05-10 15:21:07,321 - INFO - Step 4850 | mode=video loss=0.059814 diff=0.059814 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0559 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0900 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=3881 i_micro=969 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3300 mask_tok=99.00 y_norm=48.0900±0.0059 +2026-05-10 15:21:25,694 - INFO - Step 4860 | mode=video loss=0.068359 diff=0.068359 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1337 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0864 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=3889 i_micro=971 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.5133 mask_tok=154.00 y_norm=48.0864±0.0067 +2026-05-10 15:21:44,368 - INFO - Step 4870 | mode=video loss=0.100098 diff=0.100098 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0247 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0871 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=3897 i_micro=973 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3867 mask_tok=116.00 y_norm=48.0871±0.0060 +2026-05-10 15:22:02,355 - INFO - Step 4880 | mode=video loss=0.102539 diff=0.102539 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0918 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0890 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=3905 i_micro=975 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3367 mask_tok=101.00 y_norm=48.0890±0.0055 +2026-05-10 15:22:21,511 - INFO - Step 4890 | mode=video loss=0.172852 diff=0.172852 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1001 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0883 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=3913 i_micro=977 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3167 mask_tok=95.00 y_norm=48.0883±0.0050 +2026-05-10 15:22:39,450 - INFO - Step 4900 | mode=video loss=0.044434 diff=0.044434 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0655 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0867 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=3921 i_micro=979 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3333 mask_tok=100.00 y_norm=48.0867±0.0055 +2026-05-10 15:22:46,249 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 15:23:04,650 - INFO - Step 4910 | mode=video loss=0.130859 diff=0.130859 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0267 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0854 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=3929 i_micro=981 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4300 mask_tok=129.00 y_norm=48.0854±0.0063 +2026-05-10 15:23:23,420 - INFO - Step 4920 | mode=video loss=0.088867 diff=0.088867 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0670 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0875 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=3937 i_micro=983 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4433 mask_tok=133.00 y_norm=48.0875±0.0065 +2026-05-10 15:23:41,378 - INFO - Step 4930 | mode=video loss=0.092773 diff=0.092773 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0391 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0891 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=3945 i_micro=985 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3733 mask_tok=112.00 y_norm=48.0891±0.0056 +2026-05-10 15:24:00,702 - INFO - Step 4940 | mode=video loss=0.080078 diff=0.080078 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1310 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0883 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=3953 i_micro=987 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3633 mask_tok=109.00 y_norm=48.0883±0.0066 +2026-05-10 15:24:18,667 - INFO - Step 4950 | mode=video loss=0.068359 diff=0.068359 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0468 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0865 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=3961 i_micro=989 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4167 mask_tok=125.00 y_norm=48.0865±0.0058 +2026-05-10 15:24:37,243 - INFO - Step 4960 | mode=video loss=0.156250 diff=0.156250 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0679 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0887 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=3969 i_micro=991 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3900 mask_tok=117.00 y_norm=48.0887±0.0055 +2026-05-10 15:24:55,944 - INFO - Step 4970 | mode=video loss=0.103027 diff=0.103027 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0447 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0869 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=3977 i_micro=993 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4233 mask_tok=127.00 y_norm=48.0869±0.0056 +2026-05-10 15:25:13,889 - INFO - Step 4980 | mode=video loss=0.066406 diff=0.066406 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0439 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0870 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=3985 i_micro=995 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3300 mask_tok=99.00 y_norm=48.0870±0.0054 +2026-05-10 15:25:33,086 - INFO - Step 4990 | mode=video loss=0.096191 diff=0.096191 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0379 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0872 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=3993 i_micro=997 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.2567 mask_tok=77.00 y_norm=48.0872±0.0054 +2026-05-10 15:25:52,673 - INFO - Step 5000 | mode=video loss=0.086914 diff=0.086914 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0440 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0862 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=4001 i_micro=999 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3367 mask_tok=101.00 y_norm=48.0862±0.0059 cond_uncond_dloss=0.328125 cond_grad=0.000059 cond_pred_l2=44.530571 cond_pred_ratio=0.025426 +2026-05-10 15:25:52,711 - INFO - Step 5000 | probe_semantic mcp_offdiag(mean/min/max)=0.646494/0.534596/0.708531 smol_offdiag=0.660557 prompts=4 mcp_tok=17.50 smol_tok=17.50 +2026-05-10 15:25:58,971 - INFO - Saved checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_step5000.pt +2026-05-10 15:26:03,858 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 15:26:21,788 - INFO - Step 5010 | mode=video loss=0.060791 diff=0.060791 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0516 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0886 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=4009 i_micro=1001 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3400 mask_tok=102.00 y_norm=48.0886±0.0053 +2026-05-10 15:26:41,133 - INFO - Step 5020 | mode=video loss=0.066895 diff=0.066895 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0400 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0865 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=4017 i_micro=1003 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3367 mask_tok=101.00 y_norm=48.0865±0.0064 +2026-05-10 15:26:59,093 - INFO - Step 5030 | mode=video loss=0.107910 diff=0.107910 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0595 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0864 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=4025 i_micro=1005 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4367 mask_tok=131.00 y_norm=48.0864±0.0062 +2026-05-10 15:27:17,677 - INFO - Step 5040 | mode=video loss=0.121094 diff=0.121094 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0521 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0876 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=4033 i_micro=1007 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3333 mask_tok=100.00 y_norm=48.0876±0.0055 +2026-05-10 15:27:36,317 - INFO - Step 5050 | mode=video loss=0.083984 diff=0.083984 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0417 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0882 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=4041 i_micro=1009 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3433 mask_tok=103.00 y_norm=48.0882±0.0060 +2026-05-10 15:27:54,269 - INFO - Step 5060 | mode=video loss=0.084473 diff=0.084473 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0285 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0879 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=4049 i_micro=1011 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3433 mask_tok=103.00 y_norm=48.0879±0.0048 +2026-05-10 15:28:13,527 - INFO - Step 5070 | mode=video loss=0.090820 diff=0.090820 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0588 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0875 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=4057 i_micro=1013 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3533 mask_tok=106.00 y_norm=48.0875±0.0051 +2026-05-10 15:28:31,483 - INFO - Step 5080 | mode=video loss=0.106445 diff=0.106445 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0526 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0885 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=4065 i_micro=1015 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3533 mask_tok=106.00 y_norm=48.0885±0.0052 +2026-05-10 15:28:49,442 - INFO - Step 5090 | mode=video loss=0.104004 diff=0.104004 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0374 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0890 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=4073 i_micro=1017 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3100 mask_tok=93.00 y_norm=48.0890±0.0052 +2026-05-10 15:29:08,684 - INFO - Step 5100 | mode=video loss=0.030029 diff=0.030029 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0594 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0862 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=4081 i_micro=1019 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4567 mask_tok=137.00 y_norm=48.0862±0.0066 +2026-05-10 15:29:15,658 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 15:29:33,604 - INFO - Step 5110 | mode=video loss=0.085449 diff=0.085449 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0336 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0880 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=4089 i_micro=1021 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4800 mask_tok=144.00 y_norm=48.0880±0.0061 +2026-05-10 15:29:52,821 - INFO - Step 5120 | mode=video loss=0.141602 diff=0.141602 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1142 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0876 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=4097 i_micro=1023 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4500 mask_tok=135.00 y_norm=48.0876±0.0063 +2026-05-10 15:30:10,768 - INFO - Step 5130 | mode=video loss=0.076660 diff=0.076660 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0698 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0860 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=4105 i_micro=1025 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3467 mask_tok=104.00 y_norm=48.0860±0.0063 +2026-05-10 15:30:28,726 - INFO - Step 5140 | mode=video loss=0.123535 diff=0.123535 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0576 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0859 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=4113 i_micro=1027 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3833 mask_tok=115.00 y_norm=48.0859±0.0069 +2026-05-10 15:30:47,662 - INFO - Step 5150 | mode=video loss=0.086914 diff=0.086914 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0300 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0874 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=4121 i_micro=1029 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3767 mask_tok=113.00 y_norm=48.0874±0.0061 +2026-05-10 15:31:05,618 - INFO - Step 5160 | mode=video loss=0.129883 diff=0.129883 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0463 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0876 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=4129 i_micro=1031 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3967 mask_tok=119.00 y_norm=48.0876±0.0069 +2026-05-10 15:31:24,026 - INFO - Step 5170 | mode=video loss=0.148438 diff=0.148438 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0651 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0892 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=4137 i_micro=1033 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3267 mask_tok=98.00 y_norm=48.0892±0.0060 +2026-05-10 15:31:42,586 - INFO - Step 5180 | mode=video loss=0.091309 diff=0.091309 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0272 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0883 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=4145 i_micro=1035 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3167 mask_tok=95.00 y_norm=48.0883±0.0058 +2026-05-10 15:32:00,531 - INFO - Step 5190 | mode=video loss=0.116211 diff=0.116211 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0950 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0859 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=4153 i_micro=1037 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3767 mask_tok=113.00 y_norm=48.0859±0.0061 +2026-05-10 15:32:21,292 - INFO - Step 5200 | mode=video loss=0.172852 diff=0.172852 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0470 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0864 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=4161 i_micro=1039 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4000 mask_tok=120.00 y_norm=48.0864±0.0060 cond_uncond_dloss=0.445312 cond_grad=0.000276 cond_pred_l2=53.313972 cond_pred_ratio=0.034296 +2026-05-10 15:32:21,329 - INFO - Step 5200 | probe_semantic mcp_offdiag(mean/min/max)=0.646248/0.533781/0.708256 smol_offdiag=0.660557 prompts=4 mcp_tok=17.50 smol_tok=17.50 +2026-05-10 15:32:28,163 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 15:32:46,095 - INFO - Step 5210 | mode=video loss=0.056152 diff=0.056152 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0340 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0880 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=4169 i_micro=1041 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3133 mask_tok=94.00 y_norm=48.0880±0.0052 +2026-05-10 15:33:04,577 - INFO - Step 5220 | mode=video loss=0.065430 diff=0.065430 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0428 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0874 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=4177 i_micro=1043 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3433 mask_tok=103.00 y_norm=48.0874±0.0054 +2026-05-10 15:33:23,314 - INFO - Step 5230 | mode=video loss=0.098145 diff=0.098145 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0466 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0884 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=4185 i_micro=1045 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3767 mask_tok=113.00 y_norm=48.0884±0.0058 +2026-05-10 15:33:41,276 - INFO - Step 5240 | mode=video loss=0.086426 diff=0.086426 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0352 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0883 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=4193 i_micro=1047 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3600 mask_tok=108.00 y_norm=48.0883±0.0055 +2026-05-10 15:34:00,543 - INFO - Step 5250 | mode=video loss=0.080078 diff=0.080078 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0396 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0871 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=4201 i_micro=1049 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3700 mask_tok=111.00 y_norm=48.0871±0.0068 +2026-05-10 15:34:18,488 - INFO - Step 5260 | mode=video loss=0.108398 diff=0.108398 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0436 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0894 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=4209 i_micro=1051 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3533 mask_tok=106.00 y_norm=48.0894±0.0063 +2026-05-10 15:34:36,445 - INFO - Step 5270 | mode=video loss=0.127930 diff=0.127930 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0395 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0877 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=4217 i_micro=1053 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3333 mask_tok=100.00 y_norm=48.0877±0.0060 +2026-05-10 15:34:55,632 - INFO - Step 5280 | mode=video loss=0.158203 diff=0.158203 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0737 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0882 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=4225 i_micro=1055 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3333 mask_tok=100.00 y_norm=48.0882±0.0052 +2026-05-10 15:35:13,584 - INFO - Step 5290 | mode=video loss=0.073242 diff=0.073242 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0348 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0888 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=4233 i_micro=1057 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3333 mask_tok=100.00 y_norm=48.0888±0.0054 +2026-05-10 15:35:31,965 - INFO - Step 5300 | mode=video loss=0.083496 diff=0.083496 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0689 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0874 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=4241 i_micro=1059 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3533 mask_tok=106.00 y_norm=48.0874±0.0056 +2026-05-10 15:35:38,995 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 15:35:57,588 - INFO - Step 5310 | mode=video loss=0.089844 diff=0.089844 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0714 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0866 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=4249 i_micro=1061 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3700 mask_tok=111.00 y_norm=48.0866±0.0059 +2026-05-10 15:36:15,537 - INFO - Step 5320 | mode=video loss=0.111328 diff=0.111328 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0701 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0881 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=4257 i_micro=1063 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3800 mask_tok=114.00 y_norm=48.0881±0.0059 +2026-05-10 15:36:34,618 - INFO - Step 5330 | mode=video loss=0.045654 diff=0.045654 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0436 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0858 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=4265 i_micro=1065 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3700 mask_tok=111.00 y_norm=48.0858±0.0056 +2026-05-10 15:36:52,572 - INFO - Step 5340 | mode=video loss=0.110840 diff=0.110840 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0550 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0878 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=4273 i_micro=1067 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3500 mask_tok=105.00 y_norm=48.0878±0.0055 +2026-05-10 15:37:10,984 - INFO - Step 5350 | mode=video loss=0.244141 diff=0.244141 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.2022 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0876 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=4281 i_micro=1069 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3400 mask_tok=102.00 y_norm=48.0876±0.0057 +2026-05-10 15:37:29,608 - INFO - Step 5360 | mode=video loss=0.141602 diff=0.141602 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1345 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0889 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=4289 i_micro=1071 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3100 mask_tok=93.00 y_norm=48.0889±0.0052 +2026-05-10 15:37:47,555 - INFO - Step 5370 | mode=video loss=0.131836 diff=0.131836 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0525 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0888 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=4297 i_micro=1073 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4233 mask_tok=127.00 y_norm=48.0888±0.0066 +2026-05-10 15:38:06,480 - INFO - Step 5380 | mode=video loss=0.075195 diff=0.075195 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0771 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0885 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=4305 i_micro=1075 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3500 mask_tok=105.00 y_norm=48.0885±0.0056 +2026-05-10 15:38:24,424 - INFO - Step 5390 | mode=video loss=0.166016 diff=0.166016 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0773 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0888 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=4313 i_micro=1077 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3600 mask_tok=108.00 y_norm=48.0888±0.0063 +2026-05-10 15:38:43,996 - INFO - Step 5400 | mode=video loss=0.179688 diff=0.179688 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0991 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0872 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=4321 i_micro=1079 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3433 mask_tok=103.00 y_norm=48.0872±0.0059 cond_uncond_dloss=0.617188 cond_grad=0.001013 cond_pred_l2=76.009850 cond_pred_ratio=0.049201 +2026-05-10 15:38:44,034 - INFO - Step 5400 | probe_semantic mcp_offdiag(mean/min/max)=0.646870/0.534595/0.708793 smol_offdiag=0.660557 prompts=4 mcp_tok=17.50 smol_tok=17.50 +2026-05-10 15:38:51,132 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 15:39:10,241 - INFO - Step 5410 | mode=video loss=0.106934 diff=0.106934 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0526 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0883 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=4329 i_micro=1081 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.5000 mask_tok=150.00 y_norm=48.0883±0.0068 +2026-05-10 15:39:28,200 - INFO - Step 5420 | mode=video loss=0.135742 diff=0.135742 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1092 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0881 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=4337 i_micro=1083 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3467 mask_tok=104.00 y_norm=48.0881±0.0061 +2026-05-10 15:39:47,332 - INFO - Step 5430 | mode=video loss=0.083496 diff=0.083496 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0525 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0899 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=4345 i_micro=1085 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3400 mask_tok=102.00 y_norm=48.0899±0.0059 +2026-05-10 15:40:05,294 - INFO - Step 5440 | mode=video loss=0.070312 diff=0.070312 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1072 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0884 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=4353 i_micro=1087 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3367 mask_tok=101.00 y_norm=48.0884±0.0068 +2026-05-10 15:40:23,259 - INFO - Step 5450 | mode=video loss=0.098145 diff=0.098145 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0346 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0884 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=4361 i_micro=1089 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4400 mask_tok=132.00 y_norm=48.0884±0.0063 +2026-05-10 15:40:42,465 - INFO - Step 5460 | mode=video loss=0.141602 diff=0.141602 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0743 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0869 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=4369 i_micro=1091 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3967 mask_tok=119.00 y_norm=48.0869±0.0055 +2026-05-10 15:41:00,419 - INFO - Step 5470 | mode=video loss=0.090332 diff=0.090332 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0693 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0883 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=4377 i_micro=1093 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4833 mask_tok=145.00 y_norm=48.0883±0.0070 +2026-05-10 15:41:18,826 - INFO - Step 5480 | mode=video loss=0.078125 diff=0.078125 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0560 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0883 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=4385 i_micro=1095 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3333 mask_tok=100.00 y_norm=48.0883±0.0058 +2026-05-10 15:41:37,399 - INFO - Step 5490 | mode=video loss=0.085938 diff=0.085938 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0389 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0884 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=4393 i_micro=1097 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3933 mask_tok=118.00 y_norm=48.0884±0.0063 +2026-05-10 15:41:55,373 - INFO - Step 5500 | mode=video loss=0.068359 diff=0.068359 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0376 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0887 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=4401 i_micro=1099 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3800 mask_tok=114.00 y_norm=48.0887±0.0062 +2026-05-10 15:42:02,151 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 15:42:21,238 - INFO - Step 5510 | mode=video loss=0.129883 diff=0.129883 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1269 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0886 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=4409 i_micro=1101 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3000 mask_tok=90.00 y_norm=48.0886±0.0055 +2026-05-10 15:42:39,188 - INFO - Step 5520 | mode=video loss=0.099121 diff=0.099121 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0555 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0897 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=4417 i_micro=1103 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3867 mask_tok=116.00 y_norm=48.0897±0.0059 +2026-05-10 15:42:57,153 - INFO - Step 5530 | mode=video loss=0.105957 diff=0.105957 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0478 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0885 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=4425 i_micro=1105 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3367 mask_tok=101.00 y_norm=48.0885±0.0055 +2026-05-10 15:43:16,263 - INFO - Step 5540 | mode=video loss=0.215820 diff=0.215820 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.3505 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0898 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=4433 i_micro=1107 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3367 mask_tok=101.00 y_norm=48.0898±0.0048 +2026-05-10 15:43:34,222 - INFO - Step 5550 | mode=video loss=0.059570 diff=0.059570 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0884 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0883 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=4441 i_micro=1109 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4300 mask_tok=129.00 y_norm=48.0883±0.0065 +2026-05-10 15:43:53,243 - INFO - Step 5560 | mode=video loss=0.113770 diff=0.113770 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0745 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0866 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=4449 i_micro=1111 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3833 mask_tok=115.00 y_norm=48.0866±0.0057 +2026-05-10 15:44:11,196 - INFO - Step 5570 | mode=video loss=0.108398 diff=0.108398 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0515 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0878 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=4457 i_micro=1113 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4133 mask_tok=124.00 y_norm=48.0878±0.0065 +2026-05-10 15:44:29,168 - INFO - Step 5580 | mode=video loss=0.103027 diff=0.103027 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0281 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0858 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=4465 i_micro=1115 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3900 mask_tok=117.00 y_norm=48.0858±0.0061 +2026-05-10 15:44:48,343 - INFO - Step 5590 | mode=video loss=0.062500 diff=0.062500 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0331 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0871 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=4473 i_micro=1117 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3767 mask_tok=113.00 y_norm=48.0871±0.0062 +2026-05-10 15:45:07,922 - INFO - Step 5600 | mode=video loss=0.058350 diff=0.058350 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0345 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0867 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=4481 i_micro=1119 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4200 mask_tok=126.00 y_norm=48.0867±0.0058 cond_uncond_dloss=0.484375 cond_grad=0.000256 cond_pred_l2=38.169800 cond_pred_ratio=0.023468 +2026-05-10 15:45:07,960 - INFO - Step 5600 | probe_semantic mcp_offdiag(mean/min/max)=0.646328/0.533769/0.708430 smol_offdiag=0.660557 prompts=4 mcp_tok=17.50 smol_tok=17.50 +2026-05-10 15:45:15,599 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 15:45:33,535 - INFO - Step 5610 | mode=video loss=0.130859 diff=0.130859 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0678 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0878 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=4489 i_micro=1121 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3467 mask_tok=104.00 y_norm=48.0878±0.0058 +2026-05-10 15:45:52,877 - INFO - Step 5620 | mode=video loss=0.080078 diff=0.080078 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0708 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0869 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=4497 i_micro=1123 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3700 mask_tok=111.00 y_norm=48.0869±0.0059 +2026-05-10 15:46:10,848 - INFO - Step 5630 | mode=video loss=0.152344 diff=0.152344 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0672 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0874 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=4505 i_micro=1125 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3433 mask_tok=103.00 y_norm=48.0874±0.0054 +2026-05-10 15:46:29,993 - INFO - Step 5640 | mode=video loss=0.142578 diff=0.142578 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1550 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0878 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=4513 i_micro=1127 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3567 mask_tok=107.00 y_norm=48.0878±0.0052 +2026-05-10 15:46:47,950 - INFO - Step 5650 | mode=video loss=0.078613 diff=0.078613 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0303 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0884 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=4521 i_micro=1129 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.2367 mask_tok=71.00 y_norm=48.0884±0.0054 +2026-05-10 15:47:05,911 - INFO - Step 5660 | mode=video loss=0.004700 diff=0.004700 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0650 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0869 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=4529 i_micro=1131 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4333 mask_tok=130.00 y_norm=48.0869±0.0059 +2026-05-10 15:47:24,999 - INFO - Step 5670 | mode=video loss=0.163086 diff=0.163086 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0524 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0886 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=4537 i_micro=1133 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3433 mask_tok=103.00 y_norm=48.0886±0.0060 +2026-05-10 15:47:42,948 - INFO - Step 5680 | mode=video loss=0.086914 diff=0.086914 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0204 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0881 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=4545 i_micro=1135 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3267 mask_tok=98.00 y_norm=48.0881±0.0059 +2026-05-10 15:48:02,031 - INFO - Step 5690 | mode=video loss=0.134766 diff=0.134766 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0242 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0879 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=4553 i_micro=1137 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3200 mask_tok=96.00 y_norm=48.0879±0.0052 +2026-05-10 15:48:19,986 - INFO - Step 5700 | mode=video loss=0.127930 diff=0.127930 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0805 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0896 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=4561 i_micro=1139 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.1900 mask_tok=57.00 y_norm=48.0896±0.0048 +2026-05-10 15:48:27,015 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 15:48:44,968 - INFO - Step 5710 | mode=video loss=0.166016 diff=0.166016 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0603 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0882 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=4569 i_micro=1141 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3400 mask_tok=102.00 y_norm=48.0882±0.0054 +2026-05-10 15:49:03,758 - INFO - Step 5720 | mode=video loss=0.236328 diff=0.236328 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0384 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0889 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=4577 i_micro=1143 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3300 mask_tok=99.00 y_norm=48.0889±0.0053 +2026-05-10 15:49:21,727 - INFO - Step 5730 | mode=video loss=0.102539 diff=0.102539 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0423 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0896 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=4585 i_micro=1145 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3467 mask_tok=104.00 y_norm=48.0896±0.0054 +2026-05-10 15:49:39,699 - INFO - Step 5740 | mode=video loss=0.070801 diff=0.070801 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0578 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0891 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=4593 i_micro=1147 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.8867 mask_tok=266.00 y_norm=48.0891±0.0085 +2026-05-10 15:49:58,943 - INFO - Step 5750 | mode=video loss=0.075195 diff=0.075195 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0810 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0892 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=4601 i_micro=1149 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3300 mask_tok=99.00 y_norm=48.0892±0.0054 +2026-05-10 15:50:16,907 - INFO - Step 5760 | mode=video loss=0.198242 diff=0.198242 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1966 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0880 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=4609 i_micro=1151 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.5200 mask_tok=156.00 y_norm=48.0880±0.0067 +2026-05-10 15:50:35,910 - INFO - Step 5770 | mode=video loss=0.098145 diff=0.098145 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0448 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0883 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=4617 i_micro=1153 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3533 mask_tok=106.00 y_norm=48.0883±0.0057 +2026-05-10 15:50:53,881 - INFO - Step 5780 | mode=video loss=0.169922 diff=0.169922 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0357 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0874 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=4625 i_micro=1155 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.5000 mask_tok=150.00 y_norm=48.0874±0.0064 +2026-05-10 15:51:11,840 - INFO - Step 5790 | mode=video loss=0.104980 diff=0.104980 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0303 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0896 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=4633 i_micro=1157 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3233 mask_tok=97.00 y_norm=48.0896±0.0058 +2026-05-10 15:51:32,583 - INFO - Step 5800 | mode=video loss=0.090820 diff=0.090820 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0635 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0860 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=4641 i_micro=1159 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3467 mask_tok=104.00 y_norm=48.0860±0.0062 cond_uncond_dloss=0.304688 cond_grad=0.000083 cond_pred_l2=37.487122 cond_pred_ratio=0.023436 +2026-05-10 15:51:32,620 - INFO - Step 5800 | probe_semantic mcp_offdiag(mean/min/max)=0.646342/0.533690/0.708790 smol_offdiag=0.660557 prompts=4 mcp_tok=17.50 smol_tok=17.50 +2026-05-10 15:51:39,421 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 15:51:57,375 - INFO - Step 5810 | mode=video loss=0.175781 diff=0.175781 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0347 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0887 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=4649 i_micro=1161 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3133 mask_tok=94.00 y_norm=48.0887±0.0057 +2026-05-10 15:52:16,594 - INFO - Step 5820 | mode=video loss=0.151367 diff=0.151367 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0806 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0874 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=4657 i_micro=1163 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3700 mask_tok=111.00 y_norm=48.0874±0.0055 +2026-05-10 15:52:34,551 - INFO - Step 5830 | mode=video loss=0.082031 diff=0.082031 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0704 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0875 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=4665 i_micro=1165 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3433 mask_tok=103.00 y_norm=48.0875±0.0063 +2026-05-10 15:52:52,514 - INFO - Step 5840 | mode=video loss=0.090820 diff=0.090820 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0383 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0880 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=4673 i_micro=1167 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3600 mask_tok=108.00 y_norm=48.0880±0.0059 +2026-05-10 15:53:11,660 - INFO - Step 5850 | mode=video loss=0.064453 diff=0.064453 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0424 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0874 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=4681 i_micro=1169 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3900 mask_tok=117.00 y_norm=48.0874±0.0057 +2026-05-10 15:53:29,615 - INFO - Step 5860 | mode=video loss=0.091309 diff=0.091309 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0337 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0886 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=4689 i_micro=1171 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3233 mask_tok=97.00 y_norm=48.0886±0.0049 +2026-05-10 15:53:48,003 - INFO - Step 5870 | mode=video loss=0.166016 diff=0.166016 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.4013 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0876 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=4697 i_micro=1173 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3700 mask_tok=111.00 y_norm=48.0876±0.0056 +2026-05-10 15:54:06,696 - INFO - Step 5880 | mode=video loss=0.069824 diff=0.069824 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0701 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0870 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=4705 i_micro=1175 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4733 mask_tok=142.00 y_norm=48.0870±0.0063 +2026-05-10 15:54:24,663 - INFO - Step 5890 | mode=video loss=0.116211 diff=0.116211 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0587 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0868 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=4713 i_micro=1177 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3667 mask_tok=110.00 y_norm=48.0868±0.0055 +2026-05-10 15:54:43,809 - INFO - Step 5900 | mode=video loss=0.098145 diff=0.098145 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0697 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0880 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=4721 i_micro=1179 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3500 mask_tok=105.00 y_norm=48.0880±0.0055 +2026-05-10 15:54:50,500 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 15:55:08,441 - INFO - Step 5910 | mode=video loss=0.079102 diff=0.079102 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0631 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0886 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=4729 i_micro=1181 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4033 mask_tok=121.00 y_norm=48.0886±0.0057 +2026-05-10 15:55:26,387 - INFO - Step 5920 | mode=video loss=0.094238 diff=0.094238 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0490 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0878 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=4737 i_micro=1183 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3567 mask_tok=107.00 y_norm=48.0878±0.0062 +2026-05-10 15:55:45,625 - INFO - Step 5930 | mode=video loss=0.073242 diff=0.073242 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0327 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0886 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=4745 i_micro=1185 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3467 mask_tok=104.00 y_norm=48.0886±0.0056 +2026-05-10 15:56:03,593 - INFO - Step 5940 | mode=video loss=0.068848 diff=0.068848 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0429 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0878 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=4753 i_micro=1187 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4333 mask_tok=130.00 y_norm=48.0878±0.0061 +2026-05-10 15:56:22,763 - INFO - Step 5950 | mode=video loss=0.091797 diff=0.091797 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0393 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0870 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=4761 i_micro=1189 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3833 mask_tok=115.00 y_norm=48.0870±0.0058 +2026-05-10 15:56:40,710 - INFO - Step 5960 | mode=video loss=0.080566 diff=0.080566 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0545 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0898 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=4769 i_micro=1191 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3500 mask_tok=105.00 y_norm=48.0898±0.0056 +2026-05-10 15:56:58,666 - INFO - Step 5970 | mode=video loss=0.100586 diff=0.100586 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0513 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0896 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=4777 i_micro=1193 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3567 mask_tok=107.00 y_norm=48.0896±0.0055 +2026-05-10 15:57:17,726 - INFO - Step 5980 | mode=video loss=0.081543 diff=0.081543 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0823 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0861 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=4785 i_micro=1195 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3633 mask_tok=109.00 y_norm=48.0861±0.0058 +2026-05-10 15:57:35,696 - INFO - Step 5990 | mode=video loss=0.109375 diff=0.109375 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0330 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0886 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=4793 i_micro=1197 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3433 mask_tok=103.00 y_norm=48.0886±0.0051 +2026-05-10 15:57:55,709 - INFO - Step 6000 | mode=video loss=0.087402 diff=0.087402 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0478 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0873 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=4801 i_micro=1199 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3933 mask_tok=118.00 y_norm=48.0873±0.0059 cond_uncond_dloss=0.310547 cond_grad=0.000058 cond_pred_l2=31.373047 cond_pred_ratio=0.018350 +2026-05-10 15:57:55,747 - INFO - Step 6000 | probe_semantic mcp_offdiag(mean/min/max)=0.646224/0.532998/0.708824 smol_offdiag=0.660557 prompts=4 mcp_tok=17.50 smol_tok=17.50 +2026-05-10 15:58:02,653 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 15:58:21,400 - INFO - Step 6010 | mode=video loss=0.099609 diff=0.099609 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0493 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0889 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=4809 i_micro=1201 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3433 mask_tok=103.00 y_norm=48.0889±0.0055 +2026-05-10 15:58:39,367 - INFO - Step 6020 | mode=video loss=0.089844 diff=0.089844 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0450 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0876 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=4817 i_micro=1203 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4833 mask_tok=145.00 y_norm=48.0876±0.0065 +2026-05-10 15:58:58,680 - INFO - Step 6030 | mode=video loss=0.054199 diff=0.054199 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0545 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0859 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=4825 i_micro=1205 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3567 mask_tok=107.00 y_norm=48.0859±0.0063 +2026-05-10 15:59:16,648 - INFO - Step 6040 | mode=video loss=0.137695 diff=0.137695 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0725 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0890 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=4833 i_micro=1207 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3333 mask_tok=100.00 y_norm=48.0890±0.0050 +2026-05-10 15:59:34,601 - INFO - Step 6050 | mode=video loss=0.154297 diff=0.154297 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1124 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0888 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=4841 i_micro=1209 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4167 mask_tok=125.00 y_norm=48.0888±0.0059 +2026-05-10 15:59:53,313 - INFO - Step 6060 | mode=video loss=0.198242 diff=0.198242 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1016 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0889 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=4849 i_micro=1211 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3533 mask_tok=106.00 y_norm=48.0889±0.0060 +2026-05-10 16:00:11,265 - INFO - Step 6070 | mode=video loss=0.069824 diff=0.069824 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0343 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0880 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=4857 i_micro=1213 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3567 mask_tok=107.00 y_norm=48.0880±0.0064 +2026-05-10 16:00:30,349 - INFO - Step 6080 | mode=video loss=0.119141 diff=0.119141 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0712 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0859 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=4865 i_micro=1215 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3933 mask_tok=118.00 y_norm=48.0859±0.0064 +2026-05-10 16:00:48,295 - INFO - Step 6090 | mode=video loss=0.249023 diff=0.249023 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.2188 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0904 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=4873 i_micro=1217 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3400 mask_tok=102.00 y_norm=48.0904±0.0055 +2026-05-10 16:01:06,250 - INFO - Step 6100 | mode=video loss=0.075195 diff=0.075195 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0446 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0890 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=4881 i_micro=1219 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3533 mask_tok=106.00 y_norm=48.0890±0.0057 +2026-05-10 16:01:13,161 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 16:01:32,264 - INFO - Step 6110 | mode=video loss=0.083984 diff=0.083984 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0798 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0889 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=4889 i_micro=1221 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3267 mask_tok=98.00 y_norm=48.0889±0.0050 +2026-05-10 16:01:50,214 - INFO - Step 6120 | mode=video loss=0.134766 diff=0.134766 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0646 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0869 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=4897 i_micro=1223 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4367 mask_tok=131.00 y_norm=48.0869±0.0068 +2026-05-10 16:02:08,723 - INFO - Step 6130 | mode=video loss=0.100586 diff=0.100586 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0362 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0890 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=4905 i_micro=1225 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3633 mask_tok=109.00 y_norm=48.0890±0.0057 +2026-05-10 16:02:27,230 - INFO - Step 6140 | mode=video loss=0.073730 diff=0.073730 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0443 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0870 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=4913 i_micro=1227 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4600 mask_tok=138.00 y_norm=48.0870±0.0059 +2026-05-10 16:02:45,198 - INFO - Step 6150 | mode=video loss=0.081543 diff=0.081543 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0357 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0860 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=4921 i_micro=1229 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4933 mask_tok=148.00 y_norm=48.0860±0.0062 +2026-05-10 16:03:04,450 - INFO - Step 6160 | mode=video loss=0.171875 diff=0.171875 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0347 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0868 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=4929 i_micro=1231 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3800 mask_tok=114.00 y_norm=48.0868±0.0062 +2026-05-10 16:03:22,411 - INFO - Step 6170 | mode=video loss=0.097168 diff=0.097168 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0647 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0892 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=4937 i_micro=1233 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.2900 mask_tok=87.00 y_norm=48.0892±0.0051 +2026-05-10 16:03:40,397 - INFO - Step 6180 | mode=video loss=0.109375 diff=0.109375 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0458 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0887 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=4945 i_micro=1235 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3400 mask_tok=102.00 y_norm=48.0887±0.0050 +2026-05-10 16:03:59,615 - INFO - Step 6190 | mode=video loss=0.113770 diff=0.113770 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0595 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0885 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=4953 i_micro=1237 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3033 mask_tok=91.00 y_norm=48.0885±0.0055 +2026-05-10 16:04:19,192 - INFO - Step 6200 | mode=video loss=0.075195 diff=0.075195 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0638 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0866 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=4961 i_micro=1239 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3367 mask_tok=101.00 y_norm=48.0866±0.0060 cond_uncond_dloss=0.734375 cond_grad=0.000248 cond_pred_l2=85.986115 cond_pred_ratio=0.049941 +2026-05-10 16:04:19,230 - INFO - Step 6200 | probe_semantic mcp_offdiag(mean/min/max)=0.646046/0.532775/0.708304 smol_offdiag=0.660557 prompts=4 mcp_tok=17.50 smol_tok=17.50 +2026-05-10 16:04:26,123 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 16:04:45,349 - INFO - Step 6210 | mode=video loss=0.089844 diff=0.089844 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0880 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0884 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=4969 i_micro=1241 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4067 mask_tok=122.00 y_norm=48.0884±0.0061 +2026-05-10 16:05:03,321 - INFO - Step 6220 | mode=video loss=0.093262 diff=0.093262 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0577 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0868 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=4977 i_micro=1243 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3333 mask_tok=100.00 y_norm=48.0868±0.0061 +2026-05-10 16:05:21,282 - INFO - Step 6230 | mode=video loss=0.056885 diff=0.056885 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0260 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0886 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=4985 i_micro=1245 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3433 mask_tok=103.00 y_norm=48.0886±0.0055 +2026-05-10 16:05:40,434 - INFO - Step 6240 | mode=video loss=0.098145 diff=0.098145 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0266 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0883 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=4993 i_micro=1247 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3367 mask_tok=101.00 y_norm=48.0883±0.0057 +2026-05-10 16:05:58,401 - INFO - Step 6250 | mode=video loss=0.089844 diff=0.089844 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0450 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0875 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=5001 i_micro=1249 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3567 mask_tok=107.00 y_norm=48.0875±0.0058 +2026-05-10 16:06:16,836 - INFO - Step 6260 | mode=video loss=0.100098 diff=0.100098 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0290 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0869 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=5009 i_micro=1251 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3567 mask_tok=107.00 y_norm=48.0869±0.0055 +2026-05-10 16:06:35,529 - INFO - Step 6270 | mode=video loss=0.100586 diff=0.100586 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0397 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0863 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=5017 i_micro=1253 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4433 mask_tok=133.00 y_norm=48.0863±0.0066 +2026-05-10 16:06:53,477 - INFO - Step 6280 | mode=video loss=0.066895 diff=0.066895 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0774 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0848 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=5025 i_micro=1255 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4333 mask_tok=130.00 y_norm=48.0848±0.0066 +2026-05-10 16:07:12,235 - INFO - Step 6290 | mode=video loss=0.107910 diff=0.107910 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0847 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0876 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=5033 i_micro=1257 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3267 mask_tok=98.00 y_norm=48.0876±0.0058 +2026-05-10 16:07:30,182 - INFO - Step 6300 | mode=video loss=0.097168 diff=0.097168 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0426 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0879 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=5041 i_micro=1259 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3500 mask_tok=105.00 y_norm=48.0879±0.0051 +2026-05-10 16:07:36,992 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 16:07:54,949 - INFO - Step 6310 | mode=video loss=0.103516 diff=0.103516 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0958 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0872 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=5049 i_micro=1261 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3367 mask_tok=101.00 y_norm=48.0872±0.0054 +2026-05-10 16:08:14,129 - INFO - Step 6320 | mode=video loss=0.090820 diff=0.090820 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0457 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0899 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=5057 i_micro=1263 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4267 mask_tok=128.00 y_norm=48.0899±0.0059 +2026-05-10 16:08:32,101 - INFO - Step 6330 | mode=video loss=0.067871 diff=0.067871 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0469 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0887 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=5065 i_micro=1265 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3567 mask_tok=107.00 y_norm=48.0887±0.0049 +2026-05-10 16:08:51,277 - INFO - Step 6340 | mode=video loss=0.084473 diff=0.084473 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0520 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0861 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=5073 i_micro=1267 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.5200 mask_tok=156.00 y_norm=48.0861±0.0066 +2026-05-10 16:09:09,231 - INFO - Step 6350 | mode=video loss=0.102051 diff=0.102051 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0408 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0876 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=5081 i_micro=1269 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3567 mask_tok=107.00 y_norm=48.0876±0.0067 +2026-05-10 16:09:27,190 - INFO - Step 6360 | mode=video loss=0.052734 diff=0.052734 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0607 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0868 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=5089 i_micro=1271 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4400 mask_tok=132.00 y_norm=48.0868±0.0070 +2026-05-10 16:09:46,406 - INFO - Step 6370 | mode=video loss=0.066406 diff=0.066406 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0618 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0859 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=5097 i_micro=1273 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4200 mask_tok=126.00 y_norm=48.0859±0.0061 +2026-05-10 16:10:04,376 - INFO - Step 6380 | mode=video loss=0.064453 diff=0.064453 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1835 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0882 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=5105 i_micro=1275 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3167 mask_tok=95.00 y_norm=48.0882±0.0048 +2026-05-10 16:10:22,355 - INFO - Step 6390 | mode=video loss=0.088379 diff=0.088379 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0420 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0877 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=5113 i_micro=1277 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3633 mask_tok=109.00 y_norm=48.0877±0.0051 +2026-05-10 16:10:42,739 - INFO - Step 6400 | mode=video loss=0.115234 diff=0.115234 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0760 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0866 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=5121 i_micro=1279 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4067 mask_tok=122.00 y_norm=48.0866±0.0057 cond_uncond_dloss=0.392578 cond_grad=0.000299 cond_pred_l2=39.461185 cond_pred_ratio=0.025576 +2026-05-10 16:10:42,777 - INFO - Step 6400 | probe_semantic mcp_offdiag(mean/min/max)=0.646057/0.533095/0.707832 smol_offdiag=0.660557 prompts=4 mcp_tok=17.50 smol_tok=17.50 +2026-05-10 16:10:49,643 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 16:11:07,579 - INFO - Step 6410 | mode=video loss=0.105957 diff=0.105957 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0720 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0866 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=5129 i_micro=1281 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4667 mask_tok=140.00 y_norm=48.0866±0.0066 +2026-05-10 16:11:26,818 - INFO - Step 6420 | mode=video loss=0.101562 diff=0.101562 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0422 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0876 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=5137 i_micro=1283 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3800 mask_tok=114.00 y_norm=48.0876±0.0065 +2026-05-10 16:11:44,782 - INFO - Step 6430 | mode=video loss=0.045166 diff=0.045166 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0536 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0855 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=5145 i_micro=1285 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3400 mask_tok=102.00 y_norm=48.0855±0.0054 +2026-05-10 16:12:03,178 - INFO - Step 6440 | mode=video loss=0.098145 diff=0.098145 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0650 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0864 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=5153 i_micro=1287 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3800 mask_tok=114.00 y_norm=48.0864±0.0060 +2026-05-10 16:12:21,879 - INFO - Step 6450 | mode=video loss=0.051758 diff=0.051758 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1062 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0860 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=5161 i_micro=1289 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3467 mask_tok=104.00 y_norm=48.0860±0.0058 +2026-05-10 16:12:39,832 - INFO - Step 6460 | mode=video loss=0.080566 diff=0.080566 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0312 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0873 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=5169 i_micro=1291 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3633 mask_tok=109.00 y_norm=48.0873±0.0062 +2026-05-10 16:12:59,058 - INFO - Step 6470 | mode=video loss=0.107910 diff=0.107910 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0733 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0882 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=5177 i_micro=1293 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3133 mask_tok=94.00 y_norm=48.0882±0.0055 +2026-05-10 16:13:17,008 - INFO - Step 6480 | mode=video loss=0.146484 diff=0.146484 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.2104 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0863 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=5185 i_micro=1295 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.5133 mask_tok=154.00 y_norm=48.0863±0.0065 +2026-05-10 16:13:34,971 - INFO - Step 6490 | mode=video loss=0.091797 diff=0.091797 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0316 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0872 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=5193 i_micro=1297 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3400 mask_tok=102.00 y_norm=48.0872±0.0059 +2026-05-10 16:13:54,121 - INFO - Step 6500 | mode=video loss=0.087402 diff=0.087402 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0323 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0884 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=5201 i_micro=1299 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3433 mask_tok=103.00 y_norm=48.0884±0.0059 +2026-05-10 16:14:00,912 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 16:14:18,870 - INFO - Step 6510 | mode=video loss=0.058838 diff=0.058838 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0514 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0856 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=5209 i_micro=1301 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3900 mask_tok=117.00 y_norm=48.0856±0.0057 +2026-05-10 16:14:38,061 - INFO - Step 6520 | mode=video loss=0.135742 diff=0.135742 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0854 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0885 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=5217 i_micro=1303 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3367 mask_tok=101.00 y_norm=48.0885±0.0055 +2026-05-10 16:14:56,021 - INFO - Step 6530 | mode=video loss=0.159180 diff=0.159180 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.2152 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0874 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=5225 i_micro=1305 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3667 mask_tok=110.00 y_norm=48.0874±0.0053 +2026-05-10 16:15:13,987 - INFO - Step 6540 | mode=video loss=0.135742 diff=0.135742 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0652 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0886 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=5233 i_micro=1307 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3700 mask_tok=111.00 y_norm=48.0886±0.0061 +2026-05-10 16:15:33,086 - INFO - Step 6550 | mode=video loss=0.124023 diff=0.124023 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0335 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0885 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=5241 i_micro=1309 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3600 mask_tok=108.00 y_norm=48.0885±0.0064 +2026-05-10 16:15:51,063 - INFO - Step 6560 | mode=video loss=0.151367 diff=0.151367 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0700 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0857 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=5249 i_micro=1311 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4333 mask_tok=130.00 y_norm=48.0857±0.0062 +2026-05-10 16:16:09,515 - INFO - Step 6570 | mode=video loss=0.123535 diff=0.123535 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0466 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0863 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=5257 i_micro=1313 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4267 mask_tok=128.00 y_norm=48.0863±0.0060 +2026-05-10 16:16:28,190 - INFO - Step 6580 | mode=video loss=0.080566 diff=0.080566 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1125 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0871 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=5265 i_micro=1315 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4100 mask_tok=123.00 y_norm=48.0871±0.0061 +2026-05-10 16:16:46,159 - INFO - Step 6590 | mode=video loss=0.069824 diff=0.069824 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0834 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0871 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=5273 i_micro=1317 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4800 mask_tok=144.00 y_norm=48.0871±0.0059 +2026-05-10 16:17:06,975 - INFO - Step 6600 | mode=video loss=0.061035 diff=0.061035 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0599 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0865 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=5281 i_micro=1319 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3633 mask_tok=109.00 y_norm=48.0865±0.0060 cond_uncond_dloss=0.406250 cond_grad=0.000087 cond_pred_l2=40.406322 cond_pred_ratio=0.023008 +2026-05-10 16:17:07,012 - INFO - Step 6600 | probe_semantic mcp_offdiag(mean/min/max)=0.647483/0.535142/0.708975 smol_offdiag=0.660557 prompts=4 mcp_tok=17.50 smol_tok=17.50 +2026-05-10 16:17:14,122 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 16:17:32,064 - INFO - Step 6610 | mode=video loss=0.157227 diff=0.157227 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1105 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0884 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=5289 i_micro=1321 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3633 mask_tok=109.00 y_norm=48.0884±0.0053 +2026-05-10 16:17:50,034 - INFO - Step 6620 | mode=video loss=0.091797 diff=0.091797 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0287 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0875 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=5297 i_micro=1323 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4233 mask_tok=127.00 y_norm=48.0875±0.0069 +2026-05-10 16:18:09,227 - INFO - Step 6630 | mode=video loss=0.186523 diff=0.186523 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1031 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0885 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=5305 i_micro=1325 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3467 mask_tok=104.00 y_norm=48.0885±0.0058 +2026-05-10 16:18:27,202 - INFO - Step 6640 | mode=video loss=0.122559 diff=0.122559 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0369 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0877 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=5313 i_micro=1327 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3967 mask_tok=119.00 y_norm=48.0877±0.0060 +2026-05-10 16:18:46,417 - INFO - Step 6650 | mode=video loss=0.080566 diff=0.080566 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0506 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0890 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=5321 i_micro=1329 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3400 mask_tok=102.00 y_norm=48.0890±0.0052 +2026-05-10 16:19:04,378 - INFO - Step 6660 | mode=video loss=0.086914 diff=0.086914 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0339 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0888 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=5329 i_micro=1331 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3333 mask_tok=100.00 y_norm=48.0888±0.0048 +2026-05-10 16:19:22,348 - INFO - Step 6670 | mode=video loss=0.087402 diff=0.087402 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0381 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0883 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=5337 i_micro=1333 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4033 mask_tok=121.00 y_norm=48.0883±0.0074 +2026-05-10 16:19:41,559 - INFO - Step 6680 | mode=video loss=0.084961 diff=0.084961 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0760 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0889 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=5345 i_micro=1335 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3467 mask_tok=104.00 y_norm=48.0889±0.0055 +2026-05-10 16:19:59,510 - INFO - Step 6690 | mode=video loss=0.094727 diff=0.094727 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0335 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0874 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=5353 i_micro=1337 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3967 mask_tok=119.00 y_norm=48.0874±0.0057 +2026-05-10 16:20:17,913 - INFO - Step 6700 | mode=video loss=0.066406 diff=0.066406 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0305 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0873 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=5361 i_micro=1339 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3333 mask_tok=100.00 y_norm=48.0873±0.0063 +2026-05-10 16:20:25,306 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 16:20:43,253 - INFO - Step 6710 | mode=video loss=0.111328 diff=0.111328 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0819 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0856 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=5369 i_micro=1341 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3733 mask_tok=112.00 y_norm=48.0856±0.0058 +2026-05-10 16:21:01,209 - INFO - Step 6720 | mode=video loss=0.073730 diff=0.073730 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0456 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0874 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=5377 i_micro=1343 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4100 mask_tok=123.00 y_norm=48.0874±0.0064 +2026-05-10 16:21:20,456 - INFO - Step 6730 | mode=video loss=0.100586 diff=0.100586 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0421 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0894 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=5385 i_micro=1345 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3333 mask_tok=100.00 y_norm=48.0894±0.0056 +2026-05-10 16:21:38,426 - INFO - Step 6740 | mode=video loss=0.077148 diff=0.077148 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0391 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0865 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=5393 i_micro=1347 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3800 mask_tok=114.00 y_norm=48.0865±0.0062 +2026-05-10 16:21:56,401 - INFO - Step 6750 | mode=video loss=0.111816 diff=0.111816 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0623 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0889 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=5401 i_micro=1349 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3133 mask_tok=94.00 y_norm=48.0889±0.0062 +2026-05-10 16:22:15,599 - INFO - Step 6760 | mode=video loss=0.093262 diff=0.093262 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0284 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0899 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=5409 i_micro=1351 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.8400 mask_tok=252.00 y_norm=48.0899±0.0075 +2026-05-10 16:22:33,552 - INFO - Step 6770 | mode=video loss=0.077637 diff=0.077637 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0580 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0883 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=5417 i_micro=1353 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3733 mask_tok=112.00 y_norm=48.0883±0.0048 +2026-05-10 16:22:52,685 - INFO - Step 6780 | mode=video loss=0.119141 diff=0.119141 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0599 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0857 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=5425 i_micro=1355 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3633 mask_tok=109.00 y_norm=48.0857±0.0055 +2026-05-10 16:23:10,643 - INFO - Step 6790 | mode=video loss=0.028198 diff=0.028198 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1046 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0862 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=5433 i_micro=1357 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.5533 mask_tok=166.00 y_norm=48.0862±0.0069 +2026-05-10 16:23:30,232 - INFO - Step 6800 | mode=video loss=0.085938 diff=0.085938 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0291 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0894 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=5441 i_micro=1359 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4033 mask_tok=121.00 y_norm=48.0894±0.0060 cond_uncond_dloss=0.492188 cond_grad=0.000138 cond_pred_l2=41.306232 cond_pred_ratio=0.025035 +2026-05-10 16:23:30,270 - INFO - Step 6800 | probe_semantic mcp_offdiag(mean/min/max)=0.647621/0.535122/0.709303 smol_offdiag=0.660557 prompts=4 mcp_tok=17.50 smol_tok=17.50 +2026-05-10 16:23:37,146 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 16:23:56,377 - INFO - Step 6810 | mode=video loss=0.044434 diff=0.044434 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0235 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0866 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=5449 i_micro=1361 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4033 mask_tok=121.00 y_norm=48.0866±0.0056 +2026-05-10 16:24:14,329 - INFO - Step 6820 | mode=video loss=0.108398 diff=0.108398 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0508 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0887 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=5457 i_micro=1363 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3767 mask_tok=113.00 y_norm=48.0887±0.0064 +2026-05-10 16:24:33,052 - INFO - Step 6830 | mode=video loss=0.062012 diff=0.062012 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0401 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0885 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=5465 i_micro=1365 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3367 mask_tok=101.00 y_norm=48.0885±0.0057 +2026-05-10 16:24:51,018 - INFO - Step 6840 | mode=video loss=0.078613 diff=0.078613 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0751 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0891 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=5473 i_micro=1367 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3200 mask_tok=96.00 y_norm=48.0891±0.0057 +2026-05-10 16:25:08,990 - INFO - Step 6850 | mode=video loss=0.125000 diff=0.125000 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0906 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0871 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=5481 i_micro=1369 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.6767 mask_tok=203.00 y_norm=48.0871±0.0060 +2026-05-10 16:25:27,633 - INFO - Step 6860 | mode=video loss=0.104004 diff=0.104004 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1066 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0894 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=5489 i_micro=1371 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3833 mask_tok=115.00 y_norm=48.0894±0.0058 +2026-05-10 16:25:45,590 - INFO - Step 6870 | mode=video loss=0.063965 diff=0.063965 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0498 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0887 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=5497 i_micro=1373 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3567 mask_tok=107.00 y_norm=48.0887±0.0055 +2026-05-10 16:26:03,965 - INFO - Step 6880 | mode=video loss=0.081055 diff=0.081055 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0346 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0896 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=5505 i_micro=1375 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3567 mask_tok=107.00 y_norm=48.0896±0.0057 +2026-05-10 16:26:22,610 - INFO - Step 6890 | mode=video loss=0.100586 diff=0.100586 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0257 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0886 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=5513 i_micro=1377 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.2900 mask_tok=87.00 y_norm=48.0886±0.0049 +2026-05-10 16:26:40,563 - INFO - Step 6900 | mode=video loss=0.112793 diff=0.112793 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0715 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0850 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=5521 i_micro=1379 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4267 mask_tok=128.00 y_norm=48.0850±0.0055 +2026-05-10 16:26:47,542 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 16:27:06,309 - INFO - Step 6910 | mode=video loss=0.083008 diff=0.083008 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0700 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0883 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=5529 i_micro=1381 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3133 mask_tok=94.00 y_norm=48.0883±0.0052 +2026-05-10 16:27:24,275 - INFO - Step 6920 | mode=video loss=0.124023 diff=0.124023 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0945 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0880 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=5537 i_micro=1383 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3233 mask_tok=97.00 y_norm=48.0880±0.0052 +2026-05-10 16:27:42,249 - INFO - Step 6930 | mode=video loss=0.091797 diff=0.091797 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0679 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0874 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=5545 i_micro=1385 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4267 mask_tok=128.00 y_norm=48.0874±0.0061 +2026-05-10 16:28:00,989 - INFO - Step 6940 | mode=video loss=0.132812 diff=0.132812 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0288 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0868 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=5553 i_micro=1387 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4433 mask_tok=133.00 y_norm=48.0868±0.0065 +2026-05-10 16:28:18,960 - INFO - Step 6950 | mode=video loss=0.057373 diff=0.057373 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0587 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0858 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=5561 i_micro=1389 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3600 mask_tok=108.00 y_norm=48.0858±0.0060 +2026-05-10 16:28:37,614 - INFO - Step 6960 | mode=video loss=0.065918 diff=0.065918 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0667 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0862 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=5569 i_micro=1391 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3600 mask_tok=108.00 y_norm=48.0862±0.0067 +2026-05-10 16:28:55,569 - INFO - Step 6970 | mode=video loss=0.085449 diff=0.085449 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0744 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0881 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=5577 i_micro=1393 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3767 mask_tok=113.00 y_norm=48.0881±0.0053 +2026-05-10 16:29:13,531 - INFO - Step 6980 | mode=video loss=0.081543 diff=0.081543 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0624 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0871 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=5585 i_micro=1395 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3067 mask_tok=92.00 y_norm=48.0871±0.0061 +2026-05-10 16:29:32,712 - INFO - Step 6990 | mode=video loss=0.053223 diff=0.053223 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0315 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0865 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=5593 i_micro=1397 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4367 mask_tok=131.00 y_norm=48.0865±0.0059 +2026-05-10 16:29:52,284 - INFO - Step 7000 | mode=video loss=0.088379 diff=0.088379 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0522 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0892 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=5601 i_micro=1399 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3400 mask_tok=102.00 y_norm=48.0892±0.0057 cond_uncond_dloss=0.443359 cond_grad=0.000292 cond_pred_l2=53.700008 cond_pred_ratio=0.032638 +2026-05-10 16:29:52,322 - INFO - Step 7000 | probe_semantic mcp_offdiag(mean/min/max)=0.647277/0.534697/0.709109 smol_offdiag=0.660557 prompts=4 mcp_tok=17.50 smol_tok=17.50 +2026-05-10 16:29:59,414 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 16:30:18,668 - INFO - Step 7010 | mode=video loss=0.130859 diff=0.130859 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0426 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0871 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=5609 i_micro=1401 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3367 mask_tok=101.00 y_norm=48.0871±0.0056 +2026-05-10 16:30:36,624 - INFO - Step 7020 | mode=video loss=0.073242 diff=0.073242 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0911 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0900 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=5617 i_micro=1403 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3633 mask_tok=109.00 y_norm=48.0900±0.0054 +2026-05-10 16:30:54,591 - INFO - Step 7030 | mode=video loss=0.104492 diff=0.104492 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0325 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0887 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=5625 i_micro=1405 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.2400 mask_tok=72.00 y_norm=48.0887±0.0053 +2026-05-10 16:31:13,839 - INFO - Step 7040 | mode=video loss=0.128906 diff=0.128906 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0319 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0865 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=5633 i_micro=1407 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4667 mask_tok=140.00 y_norm=48.0865±0.0064 +2026-05-10 16:31:31,814 - INFO - Step 7050 | mode=video loss=0.081055 diff=0.081055 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1474 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0887 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=5641 i_micro=1409 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3333 mask_tok=100.00 y_norm=48.0887±0.0057 +2026-05-10 16:31:49,786 - INFO - Step 7060 | mode=video loss=0.064941 diff=0.064941 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0345 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0891 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=5649 i_micro=1411 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3367 mask_tok=101.00 y_norm=48.0891±0.0059 +2026-05-10 16:32:08,879 - INFO - Step 7070 | mode=video loss=0.076172 diff=0.076172 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0675 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0882 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=5657 i_micro=1413 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3800 mask_tok=114.00 y_norm=48.0882±0.0065 +2026-05-10 16:32:26,840 - INFO - Step 7080 | mode=video loss=0.119629 diff=0.119629 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0531 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0891 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=5665 i_micro=1415 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3667 mask_tok=110.00 y_norm=48.0891±0.0051 +2026-05-10 16:32:45,961 - INFO - Step 7090 | mode=video loss=0.081055 diff=0.081055 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0714 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0881 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=5673 i_micro=1417 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3433 mask_tok=103.00 y_norm=48.0881±0.0060 +2026-05-10 16:33:03,941 - INFO - Step 7100 | mode=video loss=0.087402 diff=0.087402 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0575 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0862 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=5681 i_micro=1419 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3433 mask_tok=103.00 y_norm=48.0862±0.0062 +2026-05-10 16:33:10,767 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 16:33:28,739 - INFO - Step 7110 | mode=video loss=0.104980 diff=0.104980 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0316 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0867 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=5689 i_micro=1421 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3700 mask_tok=111.00 y_norm=48.0867±0.0057 +2026-05-10 16:33:47,714 - INFO - Step 7120 | mode=video loss=0.134766 diff=0.134766 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0422 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0862 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=5697 i_micro=1423 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4967 mask_tok=149.00 y_norm=48.0862±0.0070 +2026-05-10 16:34:05,701 - INFO - Step 7130 | mode=video loss=0.085449 diff=0.085449 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0392 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0856 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=5705 i_micro=1425 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.5133 mask_tok=154.00 y_norm=48.0856±0.0065 +2026-05-10 16:34:24,338 - INFO - Step 7140 | mode=video loss=0.140625 diff=0.140625 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0385 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0873 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=5713 i_micro=1427 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3600 mask_tok=108.00 y_norm=48.0873±0.0069 +2026-05-10 16:34:42,809 - INFO - Step 7150 | mode=video loss=0.150391 diff=0.150391 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0263 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0896 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=5721 i_micro=1429 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3400 mask_tok=102.00 y_norm=48.0896±0.0055 +2026-05-10 16:35:00,772 - INFO - Step 7160 | mode=video loss=0.104492 diff=0.104492 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0764 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0886 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=5729 i_micro=1431 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3567 mask_tok=107.00 y_norm=48.0886±0.0056 +2026-05-10 16:35:19,891 - INFO - Step 7170 | mode=video loss=0.220703 diff=0.220703 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0259 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0891 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=5737 i_micro=1433 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.2933 mask_tok=88.00 y_norm=48.0891±0.0044 +2026-05-10 16:35:37,862 - INFO - Step 7180 | mode=video loss=0.097168 diff=0.097168 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0395 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0881 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=5745 i_micro=1435 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.2667 mask_tok=80.00 y_norm=48.0881±0.0052 +2026-05-10 16:35:56,560 - INFO - Step 7190 | mode=video loss=0.071289 diff=0.071289 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0676 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0872 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=5753 i_micro=1437 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4233 mask_tok=127.00 y_norm=48.0872±0.0059 +2026-05-10 16:36:16,653 - INFO - Step 7200 | mode=video loss=0.071777 diff=0.071777 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0269 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0894 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=5761 i_micro=1439 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3833 mask_tok=115.00 y_norm=48.0894±0.0059 cond_uncond_dloss=0.335938 cond_grad=0.000047 cond_pred_l2=36.409218 cond_pred_ratio=0.020707 +2026-05-10 16:36:16,690 - INFO - Step 7200 | probe_semantic mcp_offdiag(mean/min/max)=0.648291/0.535719/0.709983 smol_offdiag=0.660557 prompts=4 mcp_tok=17.50 smol_tok=17.50 +2026-05-10 16:36:23,413 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 16:36:41,374 - INFO - Step 7210 | mode=video loss=0.108887 diff=0.108887 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0820 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0888 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=5769 i_micro=1441 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3533 mask_tok=106.00 y_norm=48.0888±0.0055 +2026-05-10 16:37:00,642 - INFO - Step 7220 | mode=video loss=0.078125 diff=0.078125 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0302 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0872 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=5777 i_micro=1443 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3433 mask_tok=103.00 y_norm=48.0872±0.0055 +2026-05-10 16:37:18,601 - INFO - Step 7230 | mode=video loss=0.086426 diff=0.086426 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0440 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0896 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=5785 i_micro=1445 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.2867 mask_tok=86.00 y_norm=48.0896±0.0056 +2026-05-10 16:37:36,576 - INFO - Step 7240 | mode=video loss=0.091309 diff=0.091309 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0397 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0886 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=5793 i_micro=1447 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.1867 mask_tok=56.00 y_norm=48.0886±0.0050 +2026-05-10 16:37:55,661 - INFO - Step 7250 | mode=video loss=0.093750 diff=0.093750 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0250 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0898 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=5801 i_micro=1449 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3967 mask_tok=119.00 y_norm=48.0898±0.0063 +2026-05-10 16:38:13,615 - INFO - Step 7260 | mode=video loss=0.080566 diff=0.080566 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0288 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0890 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=5809 i_micro=1451 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3567 mask_tok=107.00 y_norm=48.0890±0.0053 +2026-05-10 16:38:32,319 - INFO - Step 7270 | mode=video loss=0.123047 diff=0.123047 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0698 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0873 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=5817 i_micro=1453 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.5467 mask_tok=164.00 y_norm=48.0873±0.0066 +2026-05-10 16:38:50,755 - INFO - Step 7280 | mode=video loss=0.081055 diff=0.081055 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0298 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0877 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=5825 i_micro=1455 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4167 mask_tok=125.00 y_norm=48.0877±0.0065 +2026-05-10 16:39:08,701 - INFO - Step 7290 | mode=video loss=0.102051 diff=0.102051 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0593 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0877 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=5833 i_micro=1457 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3700 mask_tok=111.00 y_norm=48.0877±0.0062 +2026-05-10 16:39:27,838 - INFO - Step 7300 | mode=video loss=0.084961 diff=0.084961 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0803 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0877 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=5841 i_micro=1459 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3500 mask_tok=105.00 y_norm=48.0877±0.0058 +2026-05-10 16:39:34,734 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 16:39:52,683 - INFO - Step 7310 | mode=video loss=0.127930 diff=0.127930 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0802 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0878 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=5849 i_micro=1461 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3400 mask_tok=102.00 y_norm=48.0878±0.0063 +2026-05-10 16:40:11,450 - INFO - Step 7320 | mode=video loss=0.109375 diff=0.109375 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0504 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0894 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=5857 i_micro=1463 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3700 mask_tok=111.00 y_norm=48.0894±0.0069 +2026-05-10 16:40:29,829 - INFO - Step 7330 | mode=video loss=0.076660 diff=0.076660 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1930 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0889 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=5865 i_micro=1465 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4133 mask_tok=124.00 y_norm=48.0889±0.0073 +2026-05-10 16:40:47,792 - INFO - Step 7340 | mode=video loss=0.083008 diff=0.083008 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0440 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0886 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=5873 i_micro=1467 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3533 mask_tok=106.00 y_norm=48.0886±0.0060 +2026-05-10 16:41:06,876 - INFO - Step 7350 | mode=video loss=0.159180 diff=0.159180 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0581 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0853 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=5881 i_micro=1469 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3933 mask_tok=118.00 y_norm=48.0853±0.0056 +2026-05-10 16:41:24,828 - INFO - Step 7360 | mode=video loss=0.117676 diff=0.117676 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0361 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0892 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=5889 i_micro=1471 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3200 mask_tok=96.00 y_norm=48.0892±0.0051 +2026-05-10 16:41:42,784 - INFO - Step 7370 | mode=video loss=0.064453 diff=0.064453 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0279 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0875 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=5897 i_micro=1473 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4933 mask_tok=148.00 y_norm=48.0875±0.0073 +2026-05-10 16:42:01,910 - INFO - Step 7380 | mode=video loss=0.081055 diff=0.081055 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0435 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0869 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=5905 i_micro=1475 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3967 mask_tok=119.00 y_norm=48.0869±0.0066 +2026-05-10 16:42:19,873 - INFO - Step 7390 | mode=video loss=0.071777 diff=0.071777 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0394 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0864 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=5913 i_micro=1477 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4000 mask_tok=120.00 y_norm=48.0864±0.0064 +2026-05-10 16:42:39,964 - INFO - Step 7400 | mode=video loss=0.070801 diff=0.070801 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0305 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0887 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=5921 i_micro=1479 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3100 mask_tok=93.00 y_norm=48.0887±0.0057 cond_uncond_dloss=0.312500 cond_grad=0.000072 cond_pred_l2=43.283890 cond_pred_ratio=0.025091 +2026-05-10 16:42:40,002 - INFO - Step 7400 | probe_semantic mcp_offdiag(mean/min/max)=0.648663/0.536437/0.710122 smol_offdiag=0.660557 prompts=4 mcp_tok=17.50 smol_tok=17.50 +2026-05-10 16:42:47,351 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 16:43:05,796 - INFO - Step 7410 | mode=video loss=0.068848 diff=0.068848 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0293 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0883 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=5929 i_micro=1481 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3633 mask_tok=109.00 y_norm=48.0883±0.0057 +2026-05-10 16:43:23,776 - INFO - Step 7420 | mode=video loss=0.107910 diff=0.107910 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0630 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0898 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=5937 i_micro=1483 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3500 mask_tok=105.00 y_norm=48.0898±0.0057 +2026-05-10 16:43:42,986 - INFO - Step 7430 | mode=video loss=0.094238 diff=0.094238 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0392 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0888 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=5945 i_micro=1485 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3833 mask_tok=115.00 y_norm=48.0888±0.0061 +2026-05-10 16:44:00,948 - INFO - Step 7440 | mode=video loss=0.098145 diff=0.098145 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0683 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0876 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=5953 i_micro=1487 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3367 mask_tok=101.00 y_norm=48.0876±0.0058 +2026-05-10 16:44:19,701 - INFO - Step 7450 | mode=video loss=0.096680 diff=0.096680 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0590 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0873 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=5961 i_micro=1489 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.5267 mask_tok=158.00 y_norm=48.0873±0.0064 +2026-05-10 16:44:38,099 - INFO - Step 7460 | mode=video loss=0.079102 diff=0.079102 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0348 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0884 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=5969 i_micro=1491 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3500 mask_tok=105.00 y_norm=48.0884±0.0052 +2026-05-10 16:44:56,049 - INFO - Step 7470 | mode=video loss=0.121582 diff=0.121582 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0888 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0871 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=5977 i_micro=1493 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3900 mask_tok=117.00 y_norm=48.0871±0.0065 +2026-05-10 16:45:15,173 - INFO - Step 7480 | mode=video loss=0.140625 diff=0.140625 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0398 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0855 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=5985 i_micro=1495 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4100 mask_tok=123.00 y_norm=48.0855±0.0058 +2026-05-10 16:45:33,143 - INFO - Step 7490 | mode=video loss=0.094727 diff=0.094727 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0701 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0887 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=5993 i_micro=1497 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4533 mask_tok=136.00 y_norm=48.0887±0.0059 +2026-05-10 16:45:51,108 - INFO - Step 7500 | mode=video loss=0.142578 diff=0.142578 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0275 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0889 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=6001 i_micro=1499 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3833 mask_tok=115.00 y_norm=48.0889±0.0057 +2026-05-10 16:45:57,935 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 16:46:16,962 - INFO - Step 7510 | mode=video loss=0.102539 diff=0.102539 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0869 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0859 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=6009 i_micro=1501 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3467 mask_tok=104.00 y_norm=48.0859±0.0061 +2026-05-10 16:46:34,925 - INFO - Step 7520 | mode=video loss=0.120605 diff=0.120605 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0311 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0885 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=6017 i_micro=1503 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3533 mask_tok=106.00 y_norm=48.0885±0.0056 +2026-05-10 16:46:53,623 - INFO - Step 7530 | mode=video loss=0.104004 diff=0.104004 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0606 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0888 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=6025 i_micro=1505 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3633 mask_tok=109.00 y_norm=48.0888±0.0059 +2026-05-10 16:47:12,086 - INFO - Step 7540 | mode=video loss=0.085449 diff=0.085449 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0630 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0880 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=6033 i_micro=1507 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4900 mask_tok=147.00 y_norm=48.0880±0.0073 +2026-05-10 16:47:30,046 - INFO - Step 7550 | mode=video loss=0.087402 diff=0.087402 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0533 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0868 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=6041 i_micro=1509 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3300 mask_tok=99.00 y_norm=48.0868±0.0052 +2026-05-10 16:47:48,986 - INFO - Step 7560 | mode=video loss=0.076172 diff=0.076172 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0415 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0882 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=6049 i_micro=1511 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3333 mask_tok=100.00 y_norm=48.0882±0.0055 +2026-05-10 16:48:06,942 - INFO - Step 7570 | mode=video loss=0.086426 diff=0.086426 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0367 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0887 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=6057 i_micro=1513 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3467 mask_tok=104.00 y_norm=48.0887±0.0052 +2026-05-10 16:48:25,395 - INFO - Step 7580 | mode=video loss=0.083496 diff=0.083496 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0683 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0875 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=6065 i_micro=1515 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3600 mask_tok=108.00 y_norm=48.0875±0.0058 +2026-05-10 16:48:43,765 - INFO - Step 7590 | mode=video loss=0.208984 diff=0.208984 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.2478 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0878 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=6073 i_micro=1517 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3333 mask_tok=100.00 y_norm=48.0878±0.0059 +2026-05-10 16:49:03,336 - INFO - Step 7600 | mode=video loss=0.092285 diff=0.092285 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0604 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0857 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=6081 i_micro=1519 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4033 mask_tok=121.00 y_norm=48.0857±0.0077 cond_uncond_dloss=0.296875 cond_grad=0.000190 cond_pred_l2=40.190876 cond_pred_ratio=0.024693 +2026-05-10 16:49:03,374 - INFO - Step 7600 | probe_semantic mcp_offdiag(mean/min/max)=0.648382/0.535900/0.709875 smol_offdiag=0.660557 prompts=4 mcp_tok=17.50 smol_tok=17.50 +2026-05-10 16:49:10,219 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 16:49:29,420 - INFO - Step 7610 | mode=video loss=0.114258 diff=0.114258 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0653 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0869 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=6089 i_micro=1521 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3467 mask_tok=104.00 y_norm=48.0869±0.0058 +2026-05-10 16:49:47,397 - INFO - Step 7620 | mode=video loss=0.144531 diff=0.144531 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0578 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0881 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=6097 i_micro=1523 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3467 mask_tok=104.00 y_norm=48.0881±0.0057 +2026-05-10 16:50:06,113 - INFO - Step 7630 | mode=video loss=0.053711 diff=0.053711 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0204 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0888 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=6105 i_micro=1525 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3667 mask_tok=110.00 y_norm=48.0888±0.0056 +2026-05-10 16:50:24,500 - INFO - Step 7640 | mode=video loss=0.076660 diff=0.076660 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0405 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0890 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=6113 i_micro=1527 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3533 mask_tok=106.00 y_norm=48.0890±0.0059 +2026-05-10 16:50:42,472 - INFO - Step 7650 | mode=video loss=0.158203 diff=0.158203 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1036 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0870 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=6121 i_micro=1529 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4367 mask_tok=131.00 y_norm=48.0870±0.0072 +2026-05-10 16:51:01,159 - INFO - Step 7660 | mode=video loss=0.148438 diff=0.148438 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0973 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0878 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=6129 i_micro=1531 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3100 mask_tok=93.00 y_norm=48.0878±0.0057 +2026-05-10 16:51:19,560 - INFO - Step 7670 | mode=video loss=0.098633 diff=0.098633 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0524 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0876 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=6137 i_micro=1533 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3667 mask_tok=110.00 y_norm=48.0876±0.0058 +2026-05-10 16:51:37,510 - INFO - Step 7680 | mode=video loss=0.101074 diff=0.101074 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0531 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0868 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=6145 i_micro=1535 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3600 mask_tok=108.00 y_norm=48.0868±0.0060 +2026-05-10 16:51:56,556 - INFO - Step 7690 | mode=video loss=0.096680 diff=0.096680 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0416 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0869 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=6153 i_micro=1537 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3300 mask_tok=99.00 y_norm=48.0869±0.0055 +2026-05-10 16:52:14,523 - INFO - Step 7700 | mode=video loss=0.171875 diff=0.171875 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0417 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0857 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=6161 i_micro=1539 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4133 mask_tok=124.00 y_norm=48.0857±0.0058 +2026-05-10 16:52:21,747 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 16:52:40,447 - INFO - Step 7710 | mode=video loss=0.126953 diff=0.126953 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1143 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0868 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=6169 i_micro=1541 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3600 mask_tok=108.00 y_norm=48.0868±0.0055 +2026-05-10 16:52:58,839 - INFO - Step 7720 | mode=video loss=0.090820 diff=0.090820 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0355 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0867 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=6177 i_micro=1543 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4133 mask_tok=124.00 y_norm=48.0867±0.0064 +2026-05-10 16:53:16,814 - INFO - Step 7730 | mode=video loss=0.072754 diff=0.072754 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0509 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0870 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=6185 i_micro=1545 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3433 mask_tok=103.00 y_norm=48.0870±0.0058 +2026-05-10 16:53:35,942 - INFO - Step 7740 | mode=video loss=0.069824 diff=0.069824 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1065 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0882 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=6193 i_micro=1547 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.2733 mask_tok=82.00 y_norm=48.0882±0.0053 +2026-05-10 16:53:53,900 - INFO - Step 7750 | mode=video loss=0.058350 diff=0.058350 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0731 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0890 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=6201 i_micro=1549 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3467 mask_tok=104.00 y_norm=48.0890±0.0061 +2026-05-10 16:54:12,621 - INFO - Step 7760 | mode=video loss=0.078613 diff=0.078613 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0386 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0884 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=6209 i_micro=1551 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3700 mask_tok=111.00 y_norm=48.0884±0.0059 +2026-05-10 16:54:31,107 - INFO - Step 7770 | mode=video loss=0.092285 diff=0.092285 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1158 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0858 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=6217 i_micro=1553 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4500 mask_tok=135.00 y_norm=48.0858±0.0067 +2026-05-10 16:54:49,089 - INFO - Step 7780 | mode=video loss=0.109863 diff=0.109863 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0259 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0881 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=6225 i_micro=1555 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4067 mask_tok=122.00 y_norm=48.0881±0.0065 +2026-05-10 16:55:07,759 - INFO - Step 7790 | mode=video loss=0.079102 diff=0.079102 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0522 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0884 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=6233 i_micro=1557 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.2633 mask_tok=79.00 y_norm=48.0884±0.0057 +2026-05-10 16:55:27,865 - INFO - Step 7800 | mode=video loss=0.189453 diff=0.189453 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0758 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0885 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=6241 i_micro=1559 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3533 mask_tok=106.00 y_norm=48.0885±0.0058 cond_uncond_dloss=0.656250 cond_grad=0.000773 cond_pred_l2=43.177227 cond_pred_ratio=0.028536 +2026-05-10 16:55:27,903 - INFO - Step 7800 | probe_semantic mcp_offdiag(mean/min/max)=0.647616/0.534702/0.709483 smol_offdiag=0.660557 prompts=4 mcp_tok=17.50 smol_tok=17.50 +2026-05-10 16:55:34,887 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 16:55:52,856 - INFO - Step 7810 | mode=video loss=0.096191 diff=0.096191 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0545 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0869 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=6249 i_micro=1561 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4700 mask_tok=141.00 y_norm=48.0869±0.0059 +2026-05-10 16:56:12,141 - INFO - Step 7820 | mode=video loss=0.164062 diff=0.164062 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0996 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0880 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=6257 i_micro=1563 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3533 mask_tok=106.00 y_norm=48.0880±0.0056 +2026-05-10 16:56:30,135 - INFO - Step 7830 | mode=video loss=0.085449 diff=0.085449 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0420 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0876 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=6265 i_micro=1565 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3633 mask_tok=109.00 y_norm=48.0876±0.0055 +2026-05-10 16:56:48,926 - INFO - Step 7840 | mode=video loss=0.081055 diff=0.081055 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0313 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0889 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=6273 i_micro=1567 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3800 mask_tok=114.00 y_norm=48.0889±0.0073 +2026-05-10 16:57:07,380 - INFO - Step 7850 | mode=video loss=0.077148 diff=0.077148 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0414 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0887 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=6281 i_micro=1569 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4300 mask_tok=129.00 y_norm=48.0887±0.0068 +2026-05-10 16:57:25,343 - INFO - Step 7860 | mode=video loss=0.153320 diff=0.153320 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0376 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0882 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=6289 i_micro=1571 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3400 mask_tok=102.00 y_norm=48.0882±0.0055 +2026-05-10 16:57:44,394 - INFO - Step 7870 | mode=video loss=0.087891 diff=0.087891 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0381 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0859 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=6297 i_micro=1573 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4167 mask_tok=125.00 y_norm=48.0859±0.0056 +2026-05-10 16:58:02,368 - INFO - Step 7880 | mode=video loss=0.059570 diff=0.059570 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0700 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0873 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=6305 i_micro=1575 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3500 mask_tok=105.00 y_norm=48.0873±0.0057 +2026-05-10 16:58:21,063 - INFO - Step 7890 | mode=video loss=0.114746 diff=0.114746 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0604 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0870 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=6313 i_micro=1577 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3800 mask_tok=114.00 y_norm=48.0870±0.0055 +2026-05-10 16:58:39,461 - INFO - Step 7900 | mode=video loss=0.110840 diff=0.110840 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0453 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0873 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=6321 i_micro=1579 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3400 mask_tok=102.00 y_norm=48.0873±0.0049 +2026-05-10 16:58:46,379 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 16:59:04,311 - INFO - Step 7910 | mode=video loss=0.055420 diff=0.055420 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0470 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0853 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=6329 i_micro=1581 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3633 mask_tok=109.00 y_norm=48.0853±0.0062 +2026-05-10 16:59:22,932 - INFO - Step 7920 | mode=video loss=0.046387 diff=0.046387 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0506 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0841 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=6337 i_micro=1583 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3333 mask_tok=100.00 y_norm=48.0841±0.0054 +2026-05-10 16:59:41,385 - INFO - Step 7930 | mode=video loss=0.091309 diff=0.091309 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0891 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0857 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=6345 i_micro=1585 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4167 mask_tok=125.00 y_norm=48.0857±0.0055 +2026-05-10 16:59:59,342 - INFO - Step 7940 | mode=video loss=0.082031 diff=0.082031 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0621 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0882 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=6353 i_micro=1587 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.2867 mask_tok=86.00 y_norm=48.0882±0.0056 +2026-05-10 17:00:18,389 - INFO - Step 7950 | mode=video loss=0.065918 diff=0.065918 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0996 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0869 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=6361 i_micro=1589 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3300 mask_tok=99.00 y_norm=48.0869±0.0058 +2026-05-10 17:00:36,330 - INFO - Step 7960 | mode=video loss=0.201172 diff=0.201172 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1010 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0880 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=6369 i_micro=1591 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3433 mask_tok=103.00 y_norm=48.0880±0.0057 +2026-05-10 17:00:54,788 - INFO - Step 7970 | mode=video loss=0.112793 diff=0.112793 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0287 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0889 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=6377 i_micro=1593 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4967 mask_tok=149.00 y_norm=48.0889±0.0061 +2026-05-10 17:01:13,184 - INFO - Step 7980 | mode=video loss=0.102539 diff=0.102539 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0384 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0856 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=6385 i_micro=1595 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4300 mask_tok=129.00 y_norm=48.0856±0.0064 +2026-05-10 17:01:31,132 - INFO - Step 7990 | mode=video loss=0.121094 diff=0.121094 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0454 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0859 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=6393 i_micro=1597 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3467 mask_tok=104.00 y_norm=48.0859±0.0068 +2026-05-10 17:01:51,760 - INFO - Step 8000 | mode=video loss=0.113770 diff=0.113770 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0824 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0866 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=6401 i_micro=1599 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.5667 mask_tok=170.00 y_norm=48.0866±0.0068 cond_uncond_dloss=0.894531 cond_grad=0.000724 cond_pred_l2=28.435213 cond_pred_ratio=0.018350 +2026-05-10 17:01:51,798 - INFO - Step 8000 | probe_semantic mcp_offdiag(mean/min/max)=0.647034/0.533801/0.709026 smol_offdiag=0.660557 prompts=4 mcp_tok=17.50 smol_tok=17.50 +2026-05-10 17:01:58,595 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 17:02:16,549 - INFO - Step 8010 | mode=video loss=0.079102 diff=0.079102 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0370 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0885 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=6409 i_micro=1601 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3600 mask_tok=108.00 y_norm=48.0885±0.0052 +2026-05-10 17:02:35,249 - INFO - Step 8020 | mode=video loss=0.226562 diff=0.226562 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1206 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0872 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=6417 i_micro=1603 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.2833 mask_tok=85.00 y_norm=48.0872±0.0053 +2026-05-10 17:02:53,698 - INFO - Step 8030 | mode=video loss=0.122559 diff=0.122559 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0321 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0882 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=6425 i_micro=1605 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3600 mask_tok=108.00 y_norm=48.0882±0.0055 +2026-05-10 17:03:11,672 - INFO - Step 8040 | mode=video loss=0.116211 diff=0.116211 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0959 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0860 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=6433 i_micro=1607 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3767 mask_tok=113.00 y_norm=48.0860±0.0054 +2026-05-10 17:03:30,275 - INFO - Step 8050 | mode=video loss=0.071289 diff=0.071289 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1174 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0871 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=6441 i_micro=1609 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4433 mask_tok=133.00 y_norm=48.0871±0.0070 +2026-05-10 17:03:48,641 - INFO - Step 8060 | mode=video loss=0.157227 diff=0.157227 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0665 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0853 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=6449 i_micro=1611 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4167 mask_tok=125.00 y_norm=48.0853±0.0056 +2026-05-10 17:04:06,598 - INFO - Step 8070 | mode=video loss=0.292969 diff=0.292969 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0410 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0854 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=6457 i_micro=1613 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4867 mask_tok=146.00 y_norm=48.0854±0.0068 +2026-05-10 17:04:25,694 - INFO - Step 8080 | mode=video loss=0.070312 diff=0.070312 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0642 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0858 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=6465 i_micro=1615 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4767 mask_tok=143.00 y_norm=48.0858±0.0061 +2026-05-10 17:04:43,645 - INFO - Step 8090 | mode=video loss=0.101074 diff=0.101074 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0522 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0874 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=6473 i_micro=1617 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.5300 mask_tok=159.00 y_norm=48.0874±0.0069 +2026-05-10 17:05:02,324 - INFO - Step 8100 | mode=video loss=0.083984 diff=0.083984 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1027 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0868 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=6481 i_micro=1619 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4467 mask_tok=134.00 y_norm=48.0868±0.0063 +2026-05-10 17:05:09,199 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 17:05:27,590 - INFO - Step 8110 | mode=video loss=0.053955 diff=0.053955 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1337 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0857 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=6489 i_micro=1621 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3967 mask_tok=119.00 y_norm=48.0857±0.0059 +2026-05-10 17:05:45,541 - INFO - Step 8120 | mode=video loss=0.069336 diff=0.069336 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0305 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0879 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=6497 i_micro=1623 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3367 mask_tok=101.00 y_norm=48.0879±0.0056 +2026-05-10 17:06:04,599 - INFO - Step 8130 | mode=video loss=0.104004 diff=0.104004 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0419 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0867 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=6505 i_micro=1625 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3800 mask_tok=114.00 y_norm=48.0867±0.0054 +2026-05-10 17:06:22,559 - INFO - Step 8140 | mode=video loss=0.050537 diff=0.050537 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0379 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0872 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=6513 i_micro=1627 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3533 mask_tok=106.00 y_norm=48.0872±0.0065 +2026-05-10 17:06:41,290 - INFO - Step 8150 | mode=video loss=0.063965 diff=0.063965 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0662 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0884 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=6521 i_micro=1629 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3567 mask_tok=107.00 y_norm=48.0884±0.0064 +2026-05-10 17:06:59,788 - INFO - Step 8160 | mode=video loss=0.112305 diff=0.112305 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0640 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0875 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=6529 i_micro=1631 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4767 mask_tok=143.00 y_norm=48.0875±0.0061 +2026-05-10 17:07:17,770 - INFO - Step 8170 | mode=video loss=0.200195 diff=0.200195 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0612 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0872 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=6537 i_micro=1633 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.2767 mask_tok=83.00 y_norm=48.0872±0.0053 +2026-05-10 17:07:36,337 - INFO - Step 8180 | mode=video loss=0.062256 diff=0.062256 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0349 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0889 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=6545 i_micro=1635 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3067 mask_tok=92.00 y_norm=48.0889±0.0046 +2026-05-10 17:07:54,815 - INFO - Step 8190 | mode=video loss=0.098145 diff=0.098145 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1060 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0874 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=6553 i_micro=1637 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3433 mask_tok=103.00 y_norm=48.0874±0.0057 +2026-05-10 17:08:14,385 - INFO - Step 8200 | mode=video loss=0.070312 diff=0.070312 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0412 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0864 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=6561 i_micro=1639 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4300 mask_tok=129.00 y_norm=48.0864±0.0059 cond_uncond_dloss=0.128906 cond_grad=0.000057 cond_pred_l2=34.656204 cond_pred_ratio=0.019144 +2026-05-10 17:08:14,422 - INFO - Step 8200 | probe_semantic mcp_offdiag(mean/min/max)=0.647486/0.534481/0.709294 smol_offdiag=0.660557 prompts=4 mcp_tok=17.50 smol_tok=17.50 +2026-05-10 17:08:21,844 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 17:08:40,930 - INFO - Step 8210 | mode=video loss=0.087891 diff=0.087891 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0834 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0883 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=6569 i_micro=1641 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3767 mask_tok=113.00 y_norm=48.0883±0.0069 +2026-05-10 17:08:58,891 - INFO - Step 8220 | mode=video loss=0.074219 diff=0.074219 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0820 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0874 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=6577 i_micro=1643 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3567 mask_tok=107.00 y_norm=48.0874±0.0056 +2026-05-10 17:09:17,559 - INFO - Step 8230 | mode=video loss=0.108398 diff=0.108398 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0396 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0881 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=6585 i_micro=1645 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3300 mask_tok=99.00 y_norm=48.0881±0.0054 +2026-05-10 17:09:35,999 - INFO - Step 8240 | mode=video loss=0.166992 diff=0.166992 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0760 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0887 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=6593 i_micro=1647 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3567 mask_tok=107.00 y_norm=48.0887±0.0052 +2026-05-10 17:09:53,962 - INFO - Step 8250 | mode=video loss=0.125000 diff=0.125000 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0254 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0885 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=6601 i_micro=1649 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3467 mask_tok=104.00 y_norm=48.0885±0.0054 +2026-05-10 17:10:13,098 - INFO - Step 8260 | mode=video loss=0.125000 diff=0.125000 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0488 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0873 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=6609 i_micro=1651 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4533 mask_tok=136.00 y_norm=48.0873±0.0068 +2026-05-10 17:10:31,056 - INFO - Step 8270 | mode=video loss=0.097656 diff=0.097656 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0317 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0897 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=6617 i_micro=1653 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3567 mask_tok=107.00 y_norm=48.0897±0.0060 +2026-05-10 17:10:49,727 - INFO - Step 8280 | mode=video loss=0.067871 diff=0.067871 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0789 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0871 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=6625 i_micro=1655 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3900 mask_tok=117.00 y_norm=48.0871±0.0053 +2026-05-10 17:11:08,175 - INFO - Step 8290 | mode=video loss=0.271484 diff=0.271484 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0377 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0862 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=6633 i_micro=1657 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3600 mask_tok=108.00 y_norm=48.0862±0.0056 +2026-05-10 17:11:26,121 - INFO - Step 8300 | mode=video loss=0.132812 diff=0.132812 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0530 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0868 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=6641 i_micro=1659 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3300 mask_tok=99.00 y_norm=48.0868±0.0068 +2026-05-10 17:11:32,845 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 17:11:52,049 - INFO - Step 8310 | mode=video loss=0.106445 diff=0.106445 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0406 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0879 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=6649 i_micro=1661 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3567 mask_tok=107.00 y_norm=48.0879±0.0054 +2026-05-10 17:12:10,002 - INFO - Step 8320 | mode=video loss=0.068359 diff=0.068359 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0226 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0887 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=6657 i_micro=1663 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4000 mask_tok=120.00 y_norm=48.0887±0.0055 +2026-05-10 17:12:27,966 - INFO - Step 8330 | mode=video loss=0.094727 diff=0.094727 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0281 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0886 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=6665 i_micro=1665 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4000 mask_tok=120.00 y_norm=48.0886±0.0053 +2026-05-10 17:12:47,041 - INFO - Step 8340 | mode=video loss=0.062500 diff=0.062500 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0428 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0874 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=6673 i_micro=1667 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4800 mask_tok=144.00 y_norm=48.0874±0.0067 +2026-05-10 17:13:05,001 - INFO - Step 8350 | mode=video loss=0.193359 diff=0.193359 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1055 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0881 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=6681 i_micro=1669 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3100 mask_tok=93.00 y_norm=48.0881±0.0063 +2026-05-10 17:13:23,453 - INFO - Step 8360 | mode=video loss=0.067383 diff=0.067383 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0354 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0856 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=6689 i_micro=1671 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3733 mask_tok=112.00 y_norm=48.0856±0.0058 +2026-05-10 17:13:41,855 - INFO - Step 8370 | mode=video loss=0.077148 diff=0.077148 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0431 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0883 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=6697 i_micro=1673 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3500 mask_tok=105.00 y_norm=48.0883±0.0054 +2026-05-10 17:13:59,808 - INFO - Step 8380 | mode=video loss=0.084473 diff=0.084473 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0669 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0848 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=6705 i_micro=1675 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4233 mask_tok=127.00 y_norm=48.0848±0.0057 +2026-05-10 17:14:18,892 - INFO - Step 8390 | mode=video loss=0.099121 diff=0.099121 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0582 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0888 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=6713 i_micro=1677 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3667 mask_tok=110.00 y_norm=48.0888±0.0067 +2026-05-10 17:14:38,469 - INFO - Step 8400 | mode=video loss=0.078125 diff=0.078125 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0563 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0884 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=6721 i_micro=1679 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3967 mask_tok=119.00 y_norm=48.0884±0.0061 cond_uncond_dloss=0.312500 cond_grad=0.000113 cond_pred_l2=37.860531 cond_pred_ratio=0.023065 +2026-05-10 17:14:38,507 - INFO - Step 8400 | probe_semantic mcp_offdiag(mean/min/max)=0.646680/0.533353/0.709038 smol_offdiag=0.660557 prompts=4 mcp_tok=17.50 smol_tok=17.50 +2026-05-10 17:14:45,450 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 17:15:04,187 - INFO - Step 8410 | mode=video loss=0.094727 diff=0.094727 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0459 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0905 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=6729 i_micro=1681 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3467 mask_tok=104.00 y_norm=48.0905±0.0055 +2026-05-10 17:15:22,590 - INFO - Step 8420 | mode=video loss=0.114746 diff=0.114746 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0892 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0890 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=6737 i_micro=1683 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3200 mask_tok=96.00 y_norm=48.0890±0.0053 +2026-05-10 17:15:40,557 - INFO - Step 8430 | mode=video loss=0.094238 diff=0.094238 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0647 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0868 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=6745 i_micro=1685 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4800 mask_tok=144.00 y_norm=48.0868±0.0062 +2026-05-10 17:15:59,707 - INFO - Step 8440 | mode=video loss=0.082520 diff=0.082520 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0231 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0884 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=6753 i_micro=1687 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3633 mask_tok=109.00 y_norm=48.0884±0.0061 +2026-05-10 17:16:17,654 - INFO - Step 8450 | mode=video loss=0.072266 diff=0.072266 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0641 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0863 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=6761 i_micro=1689 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4033 mask_tok=121.00 y_norm=48.0863±0.0065 +2026-05-10 17:16:35,628 - INFO - Step 8460 | mode=video loss=0.153320 diff=0.153320 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0380 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0884 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=6769 i_micro=1691 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3533 mask_tok=106.00 y_norm=48.0884±0.0052 +2026-05-10 17:16:54,577 - INFO - Step 8470 | mode=video loss=0.056396 diff=0.056396 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0486 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0856 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=6777 i_micro=1693 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4433 mask_tok=133.00 y_norm=48.0856±0.0062 +2026-05-10 17:17:12,516 - INFO - Step 8480 | mode=video loss=0.038574 diff=0.038574 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0509 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0891 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=6785 i_micro=1695 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.2600 mask_tok=78.00 y_norm=48.0891±0.0050 +2026-05-10 17:17:31,177 - INFO - Step 8490 | mode=video loss=0.147461 diff=0.147461 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0623 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0885 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=6793 i_micro=1697 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3433 mask_tok=103.00 y_norm=48.0885±0.0056 +2026-05-10 17:17:49,547 - INFO - Step 8500 | mode=video loss=0.097656 diff=0.097656 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0342 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0876 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=6801 i_micro=1699 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3600 mask_tok=108.00 y_norm=48.0876±0.0055 +2026-05-10 17:17:56,275 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 17:18:14,218 - INFO - Step 8510 | mode=video loss=0.094238 diff=0.094238 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0336 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0883 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=6809 i_micro=1701 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3067 mask_tok=92.00 y_norm=48.0883±0.0052 +2026-05-10 17:18:33,402 - INFO - Step 8520 | mode=video loss=0.074707 diff=0.074707 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0269 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0889 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=6817 i_micro=1703 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3667 mask_tok=110.00 y_norm=48.0889±0.0059 +2026-05-10 17:18:51,379 - INFO - Step 8530 | mode=video loss=0.080078 diff=0.080078 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0350 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0848 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=6825 i_micro=1705 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3800 mask_tok=114.00 y_norm=48.0848±0.0063 +2026-05-10 17:19:10,076 - INFO - Step 8540 | mode=video loss=0.098633 diff=0.098633 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.2956 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0879 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=6833 i_micro=1707 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.2633 mask_tok=79.00 y_norm=48.0879±0.0055 +2026-05-10 17:19:28,512 - INFO - Step 8550 | mode=video loss=0.155273 diff=0.155273 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0503 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0894 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=6841 i_micro=1709 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3700 mask_tok=111.00 y_norm=48.0894±0.0059 +2026-05-10 17:19:46,475 - INFO - Step 8560 | mode=video loss=0.086426 diff=0.086426 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0466 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0872 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=6849 i_micro=1711 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4433 mask_tok=133.00 y_norm=48.0872±0.0065 +2026-05-10 17:20:05,194 - INFO - Step 8570 | mode=video loss=0.088867 diff=0.088867 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0264 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0888 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=6857 i_micro=1713 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.5500 mask_tok=165.00 y_norm=48.0888±0.0083 +2026-05-10 17:20:23,653 - INFO - Step 8580 | mode=video loss=0.090332 diff=0.090332 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0311 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0881 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=6865 i_micro=1715 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3467 mask_tok=104.00 y_norm=48.0881±0.0055 +2026-05-10 17:20:41,652 - INFO - Step 8590 | mode=video loss=0.104492 diff=0.104492 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0682 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0865 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=6873 i_micro=1717 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4100 mask_tok=123.00 y_norm=48.0865±0.0061 +2026-05-10 17:21:02,423 - INFO - Step 8600 | mode=video loss=0.116699 diff=0.116699 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0501 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0888 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=6881 i_micro=1719 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3000 mask_tok=90.00 y_norm=48.0888±0.0059 cond_uncond_dloss=0.332031 cond_grad=0.000123 cond_pred_l2=32.971832 cond_pred_ratio=0.019951 +2026-05-10 17:21:02,460 - INFO - Step 8600 | probe_semantic mcp_offdiag(mean/min/max)=0.646544/0.533112/0.709224 smol_offdiag=0.660557 prompts=4 mcp_tok=17.50 smol_tok=17.50 +2026-05-10 17:21:09,553 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 17:21:27,501 - INFO - Step 8610 | mode=video loss=0.071777 diff=0.071777 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0601 emb_mean=0.0007 emb_std=1.0018 tok_norm=48.0872 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=6889 i_micro=1721 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4167 mask_tok=125.00 y_norm=48.0872±0.0057 +2026-05-10 17:21:46,236 - INFO - Step 8620 | mode=video loss=0.099121 diff=0.099121 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0302 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0858 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=6897 i_micro=1723 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3933 mask_tok=118.00 y_norm=48.0858±0.0064 +2026-05-10 17:22:04,679 - INFO - Step 8630 | mode=video loss=0.107910 diff=0.107910 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0419 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0881 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=6905 i_micro=1725 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3867 mask_tok=116.00 y_norm=48.0881±0.0058 +2026-05-10 17:22:22,632 - INFO - Step 8640 | mode=video loss=0.079590 diff=0.079590 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0401 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0892 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=6913 i_micro=1727 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3200 mask_tok=96.00 y_norm=48.0892±0.0056 +2026-05-10 17:22:41,728 - INFO - Step 8650 | mode=video loss=0.068848 diff=0.068848 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0381 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0877 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=6921 i_micro=1729 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3467 mask_tok=104.00 y_norm=48.0877±0.0055 +2026-05-10 17:22:59,712 - INFO - Step 8660 | mode=video loss=0.091309 diff=0.091309 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0345 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0873 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=6929 i_micro=1731 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3600 mask_tok=108.00 y_norm=48.0873±0.0053 +2026-05-10 17:23:18,454 - INFO - Step 8670 | mode=video loss=0.121582 diff=0.121582 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0992 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0882 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=6937 i_micro=1733 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3600 mask_tok=108.00 y_norm=48.0882±0.0052 +2026-05-10 17:23:36,818 - INFO - Step 8680 | mode=video loss=0.102539 diff=0.102539 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0579 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0862 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=6945 i_micro=1735 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3600 mask_tok=108.00 y_norm=48.0862±0.0072 +2026-05-10 17:23:54,788 - INFO - Step 8690 | mode=video loss=0.094238 diff=0.094238 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0393 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0876 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=6953 i_micro=1737 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3400 mask_tok=102.00 y_norm=48.0876±0.0059 +2026-05-10 17:24:13,580 - INFO - Step 8700 | mode=video loss=0.085449 diff=0.085449 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0517 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0898 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=6961 i_micro=1739 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3000 mask_tok=90.00 y_norm=48.0898±0.0055 +2026-05-10 17:24:20,265 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 17:24:38,671 - INFO - Step 8710 | mode=video loss=0.128906 diff=0.128906 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0788 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0886 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=6969 i_micro=1741 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3400 mask_tok=102.00 y_norm=48.0886±0.0052 +2026-05-10 17:24:57,411 - INFO - Step 8720 | mode=video loss=0.110352 diff=0.110352 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0404 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0888 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=6977 i_micro=1743 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3200 mask_tok=96.00 y_norm=48.0888±0.0053 +2026-05-10 17:25:15,837 - INFO - Step 8730 | mode=video loss=0.080078 diff=0.080078 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0522 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0874 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=6985 i_micro=1745 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3633 mask_tok=109.00 y_norm=48.0874±0.0059 +2026-05-10 17:25:33,807 - INFO - Step 8740 | mode=video loss=0.052246 diff=0.052246 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0585 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0867 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=6993 i_micro=1747 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4500 mask_tok=135.00 y_norm=48.0867±0.0061 +2026-05-10 17:25:52,534 - INFO - Step 8750 | mode=video loss=0.073730 diff=0.073730 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0362 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0880 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=7001 i_micro=1749 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3533 mask_tok=106.00 y_norm=48.0880±0.0064 +2026-05-10 17:26:10,945 - INFO - Step 8760 | mode=video loss=0.076660 diff=0.076660 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1261 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0888 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=7009 i_micro=1751 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3300 mask_tok=99.00 y_norm=48.0888±0.0059 +2026-05-10 17:26:28,906 - INFO - Step 8770 | mode=video loss=0.102051 diff=0.102051 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0405 emb_mean=0.0007 emb_std=1.0018 tok_norm=48.0875 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=7017 i_micro=1753 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3833 mask_tok=115.00 y_norm=48.0875±0.0069 +2026-05-10 17:26:48,134 - INFO - Step 8780 | mode=video loss=0.107910 diff=0.107910 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0728 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0893 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=7025 i_micro=1755 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3900 mask_tok=117.00 y_norm=48.0893±0.0056 +2026-05-10 17:27:06,115 - INFO - Step 8790 | mode=video loss=0.052734 diff=0.052734 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0549 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0849 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=7033 i_micro=1757 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4367 mask_tok=131.00 y_norm=48.0849±0.0059 +2026-05-10 17:27:26,471 - INFO - Step 8800 | mode=video loss=0.092773 diff=0.092773 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0693 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0889 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=7041 i_micro=1759 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3567 mask_tok=107.00 y_norm=48.0889±0.0050 cond_uncond_dloss=0.671875 cond_grad=0.000610 cond_pred_l2=59.334499 cond_pred_ratio=0.037396 +2026-05-10 17:27:26,508 - INFO - Step 8800 | probe_semantic mcp_offdiag(mean/min/max)=0.646277/0.532587/0.708858 smol_offdiag=0.660557 prompts=4 mcp_tok=17.50 smol_tok=17.50 +2026-05-10 17:27:33,923 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 17:27:52,308 - INFO - Step 8810 | mode=video loss=0.089844 diff=0.089844 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0556 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0880 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=7049 i_micro=1761 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3033 mask_tok=91.00 y_norm=48.0880±0.0056 +2026-05-10 17:28:10,284 - INFO - Step 8820 | mode=video loss=0.096680 diff=0.096680 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0356 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0895 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=7057 i_micro=1763 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3967 mask_tok=119.00 y_norm=48.0895±0.0059 +2026-05-10 17:28:29,029 - INFO - Step 8830 | mode=video loss=0.119629 diff=0.119629 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0403 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0891 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=7065 i_micro=1765 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3600 mask_tok=108.00 y_norm=48.0891±0.0057 +2026-05-10 17:28:47,470 - INFO - Step 8840 | mode=video loss=0.044922 diff=0.044922 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0300 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0854 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=7073 i_micro=1767 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4467 mask_tok=134.00 y_norm=48.0854±0.0056 +2026-05-10 17:29:06,122 - INFO - Step 8850 | mode=video loss=0.275391 diff=0.275391 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.4138 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0916 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=7081 i_micro=1769 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.5500 mask_tok=165.00 y_norm=48.0916±0.0073 +2026-05-10 17:29:24,491 - INFO - Step 8860 | mode=video loss=0.115723 diff=0.115723 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0479 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0875 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=7089 i_micro=1771 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3533 mask_tok=106.00 y_norm=48.0875±0.0059 +2026-05-10 17:29:42,440 - INFO - Step 8870 | mode=video loss=0.067383 diff=0.067383 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0285 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0876 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=7097 i_micro=1773 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.2900 mask_tok=87.00 y_norm=48.0876±0.0053 +2026-05-10 17:30:01,139 - INFO - Step 8880 | mode=video loss=0.137695 diff=0.137695 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0359 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0881 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=7105 i_micro=1775 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3633 mask_tok=109.00 y_norm=48.0881±0.0054 +2026-05-10 17:30:19,481 - INFO - Step 8890 | mode=video loss=0.103027 diff=0.103027 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0516 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0881 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=7113 i_micro=1777 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3300 mask_tok=99.00 y_norm=48.0881±0.0060 +2026-05-10 17:30:37,429 - INFO - Step 8900 | mode=video loss=0.089844 diff=0.089844 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0894 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0877 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=7121 i_micro=1779 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3400 mask_tok=102.00 y_norm=48.0877±0.0056 +2026-05-10 17:30:44,355 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 17:31:02,697 - INFO - Step 8910 | mode=video loss=0.127930 diff=0.127930 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1244 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0874 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=7129 i_micro=1781 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3600 mask_tok=108.00 y_norm=48.0874±0.0057 +2026-05-10 17:31:20,646 - INFO - Step 8920 | mode=video loss=0.104492 diff=0.104492 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0540 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0873 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=7137 i_micro=1783 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3733 mask_tok=112.00 y_norm=48.0873±0.0065 +2026-05-10 17:31:39,381 - INFO - Step 8930 | mode=video loss=0.091309 diff=0.091309 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0544 emb_mean=0.0007 emb_std=1.0019 tok_norm=48.0891 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=7145 i_micro=1785 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3700 mask_tok=111.00 y_norm=48.0891±0.0056 +2026-05-10 17:31:57,735 - INFO - Step 8940 | mode=video loss=0.072754 diff=0.072754 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0984 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0878 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=7153 i_micro=1787 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3400 mask_tok=102.00 y_norm=48.0878±0.0059 +2026-05-10 17:32:15,714 - INFO - Step 8950 | mode=video loss=0.063477 diff=0.063477 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0412 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0876 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=7161 i_micro=1789 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3033 mask_tok=91.00 y_norm=48.0876±0.0059 +2026-05-10 17:32:34,433 - INFO - Step 8960 | mode=video loss=0.071777 diff=0.071777 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0201 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0863 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=7169 i_micro=1791 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4000 mask_tok=120.00 y_norm=48.0863±0.0056 +2026-05-10 17:32:52,849 - INFO - Step 8970 | mode=video loss=0.078613 diff=0.078613 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0231 emb_mean=0.0007 emb_std=1.0018 tok_norm=48.0869 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=7177 i_micro=1793 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3100 mask_tok=93.00 y_norm=48.0869±0.0057 +2026-05-10 17:33:11,582 - INFO - Step 8980 | mode=video loss=0.065430 diff=0.065430 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0381 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0875 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=7185 i_micro=1795 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3467 mask_tok=104.00 y_norm=48.0875±0.0050 +2026-05-10 17:33:29,944 - INFO - Step 8990 | mode=video loss=0.075195 diff=0.075195 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0603 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0870 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=7193 i_micro=1797 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3433 mask_tok=103.00 y_norm=48.0870±0.0068 +2026-05-10 17:33:49,556 - INFO - Step 9000 | mode=video loss=0.150391 diff=0.150391 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1053 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0873 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=7201 i_micro=1799 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3700 mask_tok=111.00 y_norm=48.0873±0.0055 cond_uncond_dloss=0.828125 cond_grad=0.000966 cond_pred_l2=49.276932 cond_pred_ratio=0.032576 +2026-05-10 17:33:49,592 - INFO - Step 9000 | probe_semantic mcp_offdiag(mean/min/max)=0.646659/0.532746/0.709251 smol_offdiag=0.660557 prompts=4 mcp_tok=17.50 smol_tok=17.50 +2026-05-10 17:33:56,407 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 17:34:15,167 - INFO - Step 9010 | mode=video loss=0.078125 diff=0.078125 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0601 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0861 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=7209 i_micro=1801 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4233 mask_tok=127.00 y_norm=48.0861±0.0059 +2026-05-10 17:34:33,577 - INFO - Step 9020 | mode=video loss=0.139648 diff=0.139648 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0395 emb_mean=0.0007 emb_std=1.0018 tok_norm=48.0868 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=7217 i_micro=1803 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3933 mask_tok=118.00 y_norm=48.0868±0.0062 +2026-05-10 17:34:52,290 - INFO - Step 9030 | mode=video loss=0.109863 diff=0.109863 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0298 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0864 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=7225 i_micro=1805 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4533 mask_tok=136.00 y_norm=48.0864±0.0062 +2026-05-10 17:35:10,685 - INFO - Step 9040 | mode=video loss=0.083496 diff=0.083496 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0633 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0873 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=7233 i_micro=1807 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3367 mask_tok=101.00 y_norm=48.0873±0.0049 +2026-05-10 17:35:28,653 - INFO - Step 9050 | mode=video loss=0.056152 diff=0.056152 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0367 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0874 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=7241 i_micro=1809 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4967 mask_tok=149.00 y_norm=48.0874±0.0066 +2026-05-10 17:35:47,354 - INFO - Step 9060 | mode=video loss=0.212891 diff=0.212891 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1709 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0884 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=7249 i_micro=1811 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3433 mask_tok=103.00 y_norm=48.0884±0.0060 +2026-05-10 17:36:05,724 - INFO - Step 9070 | mode=video loss=0.130859 diff=0.130859 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0369 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0855 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=7257 i_micro=1813 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3633 mask_tok=109.00 y_norm=48.0855±0.0056 +2026-05-10 17:36:23,690 - INFO - Step 9080 | mode=video loss=0.324219 diff=0.324219 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0514 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0869 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=7265 i_micro=1815 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3467 mask_tok=104.00 y_norm=48.0869±0.0052 +2026-05-10 17:36:42,379 - INFO - Step 9090 | mode=video loss=0.231445 diff=0.231445 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1383 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0888 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=7273 i_micro=1817 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3867 mask_tok=116.00 y_norm=48.0888±0.0054 +2026-05-10 17:37:00,820 - INFO - Step 9100 | mode=video loss=0.066895 diff=0.066895 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0412 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0868 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=7281 i_micro=1819 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3767 mask_tok=113.00 y_norm=48.0868±0.0061 +2026-05-10 17:37:07,604 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 17:37:26,326 - INFO - Step 9110 | mode=video loss=0.096191 diff=0.096191 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0403 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0872 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=7289 i_micro=1821 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4133 mask_tok=124.00 y_norm=48.0872±0.0055 +2026-05-10 17:37:44,704 - INFO - Step 9120 | mode=video loss=0.128906 diff=0.128906 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0561 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0878 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=7297 i_micro=1823 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3367 mask_tok=101.00 y_norm=48.0878±0.0055 +2026-05-10 17:38:02,687 - INFO - Step 9130 | mode=video loss=0.083008 diff=0.083008 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0699 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0866 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=7305 i_micro=1825 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3933 mask_tok=118.00 y_norm=48.0866±0.0065 +2026-05-10 17:38:21,310 - INFO - Step 9140 | mode=video loss=0.101562 diff=0.101562 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0582 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0891 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=7313 i_micro=1827 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4133 mask_tok=124.00 y_norm=48.0891±0.0058 +2026-05-10 17:38:39,675 - INFO - Step 9150 | mode=video loss=0.101074 diff=0.101074 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0864 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0894 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=7321 i_micro=1829 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3500 mask_tok=105.00 y_norm=48.0894±0.0057 +2026-05-10 17:38:58,181 - INFO - Step 9160 | mode=video loss=0.082520 diff=0.082520 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0870 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0860 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=7329 i_micro=1831 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3633 mask_tok=109.00 y_norm=48.0860±0.0062 +2026-05-10 17:39:16,597 - INFO - Step 9170 | mode=video loss=0.125977 diff=0.125977 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0396 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0857 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=7337 i_micro=1833 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.5333 mask_tok=160.00 y_norm=48.0857±0.0062 +2026-05-10 17:39:34,545 - INFO - Step 9180 | mode=video loss=0.175781 diff=0.175781 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1051 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0888 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=7345 i_micro=1835 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4067 mask_tok=122.00 y_norm=48.0888±0.0058 +2026-05-10 17:39:53,133 - INFO - Step 9190 | mode=video loss=0.135742 diff=0.135742 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0369 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0894 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=7353 i_micro=1837 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3033 mask_tok=91.00 y_norm=48.0894±0.0060 +2026-05-10 17:40:13,146 - INFO - Step 9200 | mode=video loss=0.105957 diff=0.105957 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0367 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0880 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=7361 i_micro=1839 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3400 mask_tok=102.00 y_norm=48.0880±0.0060 cond_uncond_dloss=0.320312 cond_grad=0.000149 cond_pred_l2=42.049534 cond_pred_ratio=0.025727 +2026-05-10 17:40:13,183 - INFO - Step 9200 | probe_semantic mcp_offdiag(mean/min/max)=0.646397/0.532414/0.709084 smol_offdiag=0.660557 prompts=4 mcp_tok=17.50 smol_tok=17.50 +2026-05-10 17:40:19,923 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 17:40:38,673 - INFO - Step 9210 | mode=video loss=0.074219 diff=0.074219 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0586 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0849 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=7369 i_micro=1841 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3600 mask_tok=108.00 y_norm=48.0849±0.0064 +2026-05-10 17:40:56,639 - INFO - Step 9220 | mode=video loss=0.171875 diff=0.171875 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.3037 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0869 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=7377 i_micro=1843 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3333 mask_tok=100.00 y_norm=48.0869±0.0056 +2026-05-10 17:41:15,102 - INFO - Step 9230 | mode=video loss=0.113281 diff=0.113281 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1253 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0888 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=7385 i_micro=1845 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3400 mask_tok=102.00 y_norm=48.0888±0.0057 +2026-05-10 17:41:33,761 - INFO - Step 9240 | mode=video loss=0.123047 diff=0.123047 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0387 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0858 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=7393 i_micro=1847 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3900 mask_tok=117.00 y_norm=48.0858±0.0062 +2026-05-10 17:41:52,237 - INFO - Step 9250 | mode=video loss=0.086426 diff=0.086426 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0399 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0854 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=7401 i_micro=1849 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4400 mask_tok=132.00 y_norm=48.0854±0.0069 +2026-05-10 17:42:10,196 - INFO - Step 9260 | mode=video loss=0.107910 diff=0.107910 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0458 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0892 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=7409 i_micro=1851 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3300 mask_tok=99.00 y_norm=48.0892±0.0061 +2026-05-10 17:42:28,833 - INFO - Step 9270 | mode=video loss=0.103027 diff=0.103027 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0813 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0886 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=7417 i_micro=1853 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3867 mask_tok=116.00 y_norm=48.0886±0.0065 +2026-05-10 17:42:47,199 - INFO - Step 9280 | mode=video loss=0.113770 diff=0.113770 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0272 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0886 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=7425 i_micro=1855 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3800 mask_tok=114.00 y_norm=48.0886±0.0052 +2026-05-10 17:43:05,768 - INFO - Step 9290 | mode=video loss=0.103027 diff=0.103027 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0402 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0894 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=7433 i_micro=1857 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3300 mask_tok=99.00 y_norm=48.0894±0.0057 +2026-05-10 17:43:24,141 - INFO - Step 9300 | mode=video loss=0.128906 diff=0.128906 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0385 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0858 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=7441 i_micro=1859 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4933 mask_tok=148.00 y_norm=48.0858±0.0061 +2026-05-10 17:43:30,821 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 17:43:48,768 - INFO - Step 9310 | mode=video loss=0.106445 diff=0.106445 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0505 emb_mean=0.0007 emb_std=1.0018 tok_norm=48.0887 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=7449 i_micro=1861 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.2967 mask_tok=89.00 y_norm=48.0887±0.0053 +2026-05-10 17:44:07,550 - INFO - Step 9320 | mode=video loss=0.102539 diff=0.102539 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0754 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0870 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=7457 i_micro=1863 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3967 mask_tok=119.00 y_norm=48.0870±0.0057 +2026-05-10 17:44:25,936 - INFO - Step 9330 | mode=video loss=0.067383 diff=0.067383 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0250 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0867 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=7465 i_micro=1865 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3367 mask_tok=101.00 y_norm=48.0867±0.0065 +2026-05-10 17:44:44,659 - INFO - Step 9340 | mode=video loss=0.081543 diff=0.081543 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0342 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0873 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=7473 i_micro=1867 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4333 mask_tok=130.00 y_norm=48.0873±0.0063 +2026-05-10 17:45:02,600 - INFO - Step 9350 | mode=video loss=0.082031 diff=0.082031 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1916 emb_mean=0.0008 emb_std=1.0017 tok_norm=48.0839 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=7481 i_micro=1869 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4967 mask_tok=149.00 y_norm=48.0839±0.0064 +2026-05-10 17:45:20,947 - INFO - Step 9360 | mode=video loss=0.082031 diff=0.082031 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0492 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0870 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=7489 i_micro=1871 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3733 mask_tok=112.00 y_norm=48.0870±0.0060 +2026-05-10 17:45:39,500 - INFO - Step 9370 | mode=video loss=0.134766 diff=0.134766 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0864 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0887 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=7497 i_micro=1873 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3633 mask_tok=109.00 y_norm=48.0887±0.0050 +2026-05-10 17:45:57,856 - INFO - Step 9380 | mode=video loss=0.110840 diff=0.110840 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0934 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0870 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=7505 i_micro=1875 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3633 mask_tok=109.00 y_norm=48.0870±0.0062 +2026-05-10 17:46:16,545 - INFO - Step 9390 | mode=video loss=0.104004 diff=0.104004 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0525 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0874 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=7513 i_micro=1877 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3700 mask_tok=111.00 y_norm=48.0874±0.0057 +2026-05-10 17:46:36,120 - INFO - Step 9400 | mode=video loss=0.096191 diff=0.096191 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0281 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0884 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=7521 i_micro=1879 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.2400 mask_tok=72.00 y_norm=48.0884±0.0052 cond_uncond_dloss=0.228516 cond_grad=0.000042 cond_pred_l2=39.674355 cond_pred_ratio=0.022612 +2026-05-10 17:46:36,157 - INFO - Step 9400 | probe_semantic mcp_offdiag(mean/min/max)=0.646841/0.532948/0.709370 smol_offdiag=0.660557 prompts=4 mcp_tok=17.50 smol_tok=17.50 +2026-05-10 17:46:42,863 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 17:47:01,214 - INFO - Step 9410 | mode=video loss=0.077637 diff=0.077637 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0610 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0868 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=7529 i_micro=1881 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4400 mask_tok=132.00 y_norm=48.0868±0.0066 +2026-05-10 17:47:19,953 - INFO - Step 9420 | mode=video loss=0.090332 diff=0.090332 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0323 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0893 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=7537 i_micro=1883 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3367 mask_tok=101.00 y_norm=48.0893±0.0061 +2026-05-10 17:47:38,347 - INFO - Step 9430 | mode=video loss=0.089844 diff=0.089844 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0689 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0870 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=7545 i_micro=1885 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3633 mask_tok=109.00 y_norm=48.0870±0.0058 +2026-05-10 17:47:56,318 - INFO - Step 9440 | mode=video loss=0.063477 diff=0.063477 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0308 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0889 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=7553 i_micro=1887 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3433 mask_tok=103.00 y_norm=48.0889±0.0053 +2026-05-10 17:48:14,967 - INFO - Step 9450 | mode=video loss=0.067871 diff=0.067871 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0357 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0885 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=7561 i_micro=1889 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3400 mask_tok=102.00 y_norm=48.0885±0.0054 +2026-05-10 17:48:33,355 - INFO - Step 9460 | mode=video loss=0.100586 diff=0.100586 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0738 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0888 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=7569 i_micro=1891 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3300 mask_tok=99.00 y_norm=48.0888±0.0060 +2026-05-10 17:48:52,039 - INFO - Step 9470 | mode=video loss=0.072754 diff=0.072754 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0460 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0882 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=7577 i_micro=1893 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3367 mask_tok=101.00 y_norm=48.0882±0.0060 +2026-05-10 17:49:09,986 - INFO - Step 9480 | mode=video loss=0.123535 diff=0.123535 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0307 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0858 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=7585 i_micro=1895 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4033 mask_tok=121.00 y_norm=48.0858±0.0060 +2026-05-10 17:49:28,339 - INFO - Step 9490 | mode=video loss=0.098145 diff=0.098145 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0285 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0898 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=7593 i_micro=1897 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3733 mask_tok=112.00 y_norm=48.0898±0.0061 +2026-05-10 17:49:47,047 - INFO - Step 9500 | mode=video loss=0.341797 diff=0.341797 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.4057 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0888 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=7601 i_micro=1899 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3333 mask_tok=100.00 y_norm=48.0888±0.0055 +2026-05-10 17:49:53,809 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 17:50:12,168 - INFO - Step 9510 | mode=video loss=0.078613 diff=0.078613 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0433 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0869 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=7609 i_micro=1901 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3300 mask_tok=99.00 y_norm=48.0869±0.0054 +2026-05-10 17:50:30,893 - INFO - Step 9520 | mode=video loss=0.118652 diff=0.118652 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0530 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0884 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=7617 i_micro=1903 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3433 mask_tok=103.00 y_norm=48.0884±0.0052 +2026-05-10 17:50:48,858 - INFO - Step 9530 | mode=video loss=0.063965 diff=0.063965 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0418 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0866 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=7625 i_micro=1905 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4067 mask_tok=122.00 y_norm=48.0866±0.0070 +2026-05-10 17:51:07,225 - INFO - Step 9540 | mode=video loss=0.076660 diff=0.076660 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0428 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0904 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=7633 i_micro=1907 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3700 mask_tok=111.00 y_norm=48.0904±0.0059 +2026-05-10 17:51:25,701 - INFO - Step 9550 | mode=video loss=0.103027 diff=0.103027 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0252 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0866 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=7641 i_micro=1909 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3467 mask_tok=104.00 y_norm=48.0866±0.0061 +2026-05-10 17:51:44,090 - INFO - Step 9560 | mode=video loss=0.198242 diff=0.198242 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.2153 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0883 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=7649 i_micro=1911 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3100 mask_tok=93.00 y_norm=48.0883±0.0055 +2026-05-10 17:52:02,046 - INFO - Step 9570 | mode=video loss=0.091309 diff=0.091309 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0416 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0900 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=7657 i_micro=1913 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3867 mask_tok=116.00 y_norm=48.0900±0.0052 +2026-05-10 17:52:20,499 - INFO - Step 9580 | mode=video loss=0.261719 diff=0.261719 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0275 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0884 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=7665 i_micro=1915 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3367 mask_tok=101.00 y_norm=48.0884±0.0054 +2026-05-10 17:52:38,952 - INFO - Step 9590 | mode=video loss=0.221680 diff=0.221680 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1430 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0854 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=7673 i_micro=1917 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3767 mask_tok=113.00 y_norm=48.0854±0.0064 +2026-05-10 17:52:59,282 - INFO - Step 9600 | mode=video loss=0.077148 diff=0.077148 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0382 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0862 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=7681 i_micro=1919 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3667 mask_tok=110.00 y_norm=48.0862±0.0063 cond_uncond_dloss=0.174805 cond_grad=0.000044 cond_pred_l2=28.617594 cond_pred_ratio=0.016276 +2026-05-10 17:52:59,319 - INFO - Step 9600 | probe_semantic mcp_offdiag(mean/min/max)=0.646651/0.532984/0.709294 smol_offdiag=0.660557 prompts=4 mcp_tok=17.50 smol_tok=17.50 +2026-05-10 17:53:06,250 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 17:53:24,632 - INFO - Step 9610 | mode=video loss=0.089844 diff=0.089844 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0299 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0873 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=7689 i_micro=1921 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.2867 mask_tok=86.00 y_norm=48.0873±0.0056 +2026-05-10 17:53:42,595 - INFO - Step 9620 | mode=video loss=0.103516 diff=0.103516 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0409 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0888 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=7697 i_micro=1923 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4033 mask_tok=121.00 y_norm=48.0888±0.0059 +2026-05-10 17:54:01,307 - INFO - Step 9630 | mode=video loss=0.063965 diff=0.063965 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0318 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0880 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=7705 i_micro=1925 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3400 mask_tok=102.00 y_norm=48.0880±0.0056 +2026-05-10 17:54:19,732 - INFO - Step 9640 | mode=video loss=0.054688 diff=0.054688 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0648 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0876 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=7713 i_micro=1927 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4667 mask_tok=140.00 y_norm=48.0876±0.0064 +2026-05-10 17:54:38,413 - INFO - Step 9650 | mode=video loss=0.100098 diff=0.100098 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0406 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0876 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=7721 i_micro=1929 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3833 mask_tok=115.00 y_norm=48.0876±0.0057 +2026-05-10 17:54:56,369 - INFO - Step 9660 | mode=video loss=0.117188 diff=0.117188 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0557 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0882 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=7729 i_micro=1931 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.2767 mask_tok=83.00 y_norm=48.0882±0.0052 +2026-05-10 17:55:14,796 - INFO - Step 9670 | mode=video loss=0.091797 diff=0.091797 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0592 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0884 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=7737 i_micro=1933 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.5133 mask_tok=154.00 y_norm=48.0884±0.0066 +2026-05-10 17:55:33,422 - INFO - Step 9680 | mode=video loss=0.081055 diff=0.081055 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0614 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0887 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=7745 i_micro=1935 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3400 mask_tok=102.00 y_norm=48.0887±0.0051 +2026-05-10 17:55:51,855 - INFO - Step 9690 | mode=video loss=0.109375 diff=0.109375 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0486 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0855 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=7753 i_micro=1937 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3433 mask_tok=103.00 y_norm=48.0855±0.0062 +2026-05-10 17:56:09,810 - INFO - Step 9700 | mode=video loss=0.072266 diff=0.072266 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0788 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0875 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=7761 i_micro=1939 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4800 mask_tok=144.00 y_norm=48.0875±0.0067 +2026-05-10 17:56:16,671 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 17:56:35,396 - INFO - Step 9710 | mode=video loss=0.121094 diff=0.121094 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0788 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0879 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=7769 i_micro=1941 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3467 mask_tok=104.00 y_norm=48.0879±0.0052 +2026-05-10 17:56:53,817 - INFO - Step 9720 | mode=video loss=0.091797 diff=0.091797 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0466 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0887 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=7777 i_micro=1943 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3300 mask_tok=99.00 y_norm=48.0887±0.0050 +2026-05-10 17:57:12,594 - INFO - Step 9730 | mode=video loss=0.075684 diff=0.075684 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0500 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0857 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=7785 i_micro=1945 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4400 mask_tok=132.00 y_norm=48.0857±0.0061 +2026-05-10 17:57:30,574 - INFO - Step 9740 | mode=video loss=0.094238 diff=0.094238 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0615 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0883 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=7793 i_micro=1947 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4100 mask_tok=123.00 y_norm=48.0883±0.0056 +2026-05-10 17:57:49,009 - INFO - Step 9750 | mode=video loss=0.062256 diff=0.062256 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0434 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0869 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=7801 i_micro=1949 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4567 mask_tok=137.00 y_norm=48.0869±0.0060 +2026-05-10 17:58:07,458 - INFO - Step 9760 | mode=video loss=0.088379 diff=0.088379 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0579 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0902 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=7809 i_micro=1951 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3467 mask_tok=104.00 y_norm=48.0902±0.0053 +2026-05-10 17:58:25,846 - INFO - Step 9770 | mode=video loss=0.066895 diff=0.066895 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0422 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0890 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=7817 i_micro=1953 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.2333 mask_tok=70.00 y_norm=48.0890±0.0054 +2026-05-10 17:58:44,463 - INFO - Step 9780 | mode=video loss=0.136719 diff=0.136719 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0417 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0867 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=7825 i_micro=1955 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4533 mask_tok=136.00 y_norm=48.0867±0.0062 +2026-05-10 17:59:02,410 - INFO - Step 9790 | mode=video loss=0.094238 diff=0.094238 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0467 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0868 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=7833 i_micro=1957 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3633 mask_tok=109.00 y_norm=48.0868±0.0062 +2026-05-10 17:59:22,413 - INFO - Step 9800 | mode=video loss=0.093262 diff=0.093262 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0752 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0912 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=7841 i_micro=1959 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.8733 mask_tok=262.00 y_norm=48.0912±0.0079 cond_uncond_dloss=0.289062 cond_grad=0.000064 cond_pred_l2=38.224873 cond_pred_ratio=0.022754 +2026-05-10 17:59:22,451 - INFO - Step 9800 | probe_semantic mcp_offdiag(mean/min/max)=0.646547/0.533017/0.709109 smol_offdiag=0.660557 prompts=4 mcp_tok=17.50 smol_tok=17.50 +2026-05-10 17:59:29,561 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 17:59:48,253 - INFO - Step 9810 | mode=video loss=0.055420 diff=0.055420 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0492 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0881 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=7849 i_micro=1961 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4300 mask_tok=129.00 y_norm=48.0881±0.0058 +2026-05-10 18:00:06,693 - INFO - Step 9820 | mode=video loss=0.087402 diff=0.087402 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0821 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0893 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=7857 i_micro=1963 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3533 mask_tok=106.00 y_norm=48.0893±0.0056 +2026-05-10 18:00:25,364 - INFO - Step 9830 | mode=video loss=0.090820 diff=0.090820 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0466 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0865 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=7865 i_micro=1965 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3633 mask_tok=109.00 y_norm=48.0865±0.0056 +2026-05-10 18:00:43,313 - INFO - Step 9840 | mode=video loss=0.177734 diff=0.177734 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0775 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0865 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=7873 i_micro=1967 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4367 mask_tok=131.00 y_norm=48.0865±0.0059 +2026-05-10 18:01:01,683 - INFO - Step 9850 | mode=video loss=0.095703 diff=0.095703 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0250 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0874 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=7881 i_micro=1969 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4000 mask_tok=120.00 y_norm=48.0874±0.0063 +2026-05-10 18:01:20,370 - INFO - Step 9860 | mode=video loss=0.089844 diff=0.089844 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0976 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0884 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=7889 i_micro=1971 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4267 mask_tok=128.00 y_norm=48.0884±0.0065 +2026-05-10 18:01:38,734 - INFO - Step 9870 | mode=video loss=0.133789 diff=0.133789 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0734 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0864 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=7897 i_micro=1973 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4600 mask_tok=138.00 y_norm=48.0864±0.0060 +2026-05-10 18:01:56,690 - INFO - Step 9880 | mode=video loss=0.097656 diff=0.097656 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0898 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0873 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=7905 i_micro=1975 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.5167 mask_tok=155.00 y_norm=48.0873±0.0064 +2026-05-10 18:02:15,471 - INFO - Step 9890 | mode=video loss=0.087402 diff=0.087402 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0433 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0884 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=7913 i_micro=1977 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3467 mask_tok=104.00 y_norm=48.0884±0.0061 +2026-05-10 18:02:33,846 - INFO - Step 9900 | mode=video loss=0.056152 diff=0.056152 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0187 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0881 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=7921 i_micro=1979 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3467 mask_tok=104.00 y_norm=48.0881±0.0067 +2026-05-10 18:02:40,671 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 18:02:59,460 - INFO - Step 9910 | mode=video loss=0.071289 diff=0.071289 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0300 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0877 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=7929 i_micro=1981 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3267 mask_tok=98.00 y_norm=48.0877±0.0055 +2026-05-10 18:03:17,448 - INFO - Step 9920 | mode=video loss=0.087402 diff=0.087402 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0231 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0888 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=7937 i_micro=1983 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.2333 mask_tok=70.00 y_norm=48.0888±0.0051 +2026-05-10 18:03:35,913 - INFO - Step 9930 | mode=video loss=0.104004 diff=0.104004 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0545 emb_mean=0.0007 emb_std=1.0019 tok_norm=48.0909 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=7945 i_micro=1985 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3767 mask_tok=113.00 y_norm=48.0909±0.0064 +2026-05-10 18:03:54,574 - INFO - Step 9940 | mode=video loss=0.077637 diff=0.077637 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0419 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0868 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=7953 i_micro=1987 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4933 mask_tok=148.00 y_norm=48.0868±0.0062 +2026-05-10 18:04:13,018 - INFO - Step 9950 | mode=video loss=0.091797 diff=0.091797 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0425 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0884 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=7961 i_micro=1989 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3267 mask_tok=98.00 y_norm=48.0884±0.0051 +2026-05-10 18:04:31,743 - INFO - Step 9960 | mode=video loss=0.058350 diff=0.058350 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0364 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0860 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=7969 i_micro=1991 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4267 mask_tok=128.00 y_norm=48.0860±0.0057 +2026-05-10 18:04:49,702 - INFO - Step 9970 | mode=video loss=0.067871 diff=0.067871 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0307 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0876 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=7977 i_micro=1993 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3800 mask_tok=114.00 y_norm=48.0876±0.0061 +2026-05-10 18:05:08,122 - INFO - Step 9980 | mode=video loss=0.083984 diff=0.083984 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1922 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0888 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=7985 i_micro=1995 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3733 mask_tok=112.00 y_norm=48.0888±0.0054 +2026-05-10 18:05:26,869 - INFO - Step 9990 | mode=video loss=0.110352 diff=0.110352 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0354 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0873 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=7993 i_micro=1997 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3967 mask_tok=119.00 y_norm=48.0873±0.0066 +2026-05-10 18:05:46,896 - INFO - Step 10000 | mode=video loss=0.178711 diff=0.178711 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0400 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0864 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=8001 i_micro=1999 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4333 mask_tok=130.00 y_norm=48.0864±0.0068 cond_uncond_dloss=0.200195 cond_grad=0.000070 cond_pred_l2=25.621384 cond_pred_ratio=0.015687 +2026-05-10 18:05:46,933 - INFO - Step 10000 | probe_semantic mcp_offdiag(mean/min/max)=0.645962/0.532198/0.708327 smol_offdiag=0.660557 prompts=4 mcp_tok=17.50 smol_tok=17.50 +2026-05-10 18:05:53,397 - INFO - Saved checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_step10000.pt +2026-05-10 18:05:58,192 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 18:06:16,150 - INFO - Step 10010 | mode=video loss=0.118652 diff=0.118652 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0360 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0869 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=8009 i_micro=2001 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3400 mask_tok=102.00 y_norm=48.0869±0.0057 +2026-05-10 18:06:34,920 - INFO - Step 10020 | mode=video loss=0.094727 diff=0.094727 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0643 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0876 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=8017 i_micro=2003 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4367 mask_tok=131.00 y_norm=48.0876±0.0062 +2026-05-10 18:06:53,358 - INFO - Step 10030 | mode=video loss=0.275391 diff=0.275391 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1774 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0869 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=8025 i_micro=2005 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.5100 mask_tok=153.00 y_norm=48.0869±0.0063 +2026-05-10 18:07:12,039 - INFO - Step 10040 | mode=video loss=0.075684 diff=0.075684 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0970 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0880 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=8033 i_micro=2007 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3667 mask_tok=110.00 y_norm=48.0880±0.0057 +2026-05-10 18:07:30,461 - INFO - Step 10050 | mode=video loss=0.116699 diff=0.116699 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0386 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0877 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=8041 i_micro=2009 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4533 mask_tok=136.00 y_norm=48.0877±0.0064 +2026-05-10 18:07:48,409 - INFO - Step 10060 | mode=video loss=0.101074 diff=0.101074 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0604 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0890 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=8049 i_micro=2011 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.5100 mask_tok=153.00 y_norm=48.0890±0.0071 +2026-05-10 18:08:07,086 - INFO - Step 10070 | mode=video loss=0.096680 diff=0.096680 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0333 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0889 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=8057 i_micro=2013 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.2533 mask_tok=76.00 y_norm=48.0889±0.0055 +2026-05-10 18:08:25,473 - INFO - Step 10080 | mode=video loss=0.016724 diff=0.016724 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1146 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0889 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=8065 i_micro=2015 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3567 mask_tok=107.00 y_norm=48.0889±0.0061 +2026-05-10 18:08:44,073 - INFO - Step 10090 | mode=video loss=0.099121 diff=0.099121 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0687 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0895 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=8073 i_micro=2017 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3433 mask_tok=103.00 y_norm=48.0895±0.0059 +2026-05-10 18:09:02,029 - INFO - Step 10100 | mode=video loss=0.074219 diff=0.074219 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0749 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0887 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=8081 i_micro=2019 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3867 mask_tok=116.00 y_norm=48.0887±0.0060 +2026-05-10 18:09:09,415 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 18:09:27,802 - INFO - Step 10110 | mode=video loss=0.083496 diff=0.083496 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0479 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0880 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=8089 i_micro=2021 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3333 mask_tok=100.00 y_norm=48.0880±0.0060 +2026-05-10 18:09:46,448 - INFO - Step 10120 | mode=video loss=0.132812 diff=0.132812 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0598 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0883 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=8097 i_micro=2023 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3367 mask_tok=101.00 y_norm=48.0883±0.0057 +2026-05-10 18:10:04,863 - INFO - Step 10130 | mode=video loss=0.105957 diff=0.105957 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0409 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0877 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=8105 i_micro=2025 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4367 mask_tok=131.00 y_norm=48.0877±0.0064 +2026-05-10 18:10:22,827 - INFO - Step 10140 | mode=video loss=0.086914 diff=0.086914 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0507 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0888 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=8113 i_micro=2027 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3300 mask_tok=99.00 y_norm=48.0888±0.0061 +2026-05-10 18:10:41,568 - INFO - Step 10150 | mode=video loss=0.097656 diff=0.097656 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0319 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0900 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=8121 i_micro=2029 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3567 mask_tok=107.00 y_norm=48.0900±0.0056 +2026-05-10 18:10:59,957 - INFO - Step 10160 | mode=video loss=0.127930 diff=0.127930 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0646 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0871 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=8129 i_micro=2031 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3500 mask_tok=105.00 y_norm=48.0871±0.0060 +2026-05-10 18:11:18,642 - INFO - Step 10170 | mode=video loss=0.118164 diff=0.118164 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0327 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0858 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=8137 i_micro=2033 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3533 mask_tok=106.00 y_norm=48.0858±0.0058 +2026-05-10 18:11:36,623 - INFO - Step 10180 | mode=video loss=0.078125 diff=0.078125 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0361 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0878 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=8145 i_micro=2035 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3467 mask_tok=104.00 y_norm=48.0878±0.0053 +2026-05-10 18:11:55,023 - INFO - Step 10190 | mode=video loss=0.116699 diff=0.116699 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0517 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0883 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=8153 i_micro=2037 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3500 mask_tok=105.00 y_norm=48.0883±0.0056 +2026-05-10 18:12:15,307 - INFO - Step 10200 | mode=video loss=0.119141 diff=0.119141 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0569 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0879 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=8161 i_micro=2039 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.2700 mask_tok=81.00 y_norm=48.0879±0.0056 cond_uncond_dloss=0.478516 cond_grad=0.000405 cond_pred_l2=33.226326 cond_pred_ratio=0.020940 +2026-05-10 18:12:15,345 - INFO - Step 10200 | probe_semantic mcp_offdiag(mean/min/max)=0.646296/0.532691/0.708542 smol_offdiag=0.660557 prompts=4 mcp_tok=17.50 smol_tok=17.50 +2026-05-10 18:12:22,206 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 18:12:40,574 - INFO - Step 10210 | mode=video loss=0.065430 diff=0.065430 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0312 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0887 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=8169 i_micro=2041 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3333 mask_tok=100.00 y_norm=48.0887±0.0054 +2026-05-10 18:12:59,352 - INFO - Step 10220 | mode=video loss=0.083496 diff=0.083496 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0620 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0856 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=8177 i_micro=2043 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4233 mask_tok=127.00 y_norm=48.0856±0.0065 +2026-05-10 18:13:17,318 - INFO - Step 10230 | mode=video loss=0.104980 diff=0.104980 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0955 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0878 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=8185 i_micro=2045 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3767 mask_tok=113.00 y_norm=48.0878±0.0070 +2026-05-10 18:13:35,778 - INFO - Step 10240 | mode=video loss=0.184570 diff=0.184570 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0961 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0882 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=8193 i_micro=2047 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3433 mask_tok=103.00 y_norm=48.0882±0.0054 +2026-05-10 18:13:54,505 - INFO - Step 10250 | mode=video loss=0.071777 diff=0.071777 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0518 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0867 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=8201 i_micro=2049 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3467 mask_tok=104.00 y_norm=48.0867±0.0059 +2026-05-10 18:14:12,896 - INFO - Step 10260 | mode=video loss=0.051758 diff=0.051758 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0477 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0872 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=8209 i_micro=2051 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4200 mask_tok=126.00 y_norm=48.0872±0.0059 +2026-05-10 18:14:30,873 - INFO - Step 10270 | mode=video loss=0.059814 diff=0.059814 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0256 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0875 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=8217 i_micro=2053 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4133 mask_tok=124.00 y_norm=48.0875±0.0062 +2026-05-10 18:14:49,572 - INFO - Step 10280 | mode=video loss=0.085449 diff=0.085449 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0551 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0875 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=8225 i_micro=2055 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3200 mask_tok=96.00 y_norm=48.0875±0.0055 +2026-05-10 18:15:07,954 - INFO - Step 10290 | mode=video loss=0.060791 diff=0.060791 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0416 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0879 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=8233 i_micro=2057 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3167 mask_tok=95.00 y_norm=48.0879±0.0060 +2026-05-10 18:15:26,444 - INFO - Step 10300 | mode=video loss=0.073242 diff=0.073242 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0476 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0855 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=8241 i_micro=2059 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3033 mask_tok=91.00 y_norm=48.0855±0.0059 +2026-05-10 18:15:33,389 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 18:15:51,733 - INFO - Step 10310 | mode=video loss=0.126953 diff=0.126953 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0495 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0887 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=8249 i_micro=2061 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3467 mask_tok=104.00 y_norm=48.0887±0.0059 +2026-05-10 18:16:09,702 - INFO - Step 10320 | mode=video loss=0.145508 diff=0.145508 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0727 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0905 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=8257 i_micro=2063 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3433 mask_tok=103.00 y_norm=48.0905±0.0060 +2026-05-10 18:16:28,369 - INFO - Step 10330 | mode=video loss=0.089355 diff=0.089355 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0674 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0868 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=8265 i_micro=2065 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4867 mask_tok=146.00 y_norm=48.0868±0.0062 +2026-05-10 18:16:46,777 - INFO - Step 10340 | mode=video loss=0.070801 diff=0.070801 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0221 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0872 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=8273 i_micro=2067 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4400 mask_tok=132.00 y_norm=48.0872±0.0062 +2026-05-10 18:17:05,279 - INFO - Step 10350 | mode=video loss=0.108398 diff=0.108398 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0503 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0876 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=8281 i_micro=2069 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3533 mask_tok=106.00 y_norm=48.0876±0.0064 +2026-05-10 18:17:23,242 - INFO - Step 10360 | mode=video loss=0.060059 diff=0.060059 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0377 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0898 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=8289 i_micro=2071 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3533 mask_tok=106.00 y_norm=48.0898±0.0051 +2026-05-10 18:17:41,610 - INFO - Step 10370 | mode=video loss=0.102539 diff=0.102539 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0483 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0876 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=8297 i_micro=2073 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3500 mask_tok=105.00 y_norm=48.0876±0.0052 +2026-05-10 18:18:00,367 - INFO - Step 10380 | mode=video loss=0.142578 diff=0.142578 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0566 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0878 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=8305 i_micro=2075 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3700 mask_tok=111.00 y_norm=48.0878±0.0057 +2026-05-10 18:18:18,748 - INFO - Step 10390 | mode=video loss=0.112305 diff=0.112305 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0385 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0886 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=8313 i_micro=2077 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3533 mask_tok=106.00 y_norm=48.0886±0.0058 +2026-05-10 18:18:38,344 - INFO - Step 10400 | mode=video loss=0.090820 diff=0.090820 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0390 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0884 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=8321 i_micro=2079 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3533 mask_tok=106.00 y_norm=48.0884±0.0057 cond_uncond_dloss=0.589844 cond_grad=0.000146 cond_pred_l2=52.215752 cond_pred_ratio=0.030691 +2026-05-10 18:18:38,382 - INFO - Step 10400 | probe_semantic mcp_offdiag(mean/min/max)=0.646410/0.533215/0.708723 smol_offdiag=0.660557 prompts=4 mcp_tok=17.50 smol_tok=17.50 +2026-05-10 18:18:45,418 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 18:19:04,173 - INFO - Step 10410 | mode=video loss=0.094238 diff=0.094238 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0355 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0870 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=8329 i_micro=2081 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3567 mask_tok=107.00 y_norm=48.0870±0.0059 +2026-05-10 18:19:22,559 - INFO - Step 10420 | mode=video loss=0.081055 diff=0.081055 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0415 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0881 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=8337 i_micro=2083 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3267 mask_tok=98.00 y_norm=48.0881±0.0058 +2026-05-10 18:19:41,277 - INFO - Step 10430 | mode=video loss=0.076172 diff=0.076172 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0424 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0888 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=8345 i_micro=2085 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3300 mask_tok=99.00 y_norm=48.0888±0.0054 +2026-05-10 18:19:59,257 - INFO - Step 10440 | mode=video loss=0.141602 diff=0.141602 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0368 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0866 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=8353 i_micro=2087 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3567 mask_tok=107.00 y_norm=48.0866±0.0062 +2026-05-10 18:20:17,665 - INFO - Step 10450 | mode=video loss=0.091797 diff=0.091797 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0430 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0880 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=8361 i_micro=2089 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3333 mask_tok=100.00 y_norm=48.0880±0.0055 +2026-05-10 18:20:36,307 - INFO - Step 10460 | mode=video loss=0.066406 diff=0.066406 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0490 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0879 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=8369 i_micro=2091 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3467 mask_tok=104.00 y_norm=48.0879±0.0053 +2026-05-10 18:20:54,781 - INFO - Step 10470 | mode=video loss=0.080078 diff=0.080078 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0459 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0873 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=8377 i_micro=2093 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4000 mask_tok=120.00 y_norm=48.0873±0.0065 +2026-05-10 18:21:13,451 - INFO - Step 10480 | mode=video loss=0.093262 diff=0.093262 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0677 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0872 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=8385 i_micro=2095 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4533 mask_tok=136.00 y_norm=48.0872±0.0064 +2026-05-10 18:21:31,426 - INFO - Step 10490 | mode=video loss=0.086426 diff=0.086426 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0450 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0866 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=8393 i_micro=2097 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3633 mask_tok=109.00 y_norm=48.0866±0.0059 +2026-05-10 18:21:49,804 - INFO - Step 10500 | mode=video loss=0.141602 diff=0.141602 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0323 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0893 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=8401 i_micro=2099 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3300 mask_tok=99.00 y_norm=48.0893±0.0053 +2026-05-10 18:21:56,732 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 18:22:15,402 - INFO - Step 10510 | mode=video loss=0.115723 diff=0.115723 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0373 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0870 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=8409 i_micro=2101 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4633 mask_tok=139.00 y_norm=48.0870±0.0065 +2026-05-10 18:22:33,801 - INFO - Step 10520 | mode=video loss=0.117188 diff=0.117188 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0374 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0871 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=8417 i_micro=2103 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3800 mask_tok=114.00 y_norm=48.0871±0.0060 +2026-05-10 18:22:51,773 - INFO - Step 10530 | mode=video loss=0.134766 diff=0.134766 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0431 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0871 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=8425 i_micro=2105 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4133 mask_tok=124.00 y_norm=48.0871±0.0057 +2026-05-10 18:23:10,483 - INFO - Step 10540 | mode=video loss=0.064941 diff=0.064941 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0701 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0858 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=8433 i_micro=2107 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3933 mask_tok=118.00 y_norm=48.0858±0.0054 +2026-05-10 18:23:28,857 - INFO - Step 10550 | mode=video loss=0.138672 diff=0.138672 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0382 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0865 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=8441 i_micro=2109 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3700 mask_tok=111.00 y_norm=48.0865±0.0056 +2026-05-10 18:23:47,611 - INFO - Step 10560 | mode=video loss=0.197266 diff=0.197266 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1050 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0869 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=8449 i_micro=2111 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3867 mask_tok=116.00 y_norm=48.0869±0.0059 +2026-05-10 18:24:06,001 - INFO - Step 10570 | mode=video loss=0.080566 diff=0.080566 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0543 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0886 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=8457 i_micro=2113 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3733 mask_tok=112.00 y_norm=48.0886±0.0059 +2026-05-10 18:24:23,945 - INFO - Step 10580 | mode=video loss=0.083984 diff=0.083984 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0298 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0876 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=8465 i_micro=2115 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3800 mask_tok=114.00 y_norm=48.0876±0.0060 +2026-05-10 18:24:42,488 - INFO - Step 10590 | mode=video loss=0.152344 diff=0.152344 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0502 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0884 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=8473 i_micro=2117 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.2733 mask_tok=82.00 y_norm=48.0884±0.0054 +2026-05-10 18:25:02,575 - INFO - Step 10600 | mode=video loss=0.075684 diff=0.075684 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0390 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0884 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=8481 i_micro=2119 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3300 mask_tok=99.00 y_norm=48.0884±0.0056 cond_uncond_dloss=0.283203 cond_grad=0.000074 cond_pred_l2=44.477871 cond_pred_ratio=0.024689 +2026-05-10 18:25:02,613 - INFO - Step 10600 | probe_semantic mcp_offdiag(mean/min/max)=0.646351/0.533180/0.708503 smol_offdiag=0.660557 prompts=4 mcp_tok=17.50 smol_tok=17.50 +2026-05-10 18:25:09,398 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 18:25:28,147 - INFO - Step 10610 | mode=video loss=0.104980 diff=0.104980 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0433 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0902 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=8489 i_micro=2121 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3200 mask_tok=96.00 y_norm=48.0902±0.0057 +2026-05-10 18:25:46,112 - INFO - Step 10620 | mode=video loss=0.070312 diff=0.070312 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0577 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0886 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=8497 i_micro=2123 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3533 mask_tok=106.00 y_norm=48.0886±0.0060 +2026-05-10 18:26:04,561 - INFO - Step 10630 | mode=video loss=0.091309 diff=0.091309 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0655 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0922 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=8505 i_micro=2125 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.5500 mask_tok=165.00 y_norm=48.0922±0.0098 +2026-05-10 18:26:23,228 - INFO - Step 10640 | mode=video loss=0.097656 diff=0.097656 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1206 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0873 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=8513 i_micro=2127 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4000 mask_tok=120.00 y_norm=48.0873±0.0057 +2026-05-10 18:26:41,628 - INFO - Step 10650 | mode=video loss=0.066406 diff=0.066406 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0394 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0870 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=8521 i_micro=2129 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4400 mask_tok=132.00 y_norm=48.0870±0.0055 +2026-05-10 18:26:59,598 - INFO - Step 10660 | mode=video loss=0.083008 diff=0.083008 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0400 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0886 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=8529 i_micro=2131 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3067 mask_tok=92.00 y_norm=48.0886±0.0051 +2026-05-10 18:27:18,315 - INFO - Step 10670 | mode=video loss=0.050781 diff=0.050781 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0292 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0882 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=8537 i_micro=2133 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4767 mask_tok=143.00 y_norm=48.0882±0.0067 +2026-05-10 18:27:36,719 - INFO - Step 10680 | mode=video loss=0.122070 diff=0.122070 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0623 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0860 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=8545 i_micro=2135 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4667 mask_tok=140.00 y_norm=48.0860±0.0057 +2026-05-10 18:27:55,442 - INFO - Step 10690 | mode=video loss=0.061279 diff=0.061279 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0681 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0860 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=8553 i_micro=2137 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4433 mask_tok=133.00 y_norm=48.0860±0.0059 +2026-05-10 18:28:13,811 - INFO - Step 10700 | mode=video loss=0.121094 diff=0.121094 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0399 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0870 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=8561 i_micro=2139 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3633 mask_tok=109.00 y_norm=48.0870±0.0054 +2026-05-10 18:28:20,792 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 18:28:38,751 - INFO - Step 10710 | mode=video loss=0.069336 diff=0.069336 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1085 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0863 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=8569 i_micro=2141 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4733 mask_tok=142.00 y_norm=48.0863±0.0061 +2026-05-10 18:28:57,513 - INFO - Step 10720 | mode=video loss=0.086914 diff=0.086914 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0308 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0889 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=8577 i_micro=2143 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3333 mask_tok=100.00 y_norm=48.0889±0.0052 +2026-05-10 18:29:15,909 - INFO - Step 10730 | mode=video loss=0.148438 diff=0.148438 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0306 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0867 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=8585 i_micro=2145 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3867 mask_tok=116.00 y_norm=48.0867±0.0054 +2026-05-10 18:29:34,533 - INFO - Step 10740 | mode=video loss=0.123535 diff=0.123535 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0975 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0879 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=8593 i_micro=2147 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3400 mask_tok=102.00 y_norm=48.0879±0.0059 +2026-05-10 18:29:52,484 - INFO - Step 10750 | mode=video loss=0.173828 diff=0.173828 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0446 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0880 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=8601 i_micro=2149 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3400 mask_tok=102.00 y_norm=48.0880±0.0051 +2026-05-10 18:30:10,900 - INFO - Step 10760 | mode=video loss=0.073730 diff=0.073730 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0635 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0865 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=8609 i_micro=2151 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3933 mask_tok=118.00 y_norm=48.0865±0.0064 +2026-05-10 18:30:29,555 - INFO - Step 10770 | mode=video loss=0.087402 diff=0.087402 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0462 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0857 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=8617 i_micro=2153 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3833 mask_tok=115.00 y_norm=48.0857±0.0072 +2026-05-10 18:30:47,941 - INFO - Step 10780 | mode=video loss=0.101562 diff=0.101562 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0862 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0885 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=8625 i_micro=2155 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3233 mask_tok=97.00 y_norm=48.0885±0.0056 +2026-05-10 18:31:05,916 - INFO - Step 10790 | mode=video loss=0.092773 diff=0.092773 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0602 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0886 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=8633 i_micro=2157 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3167 mask_tok=95.00 y_norm=48.0886±0.0058 +2026-05-10 18:31:26,212 - INFO - Step 10800 | mode=video loss=0.086914 diff=0.086914 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0673 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0901 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=8641 i_micro=2159 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3400 mask_tok=102.00 y_norm=48.0901±0.0057 cond_uncond_dloss=0.335938 cond_grad=0.000094 cond_pred_l2=41.476624 cond_pred_ratio=0.025302 +2026-05-10 18:31:26,249 - INFO - Step 10800 | probe_semantic mcp_offdiag(mean/min/max)=0.646135/0.532720/0.708378 smol_offdiag=0.660557 prompts=4 mcp_tok=17.50 smol_tok=17.50 +2026-05-10 18:31:33,723 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 18:31:52,106 - INFO - Step 10810 | mode=video loss=0.101074 diff=0.101074 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0561 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0882 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=8649 i_micro=2161 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3400 mask_tok=102.00 y_norm=48.0882±0.0059 +2026-05-10 18:32:10,901 - INFO - Step 10820 | mode=video loss=0.082520 diff=0.082520 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0327 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0864 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=8657 i_micro=2163 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.5767 mask_tok=173.00 y_norm=48.0864±0.0066 +2026-05-10 18:32:29,303 - INFO - Step 10830 | mode=video loss=0.177734 diff=0.177734 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1029 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0902 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=8665 i_micro=2165 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3467 mask_tok=104.00 y_norm=48.0902±0.0058 +2026-05-10 18:32:47,263 - INFO - Step 10840 | mode=video loss=0.152344 diff=0.152344 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0656 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0900 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=8673 i_micro=2167 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.2667 mask_tok=80.00 y_norm=48.0900±0.0051 +2026-05-10 18:33:05,989 - INFO - Step 10850 | mode=video loss=0.081543 diff=0.081543 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0821 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0894 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=8681 i_micro=2169 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3367 mask_tok=101.00 y_norm=48.0894±0.0058 +2026-05-10 18:33:24,397 - INFO - Step 10860 | mode=video loss=0.083008 diff=0.083008 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0343 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0875 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=8689 i_micro=2171 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4833 mask_tok=145.00 y_norm=48.0875±0.0059 +2026-05-10 18:33:43,077 - INFO - Step 10870 | mode=video loss=0.107910 diff=0.107910 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1465 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0869 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=8697 i_micro=2173 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.5033 mask_tok=151.00 y_norm=48.0869±0.0065 +2026-05-10 18:34:01,470 - INFO - Step 10880 | mode=video loss=0.102539 diff=0.102539 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0366 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0867 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=8705 i_micro=2175 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3800 mask_tok=114.00 y_norm=48.0867±0.0059 +2026-05-10 18:34:19,416 - INFO - Step 10890 | mode=video loss=0.141602 diff=0.141602 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0463 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0871 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=8713 i_micro=2177 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4200 mask_tok=126.00 y_norm=48.0871±0.0061 +2026-05-10 18:34:38,047 - INFO - Step 10900 | mode=video loss=0.121582 diff=0.121582 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1041 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0866 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=8721 i_micro=2179 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3433 mask_tok=103.00 y_norm=48.0866±0.0059 +2026-05-10 18:34:44,808 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 18:35:03,172 - INFO - Step 10910 | mode=video loss=0.096191 diff=0.096191 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0540 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0887 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=8729 i_micro=2181 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3700 mask_tok=111.00 y_norm=48.0887±0.0058 +2026-05-10 18:35:21,889 - INFO - Step 10920 | mode=video loss=0.108398 diff=0.108398 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0368 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0871 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=8737 i_micro=2183 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.2967 mask_tok=89.00 y_norm=48.0871±0.0051 +2026-05-10 18:35:39,848 - INFO - Step 10930 | mode=video loss=0.062500 diff=0.062500 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0922 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0857 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=8745 i_micro=2185 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3767 mask_tok=113.00 y_norm=48.0857±0.0056 +2026-05-10 18:35:58,208 - INFO - Step 10940 | mode=video loss=0.188477 diff=0.188477 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1117 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0870 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=8753 i_micro=2187 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4300 mask_tok=129.00 y_norm=48.0870±0.0058 +2026-05-10 18:36:16,866 - INFO - Step 10950 | mode=video loss=0.065918 diff=0.065918 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0442 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0890 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=8761 i_micro=2189 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3233 mask_tok=97.00 y_norm=48.0890±0.0048 +2026-05-10 18:36:35,221 - INFO - Step 10960 | mode=video loss=0.092773 diff=0.092773 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0280 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0888 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=8769 i_micro=2191 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3433 mask_tok=103.00 y_norm=48.0888±0.0063 +2026-05-10 18:36:53,174 - INFO - Step 10970 | mode=video loss=0.153320 diff=0.153320 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0853 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0876 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=8777 i_micro=2193 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3500 mask_tok=105.00 y_norm=48.0876±0.0047 +2026-05-10 18:37:11,668 - INFO - Step 10980 | mode=video loss=0.104980 diff=0.104980 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0301 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0866 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=8785 i_micro=2195 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4833 mask_tok=145.00 y_norm=48.0866±0.0070 +2026-05-10 18:37:30,077 - INFO - Step 10990 | mode=video loss=0.050537 diff=0.050537 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0534 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0877 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=8793 i_micro=2197 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3200 mask_tok=96.00 y_norm=48.0877±0.0052 +2026-05-10 18:37:50,230 - INFO - Step 11000 | mode=video loss=0.091797 diff=0.091797 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0270 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0885 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=8801 i_micro=2199 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4833 mask_tok=145.00 y_norm=48.0885±0.0070 cond_uncond_dloss=0.261719 cond_grad=0.000083 cond_pred_l2=39.563377 cond_pred_ratio=0.023284 +2026-05-10 18:37:50,268 - INFO - Step 11000 | probe_semantic mcp_offdiag(mean/min/max)=0.645236/0.531502/0.707831 smol_offdiag=0.660557 prompts=4 mcp_tok=17.50 smol_tok=17.50 +2026-05-10 18:37:57,080 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 18:38:15,517 - INFO - Step 11010 | mode=video loss=0.088379 diff=0.088379 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0709 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0869 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=8809 i_micro=2201 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4400 mask_tok=132.00 y_norm=48.0869±0.0066 +2026-05-10 18:38:33,472 - INFO - Step 11020 | mode=video loss=0.107422 diff=0.107422 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1325 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0879 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=8817 i_micro=2203 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4400 mask_tok=132.00 y_norm=48.0879±0.0059 +2026-05-10 18:38:52,250 - INFO - Step 11030 | mode=video loss=0.126953 diff=0.126953 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0232 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0870 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=8825 i_micro=2205 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4600 mask_tok=138.00 y_norm=48.0870±0.0067 +2026-05-10 18:39:10,642 - INFO - Step 11040 | mode=video loss=0.124512 diff=0.124512 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0883 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0892 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=8833 i_micro=2207 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3500 mask_tok=105.00 y_norm=48.0892±0.0063 +2026-05-10 18:39:29,355 - INFO - Step 11050 | mode=video loss=0.073730 diff=0.073730 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0321 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0865 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=8841 i_micro=2209 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3500 mask_tok=105.00 y_norm=48.0865±0.0057 +2026-05-10 18:39:47,317 - INFO - Step 11060 | mode=video loss=0.091797 diff=0.091797 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0414 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0880 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=8849 i_micro=2211 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3033 mask_tok=91.00 y_norm=48.0880±0.0051 +2026-05-10 18:40:05,752 - INFO - Step 11070 | mode=video loss=0.207031 diff=0.207031 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1624 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0888 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=8857 i_micro=2213 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3533 mask_tok=106.00 y_norm=48.0888±0.0061 +2026-05-10 18:40:24,477 - INFO - Step 11080 | mode=video loss=0.078613 diff=0.078613 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0328 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0869 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=8865 i_micro=2215 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3000 mask_tok=90.00 y_norm=48.0869±0.0059 +2026-05-10 18:40:42,856 - INFO - Step 11090 | mode=video loss=0.080078 diff=0.080078 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0557 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0888 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=8873 i_micro=2217 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.5067 mask_tok=152.00 y_norm=48.0888±0.0060 +2026-05-10 18:41:00,817 - INFO - Step 11100 | mode=video loss=0.075684 diff=0.075684 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0382 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0884 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=8881 i_micro=2219 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3200 mask_tok=96.00 y_norm=48.0884±0.0051 +2026-05-10 18:41:07,448 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 18:41:25,394 - INFO - Step 11110 | mode=video loss=0.148438 diff=0.148438 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0819 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0872 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=8889 i_micro=2221 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3533 mask_tok=106.00 y_norm=48.0872±0.0058 +2026-05-10 18:41:43,766 - INFO - Step 11120 | mode=video loss=0.172852 diff=0.172852 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1847 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0869 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=8897 i_micro=2223 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4333 mask_tok=130.00 y_norm=48.0869±0.0063 +2026-05-10 18:42:02,335 - INFO - Step 11130 | mode=video loss=0.104980 diff=0.104980 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0480 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0888 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=8905 i_micro=2225 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3333 mask_tok=100.00 y_norm=48.0888±0.0064 +2026-05-10 18:42:20,726 - INFO - Step 11140 | mode=video loss=0.060547 diff=0.060547 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0332 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0869 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=8913 i_micro=2227 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3833 mask_tok=115.00 y_norm=48.0869±0.0056 +2026-05-10 18:42:38,674 - INFO - Step 11150 | mode=video loss=0.120117 diff=0.120117 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0419 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0880 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=8921 i_micro=2229 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3533 mask_tok=106.00 y_norm=48.0880±0.0059 +2026-05-10 18:42:57,415 - INFO - Step 11160 | mode=video loss=0.077637 diff=0.077637 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0405 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0864 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=8929 i_micro=2231 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3500 mask_tok=105.00 y_norm=48.0864±0.0055 +2026-05-10 18:43:15,836 - INFO - Step 11170 | mode=video loss=0.115234 diff=0.115234 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0610 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0888 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=8937 i_micro=2233 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3533 mask_tok=106.00 y_norm=48.0888±0.0056 +2026-05-10 18:43:34,496 - INFO - Step 11180 | mode=video loss=0.090820 diff=0.090820 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0472 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0884 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=8945 i_micro=2235 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.5033 mask_tok=151.00 y_norm=48.0884±0.0068 +2026-05-10 18:43:52,454 - INFO - Step 11190 | mode=video loss=0.112305 diff=0.112305 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0450 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0864 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=8953 i_micro=2237 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4333 mask_tok=130.00 y_norm=48.0864±0.0060 +2026-05-10 18:44:12,492 - INFO - Step 11200 | mode=video loss=0.111816 diff=0.111816 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0395 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0875 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=8961 i_micro=2239 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3500 mask_tok=105.00 y_norm=48.0875±0.0054 cond_uncond_dloss=0.251953 cond_grad=0.000040 cond_pred_l2=32.774952 cond_pred_ratio=0.018627 +2026-05-10 18:44:12,529 - INFO - Step 11200 | probe_semantic mcp_offdiag(mean/min/max)=0.645117/0.531442/0.707510 smol_offdiag=0.660557 prompts=4 mcp_tok=17.50 smol_tok=17.50 +2026-05-10 18:44:19,324 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 18:44:37,965 - INFO - Step 11210 | mode=video loss=0.088867 diff=0.088867 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0983 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0878 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=8969 i_micro=2241 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4167 mask_tok=125.00 y_norm=48.0878±0.0067 +2026-05-10 18:44:56,404 - INFO - Step 11220 | mode=video loss=0.073730 diff=0.073730 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0593 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0886 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=8977 i_micro=2243 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3333 mask_tok=100.00 y_norm=48.0886±0.0057 +2026-05-10 18:45:15,057 - INFO - Step 11230 | mode=video loss=0.119141 diff=0.119141 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0461 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0887 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=8985 i_micro=2245 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3500 mask_tok=105.00 y_norm=48.0887±0.0061 +2026-05-10 18:45:33,030 - INFO - Step 11240 | mode=video loss=0.085938 diff=0.085938 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0503 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0864 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=8993 i_micro=2247 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4767 mask_tok=143.00 y_norm=48.0864±0.0059 +2026-05-10 18:45:51,430 - INFO - Step 11250 | mode=video loss=0.104004 diff=0.104004 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0498 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0863 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=9001 i_micro=2249 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3733 mask_tok=112.00 y_norm=48.0863±0.0060 +2026-05-10 18:46:10,129 - INFO - Step 11260 | mode=video loss=0.073242 diff=0.073242 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0923 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0899 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=9009 i_micro=2251 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3733 mask_tok=112.00 y_norm=48.0899±0.0054 +2026-05-10 18:46:28,486 - INFO - Step 11270 | mode=video loss=0.149414 diff=0.149414 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1122 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0866 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=9017 i_micro=2253 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4200 mask_tok=126.00 y_norm=48.0866±0.0055 +2026-05-10 18:46:46,462 - INFO - Step 11280 | mode=video loss=0.066406 diff=0.066406 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0962 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0880 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=9025 i_micro=2255 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3833 mask_tok=115.00 y_norm=48.0880±0.0053 +2026-05-10 18:47:04,989 - INFO - Step 11290 | mode=video loss=0.166016 diff=0.166016 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0866 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0887 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=9033 i_micro=2257 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3500 mask_tok=105.00 y_norm=48.0887±0.0052 +2026-05-10 18:47:23,399 - INFO - Step 11300 | mode=video loss=0.153320 diff=0.153320 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1748 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0857 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=9041 i_micro=2259 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4000 mask_tok=120.00 y_norm=48.0857±0.0057 +2026-05-10 18:47:30,760 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 18:47:49,480 - INFO - Step 11310 | mode=video loss=0.043213 diff=0.043213 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1304 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0874 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=9049 i_micro=2261 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.2933 mask_tok=88.00 y_norm=48.0874±0.0047 +2026-05-10 18:48:07,439 - INFO - Step 11320 | mode=video loss=0.116211 diff=0.116211 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0474 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0876 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=9057 i_micro=2263 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4200 mask_tok=126.00 y_norm=48.0876±0.0064 +2026-05-10 18:48:25,909 - INFO - Step 11330 | mode=video loss=0.196289 diff=0.196289 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0647 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0877 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=9065 i_micro=2265 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3500 mask_tok=105.00 y_norm=48.0877±0.0055 +2026-05-10 18:48:44,610 - INFO - Step 11340 | mode=video loss=0.213867 diff=0.213867 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0474 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0871 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=9073 i_micro=2267 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3433 mask_tok=103.00 y_norm=48.0871±0.0051 +2026-05-10 18:49:03,021 - INFO - Step 11350 | mode=video loss=0.026978 diff=0.026978 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0716 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0884 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=9081 i_micro=2269 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3367 mask_tok=101.00 y_norm=48.0884±0.0051 +2026-05-10 18:49:21,653 - INFO - Step 11360 | mode=video loss=0.113770 diff=0.113770 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1017 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0868 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=9089 i_micro=2271 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3467 mask_tok=104.00 y_norm=48.0868±0.0059 +2026-05-10 18:49:39,600 - INFO - Step 11370 | mode=video loss=0.085449 diff=0.085449 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1507 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0860 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=9097 i_micro=2273 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4233 mask_tok=127.00 y_norm=48.0860±0.0061 +2026-05-10 18:49:57,991 - INFO - Step 11380 | mode=video loss=0.103516 diff=0.103516 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0458 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0889 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=9105 i_micro=2275 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3300 mask_tok=99.00 y_norm=48.0889±0.0064 +2026-05-10 18:50:16,518 - INFO - Step 11390 | mode=video loss=0.083496 diff=0.083496 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1059 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0888 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=9113 i_micro=2277 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3367 mask_tok=101.00 y_norm=48.0888±0.0056 +2026-05-10 18:50:36,530 - INFO - Step 11400 | mode=video loss=0.077637 diff=0.077637 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0530 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0879 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=9121 i_micro=2279 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3367 mask_tok=101.00 y_norm=48.0879±0.0052 cond_uncond_dloss=0.242188 cond_grad=0.000033 cond_pred_l2=33.683651 cond_pred_ratio=0.018981 +2026-05-10 18:50:36,567 - INFO - Step 11400 | probe_semantic mcp_offdiag(mean/min/max)=0.645069/0.531215/0.707557 smol_offdiag=0.660557 prompts=4 mcp_tok=17.50 smol_tok=17.50 +2026-05-10 18:50:43,599 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 18:51:02,287 - INFO - Step 11410 | mode=video loss=0.073730 diff=0.073730 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0447 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0887 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=9129 i_micro=2281 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3800 mask_tok=114.00 y_norm=48.0887±0.0062 +2026-05-10 18:51:20,244 - INFO - Step 11420 | mode=video loss=0.093262 diff=0.093262 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0377 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0879 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=9137 i_micro=2283 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.5100 mask_tok=153.00 y_norm=48.0879±0.0065 +2026-05-10 18:51:38,640 - INFO - Step 11430 | mode=video loss=0.129883 diff=0.129883 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0551 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0863 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=9145 i_micro=2285 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3333 mask_tok=100.00 y_norm=48.0863±0.0060 +2026-05-10 18:51:57,294 - INFO - Step 11440 | mode=video loss=0.136719 diff=0.136719 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0573 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0865 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=9153 i_micro=2287 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4867 mask_tok=146.00 y_norm=48.0865±0.0067 +2026-05-10 18:52:15,246 - INFO - Step 11450 | mode=video loss=0.095215 diff=0.095215 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0832 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0863 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=9161 i_micro=2289 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3467 mask_tok=104.00 y_norm=48.0863±0.0068 +2026-05-10 18:52:33,698 - INFO - Step 11460 | mode=video loss=0.104004 diff=0.104004 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0582 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0863 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=9169 i_micro=2291 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3633 mask_tok=109.00 y_norm=48.0863±0.0061 +2026-05-10 18:52:52,358 - INFO - Step 11470 | mode=video loss=0.125977 diff=0.125977 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0495 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0883 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=9177 i_micro=2293 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3733 mask_tok=112.00 y_norm=48.0883±0.0067 +2026-05-10 18:53:10,797 - INFO - Step 11480 | mode=video loss=0.068848 diff=0.068848 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0238 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0869 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=9185 i_micro=2295 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3267 mask_tok=98.00 y_norm=48.0869±0.0056 +2026-05-10 18:53:29,414 - INFO - Step 11490 | mode=video loss=0.094238 diff=0.094238 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0952 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0887 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=9193 i_micro=2297 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3533 mask_tok=106.00 y_norm=48.0887±0.0052 +2026-05-10 18:53:47,389 - INFO - Step 11500 | mode=video loss=0.128906 diff=0.128906 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0403 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0852 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=9201 i_micro=2299 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4867 mask_tok=146.00 y_norm=48.0852±0.0067 +2026-05-10 18:53:54,837 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 18:54:13,220 - INFO - Step 11510 | mode=video loss=0.113770 diff=0.113770 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0582 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0878 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=9209 i_micro=2301 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3400 mask_tok=102.00 y_norm=48.0878±0.0061 +2026-05-10 18:54:31,973 - INFO - Step 11520 | mode=video loss=0.055908 diff=0.055908 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0626 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0875 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=9217 i_micro=2303 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4867 mask_tok=146.00 y_norm=48.0875±0.0068 +2026-05-10 18:54:50,397 - INFO - Step 11530 | mode=video loss=0.079590 diff=0.079590 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1952 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0885 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=9225 i_micro=2305 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3300 mask_tok=99.00 y_norm=48.0885±0.0051 +2026-05-10 18:55:09,049 - INFO - Step 11540 | mode=video loss=0.060059 diff=0.060059 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0236 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0852 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=9233 i_micro=2307 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4333 mask_tok=130.00 y_norm=48.0852±0.0061 +2026-05-10 18:55:27,013 - INFO - Step 11550 | mode=video loss=0.146484 diff=0.146484 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0842 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0888 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=9241 i_micro=2309 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.2700 mask_tok=81.00 y_norm=48.0888±0.0057 +2026-05-10 18:55:45,413 - INFO - Step 11560 | mode=video loss=0.054932 diff=0.054932 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0338 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0872 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=9249 i_micro=2311 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3300 mask_tok=99.00 y_norm=48.0872±0.0054 +2026-05-10 18:56:03,932 - INFO - Step 11570 | mode=video loss=0.083496 diff=0.083496 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0469 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0864 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=9257 i_micro=2313 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3600 mask_tok=108.00 y_norm=48.0864±0.0057 +2026-05-10 18:56:21,878 - INFO - Step 11580 | mode=video loss=0.067871 diff=0.067871 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0450 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0891 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=9265 i_micro=2315 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.2867 mask_tok=86.00 y_norm=48.0891±0.0056 +2026-05-10 18:56:40,728 - INFO - Step 11590 | mode=video loss=0.073242 diff=0.073242 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0773 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0879 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=9273 i_micro=2317 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4067 mask_tok=122.00 y_norm=48.0879±0.0055 +2026-05-10 18:57:00,301 - INFO - Step 11600 | mode=video loss=0.167969 diff=0.167969 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1528 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0871 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=9281 i_micro=2319 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4100 mask_tok=123.00 y_norm=48.0871±0.0064 cond_uncond_dloss=0.742188 cond_grad=0.001437 cond_pred_l2=38.537712 cond_pred_ratio=0.025168 +2026-05-10 18:57:00,338 - INFO - Step 11600 | probe_semantic mcp_offdiag(mean/min/max)=0.645341/0.531613/0.707891 smol_offdiag=0.660557 prompts=4 mcp_tok=17.50 smol_tok=17.50 +2026-05-10 18:57:07,250 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 18:57:25,661 - INFO - Step 11610 | mode=video loss=0.115234 diff=0.115234 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0359 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0889 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=9289 i_micro=2321 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.2867 mask_tok=86.00 y_norm=48.0889±0.0052 +2026-05-10 18:57:44,269 - INFO - Step 11620 | mode=video loss=0.062256 diff=0.062256 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0446 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0880 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=9297 i_micro=2323 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.8533 mask_tok=256.00 y_norm=48.0880±0.0076 +2026-05-10 18:58:02,234 - INFO - Step 11630 | mode=video loss=0.140625 diff=0.140625 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0430 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0864 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=9305 i_micro=2325 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4833 mask_tok=145.00 y_norm=48.0864±0.0069 +2026-05-10 18:58:20,608 - INFO - Step 11640 | mode=video loss=0.076172 diff=0.076172 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0273 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0871 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=9313 i_micro=2327 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3267 mask_tok=98.00 y_norm=48.0871±0.0056 +2026-05-10 18:58:39,071 - INFO - Step 11650 | mode=video loss=0.085449 diff=0.085449 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0452 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0871 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=9321 i_micro=2329 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3467 mask_tok=104.00 y_norm=48.0871±0.0059 +2026-05-10 18:58:57,440 - INFO - Step 11660 | mode=video loss=0.106934 diff=0.106934 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0261 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0894 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=9329 i_micro=2331 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3300 mask_tok=99.00 y_norm=48.0894±0.0054 +2026-05-10 18:59:16,091 - INFO - Step 11670 | mode=video loss=0.121094 diff=0.121094 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0586 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0886 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=9337 i_micro=2333 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3900 mask_tok=117.00 y_norm=48.0886±0.0065 +2026-05-10 18:59:34,047 - INFO - Step 11680 | mode=video loss=0.102051 diff=0.102051 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0759 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0869 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=9345 i_micro=2335 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3700 mask_tok=111.00 y_norm=48.0869±0.0067 +2026-05-10 18:59:52,418 - INFO - Step 11690 | mode=video loss=0.133789 diff=0.133789 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0523 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0868 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=9353 i_micro=2337 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3433 mask_tok=103.00 y_norm=48.0868±0.0060 +2026-05-10 19:00:11,055 - INFO - Step 11700 | mode=video loss=0.075684 diff=0.075684 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0332 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0881 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=9361 i_micro=2339 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3533 mask_tok=106.00 y_norm=48.0881±0.0058 +2026-05-10 19:00:17,814 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 19:00:35,781 - INFO - Step 11710 | mode=video loss=0.092285 diff=0.092285 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0432 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0844 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=9369 i_micro=2341 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4233 mask_tok=127.00 y_norm=48.0844±0.0053 +2026-05-10 19:00:55,039 - INFO - Step 11720 | mode=video loss=0.134766 diff=0.134766 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0449 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0862 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=9377 i_micro=2343 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4767 mask_tok=143.00 y_norm=48.0862±0.0066 +2026-05-10 19:01:12,999 - INFO - Step 11730 | mode=video loss=0.142578 diff=0.142578 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1435 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0887 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=9385 i_micro=2345 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.2733 mask_tok=82.00 y_norm=48.0887±0.0055 +2026-05-10 19:01:31,383 - INFO - Step 11740 | mode=video loss=0.050293 diff=0.050293 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0341 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0865 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=9393 i_micro=2347 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4800 mask_tok=144.00 y_norm=48.0865±0.0061 +2026-05-10 19:01:49,975 - INFO - Step 11750 | mode=video loss=0.179688 diff=0.179688 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0957 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0883 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=9401 i_micro=2349 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3333 mask_tok=100.00 y_norm=48.0883±0.0056 +2026-05-10 19:02:07,944 - INFO - Step 11760 | mode=video loss=0.109375 diff=0.109375 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1554 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0882 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=9409 i_micro=2351 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3500 mask_tok=105.00 y_norm=48.0882±0.0054 +2026-05-10 19:02:26,342 - INFO - Step 11770 | mode=video loss=0.116211 diff=0.116211 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0704 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0887 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=9417 i_micro=2353 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3567 mask_tok=107.00 y_norm=48.0887±0.0060 +2026-05-10 19:02:44,863 - INFO - Step 11780 | mode=video loss=0.169922 diff=0.169922 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0425 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0886 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=9425 i_micro=2355 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3500 mask_tok=105.00 y_norm=48.0886±0.0058 +2026-05-10 19:03:03,290 - INFO - Step 11790 | mode=video loss=0.066895 diff=0.066895 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0415 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0881 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=9433 i_micro=2357 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3500 mask_tok=105.00 y_norm=48.0881±0.0054 +2026-05-10 19:03:23,495 - INFO - Step 11800 | mode=video loss=0.077148 diff=0.077148 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0448 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0874 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=9441 i_micro=2359 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3500 mask_tok=105.00 y_norm=48.0874±0.0055 cond_uncond_dloss=0.375000 cond_grad=0.000106 cond_pred_l2=36.784836 cond_pred_ratio=0.022350 +2026-05-10 19:03:23,532 - INFO - Step 11800 | probe_semantic mcp_offdiag(mean/min/max)=0.645194/0.531465/0.707936 smol_offdiag=0.660557 prompts=4 mcp_tok=17.50 smol_tok=17.50 +2026-05-10 19:03:30,396 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 19:03:48,351 - INFO - Step 11810 | mode=video loss=0.091309 diff=0.091309 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1937 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0875 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=9449 i_micro=2361 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4367 mask_tok=131.00 y_norm=48.0875±0.0072 +2026-05-10 19:04:06,771 - INFO - Step 11820 | mode=video loss=0.079102 diff=0.079102 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0599 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0875 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=9457 i_micro=2363 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3667 mask_tok=110.00 y_norm=48.0875±0.0054 +2026-05-10 19:04:25,521 - INFO - Step 11830 | mode=video loss=0.087891 diff=0.087891 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0586 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0859 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=9465 i_micro=2365 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3800 mask_tok=114.00 y_norm=48.0859±0.0057 +2026-05-10 19:04:43,480 - INFO - Step 11840 | mode=video loss=0.072754 diff=0.072754 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0306 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0872 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=9473 i_micro=2367 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3600 mask_tok=108.00 y_norm=48.0872±0.0059 +2026-05-10 19:05:02,599 - INFO - Step 11850 | mode=video loss=0.121582 diff=0.121582 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0489 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0884 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=9481 i_micro=2369 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.2733 mask_tok=82.00 y_norm=48.0884±0.0047 +2026-05-10 19:05:20,550 - INFO - Step 11860 | mode=video loss=0.082520 diff=0.082520 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0409 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0885 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=9489 i_micro=2371 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3333 mask_tok=100.00 y_norm=48.0885±0.0062 +2026-05-10 19:05:38,948 - INFO - Step 11870 | mode=video loss=0.081055 diff=0.081055 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0341 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0881 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=9497 i_micro=2373 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.5200 mask_tok=156.00 y_norm=48.0881±0.0064 +2026-05-10 19:05:57,543 - INFO - Step 11880 | mode=video loss=0.164062 diff=0.164062 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0869 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0864 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=9505 i_micro=2375 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3467 mask_tok=104.00 y_norm=48.0864±0.0062 +2026-05-10 19:06:15,510 - INFO - Step 11890 | mode=video loss=0.079102 diff=0.079102 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0387 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0884 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=9513 i_micro=2377 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3333 mask_tok=100.00 y_norm=48.0884±0.0053 +2026-05-10 19:06:33,923 - INFO - Step 11900 | mode=video loss=0.075684 diff=0.075684 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0303 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0879 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=9521 i_micro=2379 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4900 mask_tok=147.00 y_norm=48.0879±0.0060 +2026-05-10 19:06:40,780 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 19:06:59,470 - INFO - Step 11910 | mode=video loss=0.073242 diff=0.073242 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0752 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0859 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=9529 i_micro=2381 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4033 mask_tok=121.00 y_norm=48.0859±0.0059 +2026-05-10 19:07:17,873 - INFO - Step 11920 | mode=video loss=0.093750 diff=0.093750 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0780 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0870 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=9537 i_micro=2383 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.5000 mask_tok=150.00 y_norm=48.0870±0.0070 +2026-05-10 19:07:36,575 - INFO - Step 11930 | mode=video loss=0.088379 diff=0.088379 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0421 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0855 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=9545 i_micro=2385 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4033 mask_tok=121.00 y_norm=48.0855±0.0064 +2026-05-10 19:07:54,543 - INFO - Step 11940 | mode=video loss=0.088867 diff=0.088867 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0440 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0879 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=9553 i_micro=2387 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3333 mask_tok=100.00 y_norm=48.0879±0.0051 +2026-05-10 19:08:12,984 - INFO - Step 11950 | mode=video loss=0.097168 diff=0.097168 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0367 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0903 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=9561 i_micro=2389 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3933 mask_tok=118.00 y_norm=48.0903±0.0053 +2026-05-10 19:08:31,658 - INFO - Step 11960 | mode=video loss=0.163086 diff=0.163086 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1363 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0874 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=9569 i_micro=2391 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.5000 mask_tok=150.00 y_norm=48.0874±0.0064 +2026-05-10 19:08:50,024 - INFO - Step 11970 | mode=video loss=0.211914 diff=0.211914 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0928 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0877 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=9577 i_micro=2393 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3433 mask_tok=103.00 y_norm=48.0877±0.0051 +2026-05-10 19:09:08,733 - INFO - Step 11980 | mode=video loss=0.085938 diff=0.085938 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0514 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0879 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=9585 i_micro=2395 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3400 mask_tok=102.00 y_norm=48.0879±0.0054 +2026-05-10 19:09:26,682 - INFO - Step 11990 | mode=video loss=0.074219 diff=0.074219 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0800 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0862 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=9593 i_micro=2397 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4233 mask_tok=127.00 y_norm=48.0862±0.0058 +2026-05-10 19:09:46,723 - INFO - Step 12000 | mode=video loss=0.112793 diff=0.112793 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0715 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0866 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=9601 i_micro=2399 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4033 mask_tok=121.00 y_norm=48.0866±0.0061 cond_uncond_dloss=0.291016 cond_grad=0.000091 cond_pred_l2=35.455669 cond_pred_ratio=0.022155 +2026-05-10 19:09:46,760 - INFO - Step 12000 | probe_semantic mcp_offdiag(mean/min/max)=0.645620/0.532001/0.708304 smol_offdiag=0.660557 prompts=4 mcp_tok=17.50 smol_tok=17.50 +2026-05-10 19:09:55,707 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 19:10:14,438 - INFO - Step 12010 | mode=video loss=0.077148 diff=0.077148 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0689 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0885 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=9609 i_micro=2401 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3333 mask_tok=100.00 y_norm=48.0885±0.0054 +2026-05-10 19:10:32,433 - INFO - Step 12020 | mode=video loss=0.068359 diff=0.068359 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0571 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0862 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=9617 i_micro=2403 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4133 mask_tok=124.00 y_norm=48.0862±0.0064 +2026-05-10 19:10:51,534 - INFO - Step 12030 | mode=video loss=0.098633 diff=0.098633 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0491 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0858 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=9625 i_micro=2405 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3300 mask_tok=99.00 y_norm=48.0858±0.0057 +2026-05-10 19:11:09,509 - INFO - Step 12040 | mode=video loss=0.075195 diff=0.075195 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0331 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0867 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=9633 i_micro=2407 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4300 mask_tok=129.00 y_norm=48.0867±0.0058 +2026-05-10 19:11:27,922 - INFO - Step 12050 | mode=video loss=0.109375 diff=0.109375 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0394 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0887 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=9641 i_micro=2409 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3400 mask_tok=102.00 y_norm=48.0887±0.0058 +2026-05-10 19:11:46,490 - INFO - Step 12060 | mode=video loss=0.087402 diff=0.087402 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0480 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0875 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=9649 i_micro=2411 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3967 mask_tok=119.00 y_norm=48.0875±0.0061 +2026-05-10 19:12:04,449 - INFO - Step 12070 | mode=video loss=0.095703 diff=0.095703 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0389 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0874 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=9657 i_micro=2413 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3633 mask_tok=109.00 y_norm=48.0874±0.0056 +2026-05-10 19:12:22,885 - INFO - Step 12080 | mode=video loss=0.091309 diff=0.091309 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0510 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0901 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=9665 i_micro=2415 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3500 mask_tok=105.00 y_norm=48.0901±0.0052 +2026-05-10 19:12:41,479 - INFO - Step 12090 | mode=video loss=0.126953 diff=0.126953 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0578 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0885 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=9673 i_micro=2417 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3133 mask_tok=94.00 y_norm=48.0885±0.0052 +2026-05-10 19:12:59,433 - INFO - Step 12100 | mode=video loss=0.082031 diff=0.082031 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0299 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0873 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=9681 i_micro=2419 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3567 mask_tok=107.00 y_norm=48.0873±0.0065 +2026-05-10 19:13:06,747 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 19:13:25,300 - INFO - Step 12110 | mode=video loss=0.066406 diff=0.066406 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0358 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0858 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=9689 i_micro=2421 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4233 mask_tok=127.00 y_norm=48.0858±0.0060 +2026-05-10 19:13:43,255 - INFO - Step 12120 | mode=video loss=0.094238 diff=0.094238 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0570 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0893 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=9697 i_micro=2423 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.2667 mask_tok=80.00 y_norm=48.0893±0.0057 +2026-05-10 19:14:01,695 - INFO - Step 12130 | mode=video loss=0.073242 diff=0.073242 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0307 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0871 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=9705 i_micro=2425 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.5167 mask_tok=155.00 y_norm=48.0871±0.0066 +2026-05-10 19:14:20,317 - INFO - Step 12140 | mode=video loss=0.078613 diff=0.078613 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0350 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0880 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=9713 i_micro=2427 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3400 mask_tok=102.00 y_norm=48.0880±0.0051 +2026-05-10 19:14:38,272 - INFO - Step 12150 | mode=video loss=0.150391 diff=0.150391 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0431 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0872 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=9721 i_micro=2429 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4533 mask_tok=136.00 y_norm=48.0872±0.0058 +2026-05-10 19:14:57,203 - INFO - Step 12160 | mode=video loss=0.094238 diff=0.094238 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0754 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0894 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=9729 i_micro=2431 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3100 mask_tok=93.00 y_norm=48.0894±0.0060 +2026-05-10 19:15:15,162 - INFO - Step 12170 | mode=video loss=0.087402 diff=0.087402 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0782 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0871 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=9737 i_micro=2433 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4067 mask_tok=122.00 y_norm=48.0871±0.0062 +2026-05-10 19:15:33,593 - INFO - Step 12180 | mode=video loss=0.107422 diff=0.107422 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0318 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0872 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=9745 i_micro=2435 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3400 mask_tok=102.00 y_norm=48.0872±0.0056 +2026-05-10 19:15:52,066 - INFO - Step 12190 | mode=video loss=0.060059 diff=0.060059 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0738 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0886 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=9753 i_micro=2437 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3667 mask_tok=110.00 y_norm=48.0886±0.0060 +2026-05-10 19:16:11,653 - INFO - Step 12200 | mode=video loss=0.074707 diff=0.074707 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0434 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0880 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=9761 i_micro=2439 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4800 mask_tok=144.00 y_norm=48.0880±0.0054 cond_uncond_dloss=0.341797 cond_grad=0.000056 cond_pred_l2=25.596077 cond_pred_ratio=0.013700 +2026-05-10 19:16:11,690 - INFO - Step 12200 | probe_semantic mcp_offdiag(mean/min/max)=0.645528/0.531958/0.708175 smol_offdiag=0.660557 prompts=4 mcp_tok=17.50 smol_tok=17.50 +2026-05-10 19:16:18,435 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 19:16:36,822 - INFO - Step 12210 | mode=video loss=0.109863 diff=0.109863 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0668 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0892 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=9769 i_micro=2441 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3267 mask_tok=98.00 y_norm=48.0892±0.0055 +2026-05-10 19:16:55,407 - INFO - Step 12220 | mode=video loss=0.078125 diff=0.078125 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0761 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0883 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=9777 i_micro=2443 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3333 mask_tok=100.00 y_norm=48.0883±0.0061 +2026-05-10 19:17:13,768 - INFO - Step 12230 | mode=video loss=0.468750 diff=0.468750 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.4441 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0872 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=9785 i_micro=2445 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4300 mask_tok=129.00 y_norm=48.0872±0.0064 +2026-05-10 19:17:32,397 - INFO - Step 12240 | mode=video loss=0.059814 diff=0.059814 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0413 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0874 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=9793 i_micro=2447 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3433 mask_tok=103.00 y_norm=48.0874±0.0066 +2026-05-10 19:17:50,384 - INFO - Step 12250 | mode=video loss=0.096191 diff=0.096191 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0585 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0866 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=9801 i_micro=2449 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3633 mask_tok=109.00 y_norm=48.0866±0.0062 +2026-05-10 19:18:08,749 - INFO - Step 12260 | mode=video loss=0.099121 diff=0.099121 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0312 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0865 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=9809 i_micro=2451 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4333 mask_tok=130.00 y_norm=48.0865±0.0065 +2026-05-10 19:18:27,411 - INFO - Step 12270 | mode=video loss=0.067383 diff=0.067383 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0492 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0877 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=9817 i_micro=2453 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3833 mask_tok=115.00 y_norm=48.0877±0.0068 +2026-05-10 19:18:45,379 - INFO - Step 12280 | mode=video loss=0.089844 diff=0.089844 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0788 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0864 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=9825 i_micro=2455 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3800 mask_tok=114.00 y_norm=48.0864±0.0054 +2026-05-10 19:19:04,525 - INFO - Step 12290 | mode=video loss=0.128906 diff=0.128906 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0498 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0858 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=9833 i_micro=2457 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3867 mask_tok=116.00 y_norm=48.0858±0.0060 +2026-05-10 19:19:22,475 - INFO - Step 12300 | mode=video loss=0.098633 diff=0.098633 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0291 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0877 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=9841 i_micro=2459 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3367 mask_tok=101.00 y_norm=48.0877±0.0053 +2026-05-10 19:19:29,417 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 19:19:47,756 - INFO - Step 12310 | mode=video loss=0.091309 diff=0.091309 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0522 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0876 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=9849 i_micro=2461 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3467 mask_tok=104.00 y_norm=48.0876±0.0060 +2026-05-10 19:20:06,440 - INFO - Step 12320 | mode=video loss=0.129883 diff=0.129883 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0273 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0868 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=9857 i_micro=2463 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3767 mask_tok=113.00 y_norm=48.0868±0.0058 +2026-05-10 19:20:24,405 - INFO - Step 12330 | mode=video loss=0.070801 diff=0.070801 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0305 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0896 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=9865 i_micro=2465 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3067 mask_tok=92.00 y_norm=48.0896±0.0053 +2026-05-10 19:20:42,779 - INFO - Step 12340 | mode=video loss=0.057861 diff=0.057861 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0400 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0878 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=9873 i_micro=2467 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3367 mask_tok=101.00 y_norm=48.0878±0.0046 +2026-05-10 19:21:01,500 - INFO - Step 12350 | mode=video loss=0.133789 diff=0.133789 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0672 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0873 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=9881 i_micro=2469 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4800 mask_tok=144.00 y_norm=48.0873±0.0064 +2026-05-10 19:21:19,856 - INFO - Step 12360 | mode=video loss=0.025146 diff=0.025146 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0901 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0865 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=9889 i_micro=2471 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4767 mask_tok=143.00 y_norm=48.0865±0.0067 +2026-05-10 19:21:38,394 - INFO - Step 12370 | mode=video loss=0.117676 diff=0.117676 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0802 emb_mean=0.0008 emb_std=1.0017 tok_norm=48.0834 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=9897 i_micro=2473 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3900 mask_tok=117.00 y_norm=48.0834±0.0063 +2026-05-10 19:21:56,347 - INFO - Step 12380 | mode=video loss=0.101074 diff=0.101074 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0786 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0873 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=9905 i_micro=2475 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4367 mask_tok=131.00 y_norm=48.0873±0.0057 +2026-05-10 19:22:14,727 - INFO - Step 12390 | mode=video loss=0.159180 diff=0.159180 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0258 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0887 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=9913 i_micro=2477 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3267 mask_tok=98.00 y_norm=48.0887±0.0058 +2026-05-10 19:22:34,833 - INFO - Step 12400 | mode=video loss=0.085449 diff=0.085449 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0255 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0881 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=9921 i_micro=2479 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4367 mask_tok=131.00 y_norm=48.0881±0.0064 cond_uncond_dloss=0.412109 cond_grad=0.000147 cond_pred_l2=45.909760 cond_pred_ratio=0.029250 +2026-05-10 19:22:34,870 - INFO - Step 12400 | probe_semantic mcp_offdiag(mean/min/max)=0.645192/0.531708/0.707745 smol_offdiag=0.660557 prompts=4 mcp_tok=17.50 smol_tok=17.50 +2026-05-10 19:22:41,720 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 19:22:59,671 - INFO - Step 12410 | mode=video loss=0.082031 diff=0.082031 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0524 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0882 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=9929 i_micro=2481 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3167 mask_tok=95.00 y_norm=48.0882±0.0060 +2026-05-10 19:23:18,850 - INFO - Step 12420 | mode=video loss=0.155273 diff=0.155273 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0663 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0890 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=9937 i_micro=2483 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.2933 mask_tok=88.00 y_norm=48.0890±0.0059 +2026-05-10 19:23:36,810 - INFO - Step 12430 | mode=video loss=0.080078 diff=0.080078 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0228 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0868 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=9945 i_micro=2485 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3300 mask_tok=99.00 y_norm=48.0868±0.0052 +2026-05-10 19:23:55,215 - INFO - Step 12440 | mode=video loss=0.095703 diff=0.095703 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0601 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0869 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=9953 i_micro=2487 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3600 mask_tok=108.00 y_norm=48.0869±0.0059 +2026-05-10 19:24:13,888 - INFO - Step 12450 | mode=video loss=0.099121 diff=0.099121 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0312 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0879 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=9961 i_micro=2489 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3667 mask_tok=110.00 y_norm=48.0879±0.0069 +2026-05-10 19:24:31,849 - INFO - Step 12460 | mode=video loss=0.066406 diff=0.066406 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0390 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0870 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=9969 i_micro=2491 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3567 mask_tok=107.00 y_norm=48.0870±0.0068 +2026-05-10 19:24:50,247 - INFO - Step 12470 | mode=video loss=0.127930 diff=0.127930 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0553 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0886 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=9977 i_micro=2493 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.2867 mask_tok=86.00 y_norm=48.0886±0.0052 +2026-05-10 19:25:08,940 - INFO - Step 12480 | mode=video loss=0.114746 diff=0.114746 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0385 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0856 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=9985 i_micro=2495 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4033 mask_tok=121.00 y_norm=48.0856±0.0061 +2026-05-10 19:25:27,321 - INFO - Step 12490 | mode=video loss=0.080078 diff=0.080078 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0937 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0872 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=9993 i_micro=2497 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4233 mask_tok=127.00 y_norm=48.0872±0.0058 +2026-05-10 19:25:45,834 - INFO - Step 12500 | mode=video loss=0.155273 diff=0.155273 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1186 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0861 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=10001 i_micro=2499 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3900 mask_tok=117.00 y_norm=48.0861±0.0064 +2026-05-10 19:25:52,807 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 19:26:10,752 - INFO - Step 12510 | mode=video loss=0.090820 diff=0.090820 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0378 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0890 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=10009 i_micro=2501 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3133 mask_tok=94.00 y_norm=48.0890±0.0053 +2026-05-10 19:26:29,168 - INFO - Step 12520 | mode=video loss=0.131836 diff=0.131836 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0957 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0862 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=10017 i_micro=2503 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4000 mask_tok=120.00 y_norm=48.0862±0.0064 +2026-05-10 19:26:47,826 - INFO - Step 12530 | mode=video loss=0.069824 diff=0.069824 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0790 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0864 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=10025 i_micro=2505 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4367 mask_tok=131.00 y_norm=48.0864±0.0068 +2026-05-10 19:27:06,225 - INFO - Step 12540 | mode=video loss=0.097168 diff=0.097168 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0498 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0896 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=10033 i_micro=2507 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3367 mask_tok=101.00 y_norm=48.0896±0.0052 +2026-05-10 19:27:24,965 - INFO - Step 12550 | mode=video loss=0.121094 diff=0.121094 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0537 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0881 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=10041 i_micro=2509 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3600 mask_tok=108.00 y_norm=48.0881±0.0056 +2026-05-10 19:27:42,928 - INFO - Step 12560 | mode=video loss=0.080566 diff=0.080566 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0731 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0879 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=10049 i_micro=2511 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4767 mask_tok=143.00 y_norm=48.0879±0.0070 +2026-05-10 19:28:01,347 - INFO - Step 12570 | mode=video loss=0.115723 diff=0.115723 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0399 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0876 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=10057 i_micro=2513 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4167 mask_tok=125.00 y_norm=48.0876±0.0055 +2026-05-10 19:28:20,072 - INFO - Step 12580 | mode=video loss=0.165039 diff=0.165039 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0913 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0883 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=10065 i_micro=2515 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.5300 mask_tok=159.00 y_norm=48.0883±0.0080 +2026-05-10 19:28:38,038 - INFO - Step 12590 | mode=video loss=0.082031 diff=0.082031 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0494 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0886 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=10073 i_micro=2517 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3567 mask_tok=107.00 y_norm=48.0886±0.0060 +2026-05-10 19:28:58,038 - INFO - Step 12600 | mode=video loss=0.089355 diff=0.089355 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0354 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0881 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=10081 i_micro=2519 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3533 mask_tok=106.00 y_norm=48.0881±0.0062 cond_uncond_dloss=1.234375 cond_grad=0.000113 cond_pred_l2=56.313931 cond_pred_ratio=0.033726 +2026-05-10 19:28:58,075 - INFO - Step 12600 | probe_semantic mcp_offdiag(mean/min/max)=0.645302/0.532064/0.707720 smol_offdiag=0.660557 prompts=4 mcp_tok=17.50 smol_tok=17.50 +2026-05-10 19:29:04,921 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 19:29:23,543 - INFO - Step 12610 | mode=video loss=0.092285 diff=0.092285 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0294 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0893 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=10089 i_micro=2521 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3467 mask_tok=104.00 y_norm=48.0893±0.0058 +2026-05-10 19:29:41,962 - INFO - Step 12620 | mode=video loss=0.129883 diff=0.129883 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0772 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0865 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=10097 i_micro=2523 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4400 mask_tok=132.00 y_norm=48.0865±0.0062 +2026-05-10 19:30:00,731 - INFO - Step 12630 | mode=video loss=0.098633 diff=0.098633 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0438 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0868 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=10105 i_micro=2525 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3233 mask_tok=97.00 y_norm=48.0868±0.0060 +2026-05-10 19:30:18,694 - INFO - Step 12640 | mode=video loss=0.132812 diff=0.132812 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1266 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0857 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=10113 i_micro=2527 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4300 mask_tok=129.00 y_norm=48.0857±0.0058 +2026-05-10 19:30:37,092 - INFO - Step 12650 | mode=video loss=0.108398 diff=0.108398 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0299 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0880 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=10121 i_micro=2529 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3700 mask_tok=111.00 y_norm=48.0880±0.0061 +2026-05-10 19:30:55,620 - INFO - Step 12660 | mode=video loss=0.097168 diff=0.097168 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0552 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0880 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=10129 i_micro=2531 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3700 mask_tok=111.00 y_norm=48.0880±0.0056 +2026-05-10 19:31:14,027 - INFO - Step 12670 | mode=video loss=0.148438 diff=0.148438 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1314 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0871 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=10137 i_micro=2533 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3667 mask_tok=110.00 y_norm=48.0871±0.0052 +2026-05-10 19:31:32,586 - INFO - Step 12680 | mode=video loss=0.122559 diff=0.122559 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0840 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0864 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=10145 i_micro=2535 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3667 mask_tok=110.00 y_norm=48.0864±0.0067 +2026-05-10 19:31:50,563 - INFO - Step 12690 | mode=video loss=0.099121 diff=0.099121 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0423 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0880 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=10153 i_micro=2537 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4733 mask_tok=142.00 y_norm=48.0880±0.0065 +2026-05-10 19:32:08,964 - INFO - Step 12700 | mode=video loss=0.090820 diff=0.090820 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0389 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0885 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=10161 i_micro=2539 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3433 mask_tok=103.00 y_norm=48.0885±0.0055 +2026-05-10 19:32:16,040 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 19:32:34,729 - INFO - Step 12710 | mode=video loss=0.090332 diff=0.090332 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0446 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0871 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=10169 i_micro=2541 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.2933 mask_tok=88.00 y_norm=48.0871±0.0055 +2026-05-10 19:32:52,685 - INFO - Step 12720 | mode=video loss=0.084473 diff=0.084473 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0586 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0883 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=10177 i_micro=2543 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3333 mask_tok=100.00 y_norm=48.0883±0.0058 +2026-05-10 19:33:11,078 - INFO - Step 12730 | mode=video loss=0.093262 diff=0.093262 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0404 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0853 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=10185 i_micro=2545 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4300 mask_tok=129.00 y_norm=48.0853±0.0060 +2026-05-10 19:33:29,760 - INFO - Step 12740 | mode=video loss=0.090820 diff=0.090820 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0686 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0870 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=10193 i_micro=2547 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.2800 mask_tok=84.00 y_norm=48.0870±0.0049 +2026-05-10 19:33:48,114 - INFO - Step 12750 | mode=video loss=0.044189 diff=0.044189 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0846 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0855 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=10201 i_micro=2549 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3800 mask_tok=114.00 y_norm=48.0855±0.0057 +2026-05-10 19:34:06,649 - INFO - Step 12760 | mode=video loss=0.063965 diff=0.063965 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0260 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0851 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=10209 i_micro=2551 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4133 mask_tok=124.00 y_norm=48.0851±0.0056 +2026-05-10 19:34:24,609 - INFO - Step 12770 | mode=video loss=0.242188 diff=0.242188 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.2951 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0881 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=10217 i_micro=2553 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.2500 mask_tok=75.00 y_norm=48.0881±0.0049 +2026-05-10 19:34:42,986 - INFO - Step 12780 | mode=video loss=0.086914 diff=0.086914 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0392 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0888 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=10225 i_micro=2555 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.2700 mask_tok=81.00 y_norm=48.0888±0.0058 +2026-05-10 19:35:01,447 - INFO - Step 12790 | mode=video loss=0.065918 diff=0.065918 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0417 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0875 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=10233 i_micro=2557 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.8800 mask_tok=264.00 y_norm=48.0875±0.0074 +2026-05-10 19:35:21,466 - INFO - Step 12800 | mode=video loss=0.130859 diff=0.130859 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1355 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0861 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=10241 i_micro=2559 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3600 mask_tok=108.00 y_norm=48.0861±0.0058 cond_uncond_dloss=0.898438 cond_grad=0.000708 cond_pred_l2=40.139278 cond_pred_ratio=0.025331 +2026-05-10 19:35:21,502 - INFO - Step 12800 | probe_semantic mcp_offdiag(mean/min/max)=0.645325/0.532080/0.707761 smol_offdiag=0.660557 prompts=4 mcp_tok=17.50 smol_tok=17.50 +2026-05-10 19:35:28,316 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 19:35:47,060 - INFO - Step 12810 | mode=video loss=0.123047 diff=0.123047 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0478 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0860 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=10249 i_micro=2561 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3867 mask_tok=116.00 y_norm=48.0860±0.0060 +2026-05-10 19:36:05,033 - INFO - Step 12820 | mode=video loss=0.074219 diff=0.074219 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0423 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0872 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=10257 i_micro=2563 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3367 mask_tok=101.00 y_norm=48.0872±0.0052 +2026-05-10 19:36:23,478 - INFO - Step 12830 | mode=video loss=0.140625 diff=0.140625 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0359 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0859 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=10265 i_micro=2565 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4767 mask_tok=143.00 y_norm=48.0859±0.0058 +2026-05-10 19:36:42,221 - INFO - Step 12840 | mode=video loss=0.109863 diff=0.109863 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0296 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0863 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=10273 i_micro=2567 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3400 mask_tok=102.00 y_norm=48.0863±0.0058 +2026-05-10 19:37:00,173 - INFO - Step 12850 | mode=video loss=0.100098 diff=0.100098 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0535 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0860 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=10281 i_micro=2569 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3333 mask_tok=100.00 y_norm=48.0860±0.0060 +2026-05-10 19:37:18,554 - INFO - Step 12860 | mode=video loss=0.090332 diff=0.090332 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0535 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0858 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=10289 i_micro=2571 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3500 mask_tok=105.00 y_norm=48.0858±0.0065 +2026-05-10 19:37:37,313 - INFO - Step 12870 | mode=video loss=0.066895 diff=0.066895 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0541 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0884 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=10297 i_micro=2573 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3533 mask_tok=106.00 y_norm=48.0884±0.0060 +2026-05-10 19:37:55,673 - INFO - Step 12880 | mode=video loss=0.082520 diff=0.082520 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0307 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0854 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=10305 i_micro=2575 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3700 mask_tok=111.00 y_norm=48.0854±0.0061 +2026-05-10 19:38:14,374 - INFO - Step 12890 | mode=video loss=0.123535 diff=0.123535 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0487 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0866 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=10313 i_micro=2577 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3567 mask_tok=107.00 y_norm=48.0866±0.0060 +2026-05-10 19:38:32,339 - INFO - Step 12900 | mode=video loss=0.103516 diff=0.103516 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0617 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0885 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=10321 i_micro=2579 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3433 mask_tok=103.00 y_norm=48.0885±0.0057 +2026-05-10 19:38:39,182 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 19:38:57,540 - INFO - Step 12910 | mode=video loss=0.071289 diff=0.071289 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0411 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0851 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=10329 i_micro=2581 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4267 mask_tok=128.00 y_norm=48.0851±0.0059 +2026-05-10 19:39:16,292 - INFO - Step 12920 | mode=video loss=0.063965 diff=0.063965 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0646 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0871 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=10337 i_micro=2583 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4833 mask_tok=145.00 y_norm=48.0871±0.0064 +2026-05-10 19:39:34,668 - INFO - Step 12930 | mode=video loss=0.081543 diff=0.081543 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0772 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0870 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=10345 i_micro=2585 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3467 mask_tok=104.00 y_norm=48.0870±0.0065 +2026-05-10 19:39:53,302 - INFO - Step 12940 | mode=video loss=0.095215 diff=0.095215 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0329 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0876 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=10353 i_micro=2587 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3300 mask_tok=99.00 y_norm=48.0876±0.0058 +2026-05-10 19:40:11,263 - INFO - Step 12950 | mode=video loss=0.111816 diff=0.111816 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0538 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0887 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=10361 i_micro=2589 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3367 mask_tok=101.00 y_norm=48.0887±0.0058 +2026-05-10 19:40:29,658 - INFO - Step 12960 | mode=video loss=0.105957 diff=0.105957 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0639 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0884 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=10369 i_micro=2591 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3000 mask_tok=90.00 y_norm=48.0884±0.0055 +2026-05-10 19:40:48,386 - INFO - Step 12970 | mode=video loss=0.092773 diff=0.092773 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0751 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0869 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=10377 i_micro=2593 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4167 mask_tok=125.00 y_norm=48.0869±0.0065 +2026-05-10 19:41:06,345 - INFO - Step 12980 | mode=video loss=0.108887 diff=0.108887 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0802 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0867 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=10385 i_micro=2595 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3533 mask_tok=106.00 y_norm=48.0867±0.0056 +2026-05-10 19:41:24,768 - INFO - Step 12990 | mode=video loss=0.100586 diff=0.100586 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0781 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0879 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=10393 i_micro=2597 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3667 mask_tok=110.00 y_norm=48.0879±0.0053 +2026-05-10 19:41:44,941 - INFO - Step 13000 | mode=video loss=0.088379 diff=0.088379 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0534 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0867 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=10401 i_micro=2599 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4067 mask_tok=122.00 y_norm=48.0867±0.0070 cond_uncond_dloss=0.220703 cond_grad=0.000068 cond_pred_l2=36.685341 cond_pred_ratio=0.022224 +2026-05-10 19:41:44,978 - INFO - Step 13000 | probe_semantic mcp_offdiag(mean/min/max)=0.645283/0.531975/0.707867 smol_offdiag=0.660557 prompts=4 mcp_tok=17.50 smol_tok=17.50 +2026-05-10 19:41:51,841 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 19:42:10,218 - INFO - Step 13010 | mode=video loss=0.067871 diff=0.067871 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0325 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0874 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=10409 i_micro=2601 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3267 mask_tok=98.00 y_norm=48.0874±0.0053 +2026-05-10 19:42:28,855 - INFO - Step 13020 | mode=video loss=0.106934 diff=0.106934 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0273 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0868 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=10417 i_micro=2603 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3533 mask_tok=106.00 y_norm=48.0868±0.0058 +2026-05-10 19:42:46,819 - INFO - Step 13030 | mode=video loss=0.106934 diff=0.106934 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0511 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0878 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=10425 i_micro=2605 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3300 mask_tok=99.00 y_norm=48.0878±0.0056 +2026-05-10 19:43:05,230 - INFO - Step 13040 | mode=video loss=0.080566 diff=0.080566 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0664 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0883 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=10433 i_micro=2607 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3667 mask_tok=110.00 y_norm=48.0883±0.0058 +2026-05-10 19:43:23,845 - INFO - Step 13050 | mode=video loss=0.059814 diff=0.059814 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0596 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0876 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=10441 i_micro=2609 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4433 mask_tok=133.00 y_norm=48.0876±0.0072 +2026-05-10 19:43:42,319 - INFO - Step 13060 | mode=video loss=0.085449 diff=0.085449 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0610 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0872 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=10449 i_micro=2611 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3400 mask_tok=102.00 y_norm=48.0872±0.0053 +2026-05-10 19:44:00,999 - INFO - Step 13070 | mode=video loss=0.092285 diff=0.092285 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0472 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0871 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=10457 i_micro=2613 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3767 mask_tok=113.00 y_norm=48.0871±0.0077 +2026-05-10 19:44:18,947 - INFO - Step 13080 | mode=video loss=0.132812 diff=0.132812 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0275 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0860 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=10465 i_micro=2615 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4567 mask_tok=137.00 y_norm=48.0860±0.0054 +2026-05-10 19:44:37,420 - INFO - Step 13090 | mode=video loss=0.232422 diff=0.232422 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1142 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0867 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=10473 i_micro=2617 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3633 mask_tok=109.00 y_norm=48.0867±0.0057 +2026-05-10 19:44:56,059 - INFO - Step 13100 | mode=video loss=0.080566 diff=0.080566 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0468 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0864 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=10481 i_micro=2619 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3833 mask_tok=115.00 y_norm=48.0864±0.0080 +2026-05-10 19:45:02,929 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 19:45:20,869 - INFO - Step 13110 | mode=video loss=0.077148 diff=0.077148 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0356 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0881 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=10489 i_micro=2621 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3333 mask_tok=100.00 y_norm=48.0881±0.0053 +2026-05-10 19:45:40,027 - INFO - Step 13120 | mode=video loss=0.064453 diff=0.064453 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.2370 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0866 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=10497 i_micro=2623 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3633 mask_tok=109.00 y_norm=48.0866±0.0058 +2026-05-10 19:45:57,987 - INFO - Step 13130 | mode=video loss=0.122559 diff=0.122559 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0532 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0883 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=10505 i_micro=2625 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3367 mask_tok=101.00 y_norm=48.0883±0.0057 +2026-05-10 19:46:16,418 - INFO - Step 13140 | mode=video loss=0.099121 diff=0.099121 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0469 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0875 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=10513 i_micro=2627 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3767 mask_tok=113.00 y_norm=48.0875±0.0059 +2026-05-10 19:46:35,105 - INFO - Step 13150 | mode=video loss=0.099609 diff=0.099609 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0605 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0859 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=10521 i_micro=2629 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4100 mask_tok=123.00 y_norm=48.0859±0.0059 +2026-05-10 19:46:53,061 - INFO - Step 13160 | mode=video loss=0.078125 diff=0.078125 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0271 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0873 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=10529 i_micro=2631 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.5400 mask_tok=162.00 y_norm=48.0873±0.0071 +2026-05-10 19:47:11,501 - INFO - Step 13170 | mode=video loss=0.105469 diff=0.105469 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0550 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0883 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=10537 i_micro=2633 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4100 mask_tok=123.00 y_norm=48.0883±0.0066 +2026-05-10 19:47:30,007 - INFO - Step 13180 | mode=video loss=0.121582 diff=0.121582 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0391 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0854 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=10545 i_micro=2635 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4067 mask_tok=122.00 y_norm=48.0854±0.0063 +2026-05-10 19:47:48,466 - INFO - Step 13190 | mode=video loss=0.095215 diff=0.095215 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0652 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0877 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=10553 i_micro=2637 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3533 mask_tok=106.00 y_norm=48.0877±0.0056 +2026-05-10 19:48:08,561 - INFO - Step 13200 | mode=video loss=0.086914 diff=0.086914 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0355 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0892 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=10561 i_micro=2639 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3567 mask_tok=107.00 y_norm=48.0892±0.0055 cond_uncond_dloss=0.229492 cond_grad=0.000074 cond_pred_l2=40.662003 cond_pred_ratio=0.023774 +2026-05-10 19:48:08,598 - INFO - Step 13200 | probe_semantic mcp_offdiag(mean/min/max)=0.645964/0.532823/0.708118 smol_offdiag=0.660557 prompts=4 mcp_tok=17.50 smol_tok=17.50 +2026-05-10 19:48:16,755 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 19:48:34,710 - INFO - Step 13210 | mode=video loss=0.101074 diff=0.101074 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0671 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0883 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=10569 i_micro=2641 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3200 mask_tok=96.00 y_norm=48.0883±0.0053 +2026-05-10 19:48:53,191 - INFO - Step 13220 | mode=video loss=0.075684 diff=0.075684 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0546 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0866 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=10577 i_micro=2643 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3600 mask_tok=108.00 y_norm=48.0866±0.0057 +2026-05-10 19:49:11,884 - INFO - Step 13230 | mode=video loss=0.086914 diff=0.086914 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0428 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0854 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=10585 i_micro=2645 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4867 mask_tok=146.00 y_norm=48.0854±0.0057 +2026-05-10 19:49:29,850 - INFO - Step 13240 | mode=video loss=0.103027 diff=0.103027 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0321 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0889 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=10593 i_micro=2647 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4167 mask_tok=125.00 y_norm=48.0889±0.0065 +2026-05-10 19:49:48,715 - INFO - Step 13250 | mode=video loss=0.100586 diff=0.100586 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0621 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0881 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=10601 i_micro=2649 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.2800 mask_tok=84.00 y_norm=48.0881±0.0046 +2026-05-10 19:50:06,666 - INFO - Step 13260 | mode=video loss=0.068359 diff=0.068359 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0347 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0851 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=10609 i_micro=2651 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3633 mask_tok=109.00 y_norm=48.0851±0.0054 +2026-05-10 19:50:25,048 - INFO - Step 13270 | mode=video loss=0.121582 diff=0.121582 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0410 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0859 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=10617 i_micro=2653 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4267 mask_tok=128.00 y_norm=48.0859±0.0062 +2026-05-10 19:50:43,766 - INFO - Step 13280 | mode=video loss=0.096191 diff=0.096191 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0495 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0857 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=10625 i_micro=2655 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3600 mask_tok=108.00 y_norm=48.0857±0.0072 +2026-05-10 19:51:01,722 - INFO - Step 13290 | mode=video loss=0.072754 diff=0.072754 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0596 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0869 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=10633 i_micro=2657 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4967 mask_tok=149.00 y_norm=48.0869±0.0064 +2026-05-10 19:51:20,088 - INFO - Step 13300 | mode=video loss=0.150391 diff=0.150391 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0477 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0870 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=10641 i_micro=2659 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4700 mask_tok=141.00 y_norm=48.0870±0.0060 +2026-05-10 19:51:26,879 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 19:51:44,840 - INFO - Step 13310 | mode=video loss=0.081055 diff=0.081055 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0732 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0901 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=10649 i_micro=2661 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3467 mask_tok=104.00 y_norm=48.0901±0.0058 +2026-05-10 19:52:03,238 - INFO - Step 13320 | mode=video loss=0.145508 diff=0.145508 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0680 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0886 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=10657 i_micro=2663 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3500 mask_tok=105.00 y_norm=48.0886±0.0057 +2026-05-10 19:52:21,789 - INFO - Step 13330 | mode=video loss=0.114746 diff=0.114746 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0496 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0858 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=10665 i_micro=2665 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4733 mask_tok=142.00 y_norm=48.0858±0.0061 +2026-05-10 19:52:39,753 - INFO - Step 13340 | mode=video loss=0.128906 diff=0.128906 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0520 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0864 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=10673 i_micro=2667 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4200 mask_tok=126.00 y_norm=48.0864±0.0057 +2026-05-10 19:52:58,204 - INFO - Step 13350 | mode=video loss=0.083008 diff=0.083008 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0351 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0848 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=10681 i_micro=2669 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4867 mask_tok=146.00 y_norm=48.0848±0.0064 +2026-05-10 19:53:16,766 - INFO - Step 13360 | mode=video loss=0.136719 diff=0.136719 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0449 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0879 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=10689 i_micro=2671 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3367 mask_tok=101.00 y_norm=48.0879±0.0054 +2026-05-10 19:53:34,718 - INFO - Step 13370 | mode=video loss=0.066895 diff=0.066895 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1107 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0885 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=10697 i_micro=2673 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.8667 mask_tok=260.00 y_norm=48.0885±0.0073 +2026-05-10 19:53:53,742 - INFO - Step 13380 | mode=video loss=0.090332 diff=0.090332 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0594 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0872 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=10705 i_micro=2675 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3733 mask_tok=112.00 y_norm=48.0872±0.0060 +2026-05-10 19:54:11,688 - INFO - Step 13390 | mode=video loss=0.105469 diff=0.105469 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0622 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0875 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=10713 i_micro=2677 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3333 mask_tok=100.00 y_norm=48.0875±0.0062 +2026-05-10 19:54:31,769 - INFO - Step 13400 | mode=video loss=0.092285 diff=0.092285 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0463 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0860 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=10721 i_micro=2679 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4000 mask_tok=120.00 y_norm=48.0860±0.0058 cond_uncond_dloss=0.351562 cond_grad=0.000144 cond_pred_l2=38.658703 cond_pred_ratio=0.023402 +2026-05-10 19:54:31,806 - INFO - Step 13400 | probe_semantic mcp_offdiag(mean/min/max)=0.646156/0.532836/0.708257 smol_offdiag=0.660557 prompts=4 mcp_tok=17.50 smol_tok=17.50 +2026-05-10 19:54:38,727 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 19:54:57,347 - INFO - Step 13410 | mode=video loss=0.086914 diff=0.086914 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0475 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0886 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=10729 i_micro=2681 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3467 mask_tok=104.00 y_norm=48.0886±0.0055 +2026-05-10 19:55:15,303 - INFO - Step 13420 | mode=video loss=0.078613 diff=0.078613 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0413 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0891 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=10737 i_micro=2683 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.2833 mask_tok=85.00 y_norm=48.0891±0.0057 +2026-05-10 19:55:34,454 - INFO - Step 13430 | mode=video loss=0.102539 diff=0.102539 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0368 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0874 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=10745 i_micro=2685 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3033 mask_tok=91.00 y_norm=48.0874±0.0052 +2026-05-10 19:55:52,424 - INFO - Step 13440 | mode=video loss=0.116699 diff=0.116699 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0579 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0876 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=10753 i_micro=2687 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3567 mask_tok=107.00 y_norm=48.0876±0.0060 +2026-05-10 19:56:10,842 - INFO - Step 13450 | mode=video loss=0.076660 diff=0.076660 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0318 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0886 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=10761 i_micro=2689 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3800 mask_tok=114.00 y_norm=48.0886±0.0059 +2026-05-10 19:56:29,642 - INFO - Step 13460 | mode=video loss=0.090332 diff=0.090332 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0822 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0878 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=10769 i_micro=2691 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3700 mask_tok=111.00 y_norm=48.0878±0.0061 +2026-05-10 19:56:47,609 - INFO - Step 13470 | mode=video loss=0.063477 diff=0.063477 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0340 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0884 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=10777 i_micro=2693 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3233 mask_tok=97.00 y_norm=48.0884±0.0055 +2026-05-10 19:57:06,014 - INFO - Step 13480 | mode=video loss=0.095703 diff=0.095703 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0406 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0859 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=10785 i_micro=2695 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.5033 mask_tok=151.00 y_norm=48.0859±0.0059 +2026-05-10 19:57:24,838 - INFO - Step 13490 | mode=video loss=0.093262 diff=0.093262 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0629 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0862 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=10793 i_micro=2697 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4500 mask_tok=135.00 y_norm=48.0862±0.0056 +2026-05-10 19:57:42,813 - INFO - Step 13500 | mode=video loss=0.086426 diff=0.086426 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0353 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0885 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=10801 i_micro=2699 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3433 mask_tok=103.00 y_norm=48.0885±0.0058 +2026-05-10 19:57:49,643 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 19:58:08,917 - INFO - Step 13510 | mode=video loss=0.107910 diff=0.107910 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0800 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0880 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=10809 i_micro=2701 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3333 mask_tok=100.00 y_norm=48.0880±0.0048 +2026-05-10 19:58:26,894 - INFO - Step 13520 | mode=video loss=0.111816 diff=0.111816 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0318 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0866 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=10817 i_micro=2703 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3867 mask_tok=116.00 y_norm=48.0866±0.0064 +2026-05-10 19:58:45,335 - INFO - Step 13530 | mode=video loss=0.285156 diff=0.285156 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1277 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0852 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=10825 i_micro=2705 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3533 mask_tok=106.00 y_norm=48.0852±0.0062 +2026-05-10 19:59:04,028 - INFO - Step 13540 | mode=video loss=0.058594 diff=0.058594 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0553 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0871 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=10833 i_micro=2707 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3300 mask_tok=99.00 y_norm=48.0871±0.0057 +2026-05-10 19:59:22,002 - INFO - Step 13550 | mode=video loss=0.066406 diff=0.066406 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0396 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0864 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=10841 i_micro=2709 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3800 mask_tok=114.00 y_norm=48.0864±0.0060 +2026-05-10 19:59:41,256 - INFO - Step 13560 | mode=video loss=0.094238 diff=0.094238 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0685 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0888 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=10849 i_micro=2711 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3433 mask_tok=103.00 y_norm=48.0888±0.0056 +2026-05-10 19:59:59,215 - INFO - Step 13570 | mode=video loss=0.089844 diff=0.089844 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0419 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0877 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=10857 i_micro=2713 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3600 mask_tok=108.00 y_norm=48.0877±0.0060 +2026-05-10 20:00:17,190 - INFO - Step 13580 | mode=video loss=0.102051 diff=0.102051 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0509 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0871 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=10865 i_micro=2715 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3367 mask_tok=101.00 y_norm=48.0871±0.0055 +2026-05-10 20:00:36,319 - INFO - Step 13590 | mode=video loss=0.124023 diff=0.124023 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0392 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0880 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=10873 i_micro=2717 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3533 mask_tok=106.00 y_norm=48.0880±0.0055 +2026-05-10 20:00:55,925 - INFO - Step 13600 | mode=video loss=0.174805 diff=0.174805 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1854 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0870 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=10881 i_micro=2719 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.5467 mask_tok=164.00 y_norm=48.0870±0.0064 cond_uncond_dloss=0.429688 cond_grad=0.001936 cond_pred_l2=47.136814 cond_pred_ratio=0.031792 +2026-05-10 20:00:55,961 - INFO - Step 13600 | probe_semantic mcp_offdiag(mean/min/max)=0.645976/0.532624/0.707930 smol_offdiag=0.660557 prompts=4 mcp_tok=17.50 smol_tok=17.50 +2026-05-10 20:01:02,772 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 20:01:21,944 - INFO - Step 13610 | mode=video loss=0.041992 diff=0.041992 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0426 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0856 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=10889 i_micro=2721 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4200 mask_tok=126.00 y_norm=48.0856±0.0063 +2026-05-10 20:01:39,920 - INFO - Step 13620 | mode=video loss=0.085449 diff=0.085449 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0528 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0883 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=10897 i_micro=2723 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3500 mask_tok=105.00 y_norm=48.0883±0.0054 +2026-05-10 20:01:57,893 - INFO - Step 13630 | mode=video loss=0.083008 diff=0.083008 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0442 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0886 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=10905 i_micro=2725 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3100 mask_tok=93.00 y_norm=48.0886±0.0051 +2026-05-10 20:02:17,061 - INFO - Step 13640 | mode=video loss=0.110840 diff=0.110840 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0264 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0864 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=10913 i_micro=2727 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3433 mask_tok=103.00 y_norm=48.0864±0.0062 +2026-05-10 20:02:35,032 - INFO - Step 13650 | mode=video loss=0.111328 diff=0.111328 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0377 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0860 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=10921 i_micro=2729 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4533 mask_tok=136.00 y_norm=48.0860±0.0064 +2026-05-10 20:02:53,424 - INFO - Step 13660 | mode=video loss=0.111816 diff=0.111816 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0922 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0874 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=10929 i_micro=2731 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3700 mask_tok=111.00 y_norm=48.0874±0.0061 +2026-05-10 20:03:12,085 - INFO - Step 13670 | mode=video loss=0.099121 diff=0.099121 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1154 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0892 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=10937 i_micro=2733 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.2900 mask_tok=87.00 y_norm=48.0892±0.0056 +2026-05-10 20:03:30,049 - INFO - Step 13680 | mode=video loss=0.121094 diff=0.121094 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1568 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0867 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=10945 i_micro=2735 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3433 mask_tok=103.00 y_norm=48.0867±0.0066 +2026-05-10 20:03:49,099 - INFO - Step 13690 | mode=video loss=0.078125 diff=0.078125 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0794 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0887 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=10953 i_micro=2737 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3367 mask_tok=101.00 y_norm=48.0887±0.0055 +2026-05-10 20:04:07,044 - INFO - Step 13700 | mode=video loss=0.063477 diff=0.063477 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0477 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0864 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=10961 i_micro=2739 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3667 mask_tok=110.00 y_norm=48.0864±0.0060 +2026-05-10 20:04:13,873 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 20:04:32,244 - INFO - Step 13710 | mode=video loss=0.082520 diff=0.082520 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0291 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0865 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=10969 i_micro=2741 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4833 mask_tok=145.00 y_norm=48.0865±0.0070 +2026-05-10 20:04:50,889 - INFO - Step 13720 | mode=video loss=0.100586 diff=0.100586 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0398 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0855 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=10977 i_micro=2743 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3633 mask_tok=109.00 y_norm=48.0855±0.0057 +2026-05-10 20:05:08,858 - INFO - Step 13730 | mode=video loss=0.050781 diff=0.050781 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0396 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0871 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=10985 i_micro=2745 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4700 mask_tok=141.00 y_norm=48.0871±0.0072 +2026-05-10 20:05:28,069 - INFO - Step 13740 | mode=video loss=0.086914 diff=0.086914 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0302 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0895 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=10993 i_micro=2747 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4433 mask_tok=133.00 y_norm=48.0895±0.0070 +2026-05-10 20:05:46,034 - INFO - Step 13750 | mode=video loss=0.080566 diff=0.080566 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0544 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0892 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=11001 i_micro=2749 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.2533 mask_tok=76.00 y_norm=48.0892±0.0049 +2026-05-10 20:06:04,004 - INFO - Step 13760 | mode=video loss=0.078125 diff=0.078125 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0425 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0862 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=11009 i_micro=2751 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3833 mask_tok=115.00 y_norm=48.0862±0.0066 +2026-05-10 20:06:23,785 - INFO - Step 13770 | mode=video loss=0.126953 diff=0.126953 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0659 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0884 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=11017 i_micro=2753 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3467 mask_tok=104.00 y_norm=48.0884±0.0057 +2026-05-10 20:06:41,746 - INFO - Step 13780 | mode=video loss=0.094727 diff=0.094727 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0673 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0879 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=11025 i_micro=2755 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3467 mask_tok=104.00 y_norm=48.0879±0.0053 +2026-05-10 20:07:00,895 - INFO - Step 13790 | mode=video loss=0.081055 diff=0.081055 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0931 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0878 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=11033 i_micro=2757 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3400 mask_tok=102.00 y_norm=48.0878±0.0058 +2026-05-10 20:07:20,485 - INFO - Step 13800 | mode=video loss=0.096191 diff=0.096191 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0620 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0885 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=11041 i_micro=2759 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3333 mask_tok=100.00 y_norm=48.0885±0.0053 cond_uncond_dloss=0.294922 cond_grad=0.000099 cond_pred_l2=36.112148 cond_pred_ratio=0.021750 +2026-05-10 20:07:20,523 - INFO - Step 13800 | probe_semantic mcp_offdiag(mean/min/max)=0.646046/0.532907/0.707893 smol_offdiag=0.660557 prompts=4 mcp_tok=17.50 smol_tok=17.50 +2026-05-10 20:07:27,493 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 20:07:45,469 - INFO - Step 13810 | mode=video loss=0.061523 diff=0.061523 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0307 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0867 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=11049 i_micro=2761 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3367 mask_tok=101.00 y_norm=48.0867±0.0063 +2026-05-10 20:08:05,269 - INFO - Step 13820 | mode=video loss=0.112305 diff=0.112305 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0457 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0857 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=11057 i_micro=2763 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4267 mask_tok=128.00 y_norm=48.0857±0.0065 +2026-05-10 20:08:23,241 - INFO - Step 13830 | mode=video loss=0.068359 diff=0.068359 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0740 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0867 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=11065 i_micro=2765 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4167 mask_tok=125.00 y_norm=48.0867±0.0061 +2026-05-10 20:08:41,256 - INFO - Step 13840 | mode=video loss=0.094238 diff=0.094238 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0401 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0891 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=11073 i_micro=2767 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3667 mask_tok=110.00 y_norm=48.0891±0.0058 +2026-05-10 20:09:00,147 - INFO - Step 13850 | mode=video loss=0.198242 diff=0.198242 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0950 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0874 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=11081 i_micro=2769 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3333 mask_tok=100.00 y_norm=48.0874±0.0058 +2026-05-10 20:09:18,129 - INFO - Step 13860 | mode=video loss=0.033203 diff=0.033203 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0600 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0871 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=11089 i_micro=2771 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.9000 mask_tok=270.00 y_norm=48.0871±0.0083 +2026-05-10 20:09:36,813 - INFO - Step 13870 | mode=video loss=0.063965 diff=0.063965 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0294 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0885 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=11097 i_micro=2773 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3333 mask_tok=100.00 y_norm=48.0885±0.0054 +2026-05-10 20:09:54,792 - INFO - Step 13880 | mode=video loss=0.118164 diff=0.118164 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0544 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0891 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=11105 i_micro=2775 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3133 mask_tok=94.00 y_norm=48.0891±0.0059 +2026-05-10 20:10:12,771 - INFO - Step 13890 | mode=video loss=0.075195 diff=0.075195 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0286 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0894 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=11113 i_micro=2777 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3933 mask_tok=118.00 y_norm=48.0894±0.0058 +2026-05-10 20:10:31,438 - INFO - Step 13900 | mode=video loss=0.077148 diff=0.077148 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0501 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0882 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=11121 i_micro=2779 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3400 mask_tok=102.00 y_norm=48.0882±0.0054 +2026-05-10 20:10:38,287 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 20:10:56,244 - INFO - Step 13910 | mode=video loss=0.093262 diff=0.093262 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0284 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0862 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=11129 i_micro=2781 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4200 mask_tok=126.00 y_norm=48.0862±0.0063 +2026-05-10 20:11:14,916 - INFO - Step 13920 | mode=video loss=0.052246 diff=0.052246 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0331 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0881 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=11137 i_micro=2783 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3233 mask_tok=97.00 y_norm=48.0881±0.0060 +2026-05-10 20:11:32,892 - INFO - Step 13930 | mode=video loss=0.068848 diff=0.068848 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1269 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0861 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=11145 i_micro=2785 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4067 mask_tok=122.00 y_norm=48.0861±0.0059 +2026-05-10 20:11:50,877 - INFO - Step 13940 | mode=video loss=0.096191 diff=0.096191 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0541 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0871 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=11153 i_micro=2787 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.5100 mask_tok=153.00 y_norm=48.0871±0.0063 +2026-05-10 20:12:09,938 - INFO - Step 13950 | mode=video loss=0.099121 diff=0.099121 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0721 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0874 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=11161 i_micro=2789 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3167 mask_tok=95.00 y_norm=48.0874±0.0062 +2026-05-10 20:12:27,931 - INFO - Step 13960 | mode=video loss=0.056641 diff=0.056641 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0258 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0879 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=11169 i_micro=2791 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3400 mask_tok=102.00 y_norm=48.0879±0.0062 +2026-05-10 20:12:46,477 - INFO - Step 13970 | mode=video loss=0.136719 diff=0.136719 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0445 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0868 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=11177 i_micro=2793 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3933 mask_tok=118.00 y_norm=48.0868±0.0062 +2026-05-10 20:13:04,980 - INFO - Step 13980 | mode=video loss=0.079102 diff=0.079102 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0317 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0864 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=11185 i_micro=2795 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3700 mask_tok=111.00 y_norm=48.0864±0.0057 +2026-05-10 20:13:22,951 - INFO - Step 13990 | mode=video loss=0.100098 diff=0.100098 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0653 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0877 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=11193 i_micro=2797 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3833 mask_tok=115.00 y_norm=48.0877±0.0064 +2026-05-10 20:13:43,859 - INFO - Step 14000 | mode=video loss=0.076172 diff=0.076172 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0408 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0871 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=11201 i_micro=2799 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3267 mask_tok=98.00 y_norm=48.0871±0.0058 cond_uncond_dloss=0.718750 cond_grad=0.000074 cond_pred_l2=46.414028 cond_pred_ratio=0.027774 +2026-05-10 20:13:43,898 - INFO - Step 14000 | probe_semantic mcp_offdiag(mean/min/max)=0.646383/0.533286/0.708277 smol_offdiag=0.660557 prompts=4 mcp_tok=17.50 smol_tok=17.50 +2026-05-10 20:13:50,708 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 20:14:08,694 - INFO - Step 14010 | mode=video loss=0.071777 diff=0.071777 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0255 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0881 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=11209 i_micro=2801 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3667 mask_tok=110.00 y_norm=48.0881±0.0061 +2026-05-10 20:14:26,682 - INFO - Step 14020 | mode=video loss=0.044678 diff=0.044678 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0714 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0852 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=11217 i_micro=2803 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4233 mask_tok=127.00 y_norm=48.0852±0.0060 +2026-05-10 20:14:46,170 - INFO - Step 14030 | mode=video loss=0.065430 diff=0.065430 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0439 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0882 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=11225 i_micro=2805 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3400 mask_tok=102.00 y_norm=48.0882±0.0052 +2026-05-10 20:15:04,155 - INFO - Step 14040 | mode=video loss=0.139648 diff=0.139648 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0903 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0895 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=11233 i_micro=2807 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3467 mask_tok=104.00 y_norm=48.0895±0.0054 +2026-05-10 20:15:23,505 - INFO - Step 14050 | mode=video loss=0.080078 diff=0.080078 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0384 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0873 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=11241 i_micro=2809 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4767 mask_tok=143.00 y_norm=48.0873±0.0069 +2026-05-10 20:15:41,496 - INFO - Step 14060 | mode=video loss=0.094727 diff=0.094727 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0269 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0863 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=11249 i_micro=2811 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.5400 mask_tok=162.00 y_norm=48.0863±0.0076 +2026-05-10 20:15:59,488 - INFO - Step 14070 | mode=video loss=0.066406 diff=0.066406 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0326 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0879 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=11257 i_micro=2813 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3000 mask_tok=90.00 y_norm=48.0879±0.0057 +2026-05-10 20:16:18,758 - INFO - Step 14080 | mode=video loss=0.085449 diff=0.085449 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0530 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0879 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=11265 i_micro=2815 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3067 mask_tok=92.00 y_norm=48.0879±0.0055 +2026-05-10 20:16:36,740 - INFO - Step 14090 | mode=video loss=0.104980 diff=0.104980 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0426 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0868 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=11273 i_micro=2817 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4767 mask_tok=143.00 y_norm=48.0868±0.0065 +2026-05-10 20:16:55,217 - INFO - Step 14100 | mode=video loss=0.111328 diff=0.111328 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0691 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0882 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=11281 i_micro=2819 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3400 mask_tok=102.00 y_norm=48.0882±0.0049 +2026-05-10 20:17:02,299 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 20:17:21,076 - INFO - Step 14110 | mode=video loss=0.083984 diff=0.083984 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0321 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0871 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=11289 i_micro=2821 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3533 mask_tok=106.00 y_norm=48.0871±0.0071 +2026-05-10 20:17:39,057 - INFO - Step 14120 | mode=video loss=0.140625 diff=0.140625 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0419 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0898 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=11297 i_micro=2823 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.2900 mask_tok=87.00 y_norm=48.0898±0.0053 +2026-05-10 20:17:58,611 - INFO - Step 14130 | mode=video loss=0.077148 diff=0.077148 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0260 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0885 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=11305 i_micro=2825 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3267 mask_tok=98.00 y_norm=48.0885±0.0049 +2026-05-10 20:18:16,607 - INFO - Step 14140 | mode=video loss=0.200195 diff=0.200195 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0632 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0881 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=11313 i_micro=2827 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3567 mask_tok=107.00 y_norm=48.0881±0.0062 +2026-05-10 20:18:34,584 - INFO - Step 14150 | mode=video loss=0.113770 diff=0.113770 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0706 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0879 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=11321 i_micro=2829 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3133 mask_tok=94.00 y_norm=48.0879±0.0054 +2026-05-10 20:18:53,735 - INFO - Step 14160 | mode=video loss=0.069824 diff=0.069824 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0842 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0872 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=11329 i_micro=2831 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3400 mask_tok=102.00 y_norm=48.0872±0.0054 +2026-05-10 20:19:11,719 - INFO - Step 14170 | mode=video loss=0.074707 diff=0.074707 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1511 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0851 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=11337 i_micro=2833 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4333 mask_tok=130.00 y_norm=48.0851±0.0060 +2026-05-10 20:19:30,805 - INFO - Step 14180 | mode=video loss=0.084473 diff=0.084473 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0732 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0892 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=11345 i_micro=2835 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3600 mask_tok=108.00 y_norm=48.0892±0.0060 +2026-05-10 20:19:48,771 - INFO - Step 14190 | mode=video loss=0.136719 diff=0.136719 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0744 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0844 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=11353 i_micro=2837 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4100 mask_tok=123.00 y_norm=48.0844±0.0067 +2026-05-10 20:20:08,361 - INFO - Step 14200 | mode=video loss=0.092773 diff=0.092773 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0432 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0900 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=11361 i_micro=2839 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3533 mask_tok=106.00 y_norm=48.0900±0.0055 cond_uncond_dloss=0.484375 cond_grad=0.000189 cond_pred_l2=46.638943 cond_pred_ratio=0.029037 +2026-05-10 20:20:08,399 - INFO - Step 14200 | probe_semantic mcp_offdiag(mean/min/max)=0.646456/0.533416/0.708532 smol_offdiag=0.660557 prompts=4 mcp_tok=17.50 smol_tok=17.50 +2026-05-10 20:20:15,294 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 20:20:34,400 - INFO - Step 14210 | mode=video loss=0.077148 diff=0.077148 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0583 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0871 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=11369 i_micro=2841 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3400 mask_tok=102.00 y_norm=48.0871±0.0045 +2026-05-10 20:20:52,390 - INFO - Step 14220 | mode=video loss=0.102051 diff=0.102051 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1126 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0891 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=11377 i_micro=2843 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3433 mask_tok=103.00 y_norm=48.0891±0.0051 +2026-05-10 20:21:11,470 - INFO - Step 14230 | mode=video loss=0.075195 diff=0.075195 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0287 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0880 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=11385 i_micro=2845 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3367 mask_tok=101.00 y_norm=48.0880±0.0049 +2026-05-10 20:21:29,464 - INFO - Step 14240 | mode=video loss=0.084473 diff=0.084473 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0290 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0875 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=11393 i_micro=2847 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4433 mask_tok=133.00 y_norm=48.0875±0.0068 +2026-05-10 20:21:47,454 - INFO - Step 14250 | mode=video loss=0.047119 diff=0.047119 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0527 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0886 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=11401 i_micro=2849 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3267 mask_tok=98.00 y_norm=48.0886±0.0053 +2026-05-10 20:22:06,600 - INFO - Step 14260 | mode=video loss=0.089355 diff=0.089355 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0648 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0867 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=11409 i_micro=2851 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4800 mask_tok=144.00 y_norm=48.0867±0.0059 +2026-05-10 20:22:24,584 - INFO - Step 14270 | mode=video loss=0.105469 diff=0.105469 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0340 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0864 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=11417 i_micro=2853 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4233 mask_tok=127.00 y_norm=48.0864±0.0061 +2026-05-10 20:22:42,999 - INFO - Step 14280 | mode=video loss=0.097168 diff=0.097168 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0587 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0880 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=11425 i_micro=2855 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3633 mask_tok=109.00 y_norm=48.0880±0.0052 +2026-05-10 20:23:01,722 - INFO - Step 14290 | mode=video loss=0.085938 diff=0.085938 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0390 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0884 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=11433 i_micro=2857 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3533 mask_tok=106.00 y_norm=48.0884±0.0058 +2026-05-10 20:23:19,715 - INFO - Step 14300 | mode=video loss=0.148438 diff=0.148438 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0364 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0862 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=11441 i_micro=2859 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4333 mask_tok=130.00 y_norm=48.0862±0.0058 +2026-05-10 20:23:26,750 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 20:23:45,827 - INFO - Step 14310 | mode=video loss=0.072754 diff=0.072754 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0688 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0892 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=11449 i_micro=2861 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3533 mask_tok=106.00 y_norm=48.0892±0.0061 +2026-05-10 20:24:03,800 - INFO - Step 14320 | mode=video loss=0.083008 diff=0.083008 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0570 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0863 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=11457 i_micro=2863 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3467 mask_tok=104.00 y_norm=48.0863±0.0068 +2026-05-10 20:24:21,778 - INFO - Step 14330 | mode=video loss=0.151367 diff=0.151367 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0550 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0896 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=11465 i_micro=2865 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3400 mask_tok=102.00 y_norm=48.0896±0.0052 +2026-05-10 20:24:40,798 - INFO - Step 14340 | mode=video loss=0.104980 diff=0.104980 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0416 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0857 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=11473 i_micro=2867 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4333 mask_tok=130.00 y_norm=48.0857±0.0063 +2026-05-10 20:24:58,773 - INFO - Step 14350 | mode=video loss=0.131836 diff=0.131836 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0507 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0859 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=11481 i_micro=2869 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4200 mask_tok=126.00 y_norm=48.0859±0.0060 +2026-05-10 20:25:17,797 - INFO - Step 14360 | mode=video loss=0.106445 diff=0.106445 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0612 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0876 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=11489 i_micro=2871 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3733 mask_tok=112.00 y_norm=48.0876±0.0055 +2026-05-10 20:25:35,758 - INFO - Step 14370 | mode=video loss=0.082031 diff=0.082031 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0445 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0887 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=11497 i_micro=2873 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3333 mask_tok=100.00 y_norm=48.0887±0.0053 +2026-05-10 20:25:53,720 - INFO - Step 14380 | mode=video loss=0.075684 diff=0.075684 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0376 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0881 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=11505 i_micro=2875 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3033 mask_tok=91.00 y_norm=48.0881±0.0050 +2026-05-10 20:26:12,697 - INFO - Step 14390 | mode=video loss=0.139648 diff=0.139648 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0410 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0864 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=11513 i_micro=2877 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.5200 mask_tok=156.00 y_norm=48.0864±0.0066 +2026-05-10 20:26:32,295 - INFO - Step 14400 | mode=video loss=0.083008 diff=0.083008 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0342 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0871 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=11521 i_micro=2879 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3067 mask_tok=92.00 y_norm=48.0871±0.0052 cond_uncond_dloss=0.429688 cond_grad=0.000101 cond_pred_l2=40.656528 cond_pred_ratio=0.024485 +2026-05-10 20:26:32,333 - INFO - Step 14400 | probe_semantic mcp_offdiag(mean/min/max)=0.645982/0.532789/0.708212 smol_offdiag=0.660557 prompts=4 mcp_tok=17.50 smol_tok=17.50 +2026-05-10 20:26:39,263 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 20:26:57,676 - INFO - Step 14410 | mode=video loss=0.102051 diff=0.102051 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0430 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0898 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=11529 i_micro=2881 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3600 mask_tok=108.00 y_norm=48.0898±0.0056 +2026-05-10 20:27:16,393 - INFO - Step 14420 | mode=video loss=0.061279 diff=0.061279 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0434 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0860 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=11537 i_micro=2883 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4500 mask_tok=135.00 y_norm=48.0860±0.0062 +2026-05-10 20:27:34,384 - INFO - Step 14430 | mode=video loss=0.092285 diff=0.092285 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0787 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0858 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=11545 i_micro=2885 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4267 mask_tok=128.00 y_norm=48.0858±0.0058 +2026-05-10 20:27:53,618 - INFO - Step 14440 | mode=video loss=0.076660 diff=0.076660 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0474 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0875 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=11553 i_micro=2887 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3700 mask_tok=111.00 y_norm=48.0875±0.0060 +2026-05-10 20:28:11,590 - INFO - Step 14450 | mode=video loss=0.082520 diff=0.082520 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0716 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0883 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=11561 i_micro=2889 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3433 mask_tok=103.00 y_norm=48.0883±0.0050 +2026-05-10 20:28:29,556 - INFO - Step 14460 | mode=video loss=0.072266 diff=0.072266 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0648 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0873 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=11569 i_micro=2891 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.5233 mask_tok=157.00 y_norm=48.0873±0.0072 +2026-05-10 20:28:48,719 - INFO - Step 14470 | mode=video loss=0.061768 diff=0.061768 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0365 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0880 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=11577 i_micro=2893 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3267 mask_tok=98.00 y_norm=48.0880±0.0050 +2026-05-10 20:29:06,674 - INFO - Step 14480 | mode=video loss=0.067383 diff=0.067383 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0407 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0879 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=11585 i_micro=2895 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4600 mask_tok=138.00 y_norm=48.0879±0.0065 +2026-05-10 20:29:26,031 - INFO - Step 14490 | mode=video loss=0.065430 diff=0.065430 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0290 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0886 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=11593 i_micro=2897 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.2833 mask_tok=85.00 y_norm=48.0886±0.0049 +2026-05-10 20:29:43,987 - INFO - Step 14500 | mode=video loss=0.101562 diff=0.101562 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0544 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0882 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=11601 i_micro=2899 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3633 mask_tok=109.00 y_norm=48.0882±0.0061 +2026-05-10 20:29:50,831 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 20:30:08,774 - INFO - Step 14510 | mode=video loss=0.072754 diff=0.072754 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0417 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0881 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=11609 i_micro=2901 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3300 mask_tok=99.00 y_norm=48.0881±0.0050 +2026-05-10 20:30:27,998 - INFO - Step 14520 | mode=video loss=0.083008 diff=0.083008 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0536 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0872 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=11617 i_micro=2903 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3600 mask_tok=108.00 y_norm=48.0872±0.0055 +2026-05-10 20:30:45,973 - INFO - Step 14530 | mode=video loss=0.085938 diff=0.085938 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0663 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0858 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=11625 i_micro=2905 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4033 mask_tok=121.00 y_norm=48.0858±0.0062 +2026-05-10 20:31:04,403 - INFO - Step 14540 | mode=video loss=0.023438 diff=0.023438 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0507 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0886 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=11633 i_micro=2907 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4733 mask_tok=142.00 y_norm=48.0886±0.0067 +2026-05-10 20:31:23,122 - INFO - Step 14550 | mode=video loss=0.074707 diff=0.074707 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0244 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0882 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=11641 i_micro=2909 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.7667 mask_tok=230.00 y_norm=48.0882±0.0075 +2026-05-10 20:31:41,098 - INFO - Step 14560 | mode=video loss=0.090332 diff=0.090332 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1051 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0877 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=11649 i_micro=2911 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4600 mask_tok=138.00 y_norm=48.0877±0.0063 +2026-05-10 20:32:00,291 - INFO - Step 14570 | mode=video loss=0.085938 diff=0.085938 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0714 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0870 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=11657 i_micro=2913 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3833 mask_tok=115.00 y_norm=48.0870±0.0060 +2026-05-10 20:32:18,253 - INFO - Step 14580 | mode=video loss=0.084473 diff=0.084473 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1068 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0879 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=11665 i_micro=2915 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3300 mask_tok=99.00 y_norm=48.0879±0.0057 +2026-05-10 20:32:36,217 - INFO - Step 14590 | mode=video loss=0.129883 diff=0.129883 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0748 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0873 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=11673 i_micro=2917 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3000 mask_tok=90.00 y_norm=48.0873±0.0058 +2026-05-10 20:32:57,658 - INFO - Step 14600 | mode=video loss=0.146484 diff=0.146484 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0540 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0893 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=11681 i_micro=2919 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3300 mask_tok=99.00 y_norm=48.0893±0.0057 cond_uncond_dloss=0.451172 cond_grad=0.000061 cond_pred_l2=28.297897 cond_pred_ratio=0.016975 +2026-05-10 20:32:57,697 - INFO - Step 14600 | probe_semantic mcp_offdiag(mean/min/max)=0.645753/0.532399/0.707899 smol_offdiag=0.660557 prompts=4 mcp_tok=17.50 smol_tok=17.50 +2026-05-10 20:33:04,666 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 20:33:22,626 - INFO - Step 14610 | mode=video loss=0.118164 diff=0.118164 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0759 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0870 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=11689 i_micro=2921 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3467 mask_tok=104.00 y_norm=48.0870±0.0056 +2026-05-10 20:33:42,019 - INFO - Step 14620 | mode=video loss=0.107910 diff=0.107910 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0886 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0892 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=11697 i_micro=2923 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4367 mask_tok=131.00 y_norm=48.0892±0.0059 +2026-05-10 20:33:59,986 - INFO - Step 14630 | mode=video loss=0.082031 diff=0.082031 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0387 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0870 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=11705 i_micro=2925 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4567 mask_tok=137.00 y_norm=48.0870±0.0066 +2026-05-10 20:34:17,958 - INFO - Step 14640 | mode=video loss=0.167969 diff=0.167969 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0428 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0896 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=11713 i_micro=2927 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3700 mask_tok=111.00 y_norm=48.0896±0.0056 +2026-05-10 20:34:37,218 - INFO - Step 14650 | mode=video loss=0.077148 diff=0.077148 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0376 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0860 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=11721 i_micro=2929 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4067 mask_tok=122.00 y_norm=48.0860±0.0060 +2026-05-10 20:34:55,193 - INFO - Step 14660 | mode=video loss=0.078613 diff=0.078613 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0294 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0873 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=11729 i_micro=2931 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4267 mask_tok=128.00 y_norm=48.0873±0.0067 +2026-05-10 20:35:13,637 - INFO - Step 14670 | mode=video loss=0.090820 diff=0.090820 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0325 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0888 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=11737 i_micro=2933 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3533 mask_tok=106.00 y_norm=48.0888±0.0057 +2026-05-10 20:35:32,218 - INFO - Step 14680 | mode=video loss=0.077148 diff=0.077148 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0417 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0859 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=11745 i_micro=2935 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4067 mask_tok=122.00 y_norm=48.0859±0.0063 +2026-05-10 20:35:50,183 - INFO - Step 14690 | mode=video loss=0.149414 diff=0.149414 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0993 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0879 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=11753 i_micro=2937 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3467 mask_tok=104.00 y_norm=48.0879±0.0056 +2026-05-10 20:36:09,319 - INFO - Step 14700 | mode=video loss=0.094727 diff=0.094727 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0488 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0874 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=11761 i_micro=2939 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3500 mask_tok=105.00 y_norm=48.0874±0.0058 +2026-05-10 20:36:16,498 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 20:36:34,459 - INFO - Step 14710 | mode=video loss=0.118164 diff=0.118164 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0955 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0879 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=11769 i_micro=2941 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3567 mask_tok=107.00 y_norm=48.0879±0.0054 +2026-05-10 20:36:52,423 - INFO - Step 14720 | mode=video loss=0.080566 diff=0.080566 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0440 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0863 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=11777 i_micro=2943 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4067 mask_tok=122.00 y_norm=48.0863±0.0063 +2026-05-10 20:37:11,676 - INFO - Step 14730 | mode=video loss=0.071777 diff=0.071777 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0773 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0877 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=11785 i_micro=2945 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4833 mask_tok=145.00 y_norm=48.0877±0.0063 +2026-05-10 20:37:29,626 - INFO - Step 14740 | mode=video loss=0.079102 diff=0.079102 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0815 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0870 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=11793 i_micro=2947 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4067 mask_tok=122.00 y_norm=48.0870±0.0064 +2026-05-10 20:37:48,766 - INFO - Step 14750 | mode=video loss=0.124023 diff=0.124023 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1174 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0880 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=11801 i_micro=2949 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3367 mask_tok=101.00 y_norm=48.0880±0.0054 +2026-05-10 20:38:06,724 - INFO - Step 14760 | mode=video loss=0.102051 diff=0.102051 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1436 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0878 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=11809 i_micro=2951 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3500 mask_tok=105.00 y_norm=48.0878±0.0054 +2026-05-10 20:38:24,677 - INFO - Step 14770 | mode=video loss=0.090820 diff=0.090820 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0494 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0860 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=11817 i_micro=2953 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3600 mask_tok=108.00 y_norm=48.0860±0.0057 +2026-05-10 20:38:43,783 - INFO - Step 14780 | mode=video loss=0.080078 diff=0.080078 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0441 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0890 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=11825 i_micro=2955 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3533 mask_tok=106.00 y_norm=48.0890±0.0065 +2026-05-10 20:39:01,757 - INFO - Step 14790 | mode=video loss=0.027466 diff=0.027466 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0559 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0857 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=11833 i_micro=2957 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4967 mask_tok=149.00 y_norm=48.0857±0.0069 +2026-05-10 20:39:21,841 - INFO - Step 14800 | mode=video loss=0.091309 diff=0.091309 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0333 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0873 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=11841 i_micro=2959 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3300 mask_tok=99.00 y_norm=48.0873±0.0058 cond_uncond_dloss=0.263672 cond_grad=0.000063 cond_pred_l2=33.399113 cond_pred_ratio=0.019174 +2026-05-10 20:39:21,879 - INFO - Step 14800 | probe_semantic mcp_offdiag(mean/min/max)=0.645668/0.532253/0.708064 smol_offdiag=0.660557 prompts=4 mcp_tok=17.50 smol_tok=17.50 +2026-05-10 20:39:29,127 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 20:39:47,774 - INFO - Step 14810 | mode=video loss=0.077637 diff=0.077637 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0730 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0883 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=11849 i_micro=2961 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3467 mask_tok=104.00 y_norm=48.0883±0.0052 +2026-05-10 20:40:05,751 - INFO - Step 14820 | mode=video loss=0.095215 diff=0.095215 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0957 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0863 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=11857 i_micro=2963 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3333 mask_tok=100.00 y_norm=48.0863±0.0057 +2026-05-10 20:40:25,348 - INFO - Step 14830 | mode=video loss=0.102539 diff=0.102539 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1513 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0866 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=11865 i_micro=2965 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3633 mask_tok=109.00 y_norm=48.0866±0.0060 +2026-05-10 20:40:43,323 - INFO - Step 14840 | mode=video loss=0.124512 diff=0.124512 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0665 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0890 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=11873 i_micro=2967 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3400 mask_tok=102.00 y_norm=48.0890±0.0051 +2026-05-10 20:41:01,305 - INFO - Step 14850 | mode=video loss=0.075684 diff=0.075684 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0248 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0883 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=11881 i_micro=2969 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3367 mask_tok=101.00 y_norm=48.0883±0.0051 +2026-05-10 20:41:20,580 - INFO - Step 14860 | mode=video loss=0.079590 diff=0.079590 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0302 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0881 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=11889 i_micro=2971 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3600 mask_tok=108.00 y_norm=48.0881±0.0059 +2026-05-10 20:41:38,557 - INFO - Step 14870 | mode=video loss=0.097656 diff=0.097656 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0595 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0851 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=11897 i_micro=2973 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4500 mask_tok=135.00 y_norm=48.0851±0.0059 +2026-05-10 20:41:58,167 - INFO - Step 14880 | mode=video loss=0.062012 diff=0.062012 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0511 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0880 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=11905 i_micro=2975 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3400 mask_tok=102.00 y_norm=48.0880±0.0060 +2026-05-10 20:42:16,129 - INFO - Step 14890 | mode=video loss=0.067871 diff=0.067871 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0403 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0873 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=11913 i_micro=2977 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3400 mask_tok=102.00 y_norm=48.0873±0.0063 +2026-05-10 20:42:34,085 - INFO - Step 14900 | mode=video loss=0.107422 diff=0.107422 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0374 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0878 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=11921 i_micro=2979 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3433 mask_tok=103.00 y_norm=48.0878±0.0057 +2026-05-10 20:42:40,872 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 20:43:00,272 - INFO - Step 14910 | mode=video loss=0.093750 diff=0.093750 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0276 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0879 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=11929 i_micro=2981 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3600 mask_tok=108.00 y_norm=48.0879±0.0068 +2026-05-10 20:43:18,232 - INFO - Step 14920 | mode=video loss=0.110352 diff=0.110352 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0596 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0854 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=11937 i_micro=2983 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4033 mask_tok=121.00 y_norm=48.0854±0.0057 +2026-05-10 20:43:36,663 - INFO - Step 14930 | mode=video loss=0.068848 diff=0.068848 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1067 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0887 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=11945 i_micro=2985 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3400 mask_tok=102.00 y_norm=48.0887±0.0052 +2026-05-10 20:43:55,576 - INFO - Step 14940 | mode=video loss=0.099121 diff=0.099121 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0358 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0862 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=11953 i_micro=2987 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4500 mask_tok=135.00 y_norm=48.0862±0.0069 +2026-05-10 20:44:13,550 - INFO - Step 14950 | mode=video loss=0.039551 diff=0.039551 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0509 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0852 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=11961 i_micro=2989 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3533 mask_tok=106.00 y_norm=48.0852±0.0061 +2026-05-10 20:44:32,612 - INFO - Step 14960 | mode=video loss=0.085449 diff=0.085449 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0302 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0865 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=11969 i_micro=2991 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3500 mask_tok=105.00 y_norm=48.0865±0.0063 +2026-05-10 20:44:50,580 - INFO - Step 14970 | mode=video loss=0.063477 diff=0.063477 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0338 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0871 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=11977 i_micro=2993 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3200 mask_tok=96.00 y_norm=48.0871±0.0050 +2026-05-10 20:45:08,544 - INFO - Step 14980 | mode=video loss=0.111328 diff=0.111328 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0851 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0865 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=11985 i_micro=2995 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3633 mask_tok=109.00 y_norm=48.0865±0.0057 +2026-05-10 20:45:27,604 - INFO - Step 14990 | mode=video loss=0.034668 diff=0.034668 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0876 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0853 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=11993 i_micro=2997 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3800 mask_tok=114.00 y_norm=48.0853±0.0058 +2026-05-10 20:45:47,231 - INFO - Step 15000 | mode=video loss=0.109863 diff=0.109863 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0884 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0875 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=12001 i_micro=2999 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3200 mask_tok=96.00 y_norm=48.0875±0.0053 cond_uncond_dloss=0.656250 cond_grad=0.000360 cond_pred_l2=57.865082 cond_pred_ratio=0.036588 +2026-05-10 20:45:47,268 - INFO - Step 15000 | probe_semantic mcp_offdiag(mean/min/max)=0.645874/0.532631/0.708151 smol_offdiag=0.660557 prompts=4 mcp_tok=17.50 smol_tok=17.50 +2026-05-10 20:45:54,347 - INFO - Saved checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_step15000.pt +2026-05-10 20:45:59,169 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 20:46:18,370 - INFO - Step 15010 | mode=video loss=0.147461 diff=0.147461 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0314 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0866 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=12009 i_micro=3001 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3567 mask_tok=107.00 y_norm=48.0866±0.0063 +2026-05-10 20:46:36,323 - INFO - Step 15020 | mode=video loss=0.108887 diff=0.108887 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0424 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0858 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=12017 i_micro=3003 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4767 mask_tok=143.00 y_norm=48.0858±0.0066 +2026-05-10 20:46:54,290 - INFO - Step 15030 | mode=video loss=0.094238 diff=0.094238 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0377 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0883 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=12025 i_micro=3005 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.2933 mask_tok=88.00 y_norm=48.0883±0.0056 +2026-05-10 20:47:13,434 - INFO - Step 15040 | mode=video loss=0.090820 diff=0.090820 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0505 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0878 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=12033 i_micro=3007 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3600 mask_tok=108.00 y_norm=48.0878±0.0070 +2026-05-10 20:47:31,398 - INFO - Step 15050 | mode=video loss=0.120605 diff=0.120605 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0467 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0868 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=12041 i_micro=3009 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3500 mask_tok=105.00 y_norm=48.0868±0.0056 +2026-05-10 20:47:49,807 - INFO - Step 15060 | mode=video loss=0.094238 diff=0.094238 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0543 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0887 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=12049 i_micro=3011 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3100 mask_tok=93.00 y_norm=48.0887±0.0052 +2026-05-10 20:48:08,548 - INFO - Step 15070 | mode=video loss=0.098145 diff=0.098145 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0535 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0878 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=12057 i_micro=3013 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3400 mask_tok=102.00 y_norm=48.0878±0.0060 +2026-05-10 20:48:26,509 - INFO - Step 15080 | mode=video loss=0.090332 diff=0.090332 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1196 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0861 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=12065 i_micro=3015 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3500 mask_tok=105.00 y_norm=48.0861±0.0053 +2026-05-10 20:48:45,655 - INFO - Step 15090 | mode=video loss=0.161133 diff=0.161133 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1442 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0865 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=12073 i_micro=3017 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.5067 mask_tok=152.00 y_norm=48.0865±0.0068 +2026-05-10 20:49:03,612 - INFO - Step 15100 | mode=video loss=0.093750 diff=0.093750 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0471 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0863 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=12081 i_micro=3019 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4567 mask_tok=137.00 y_norm=48.0863±0.0066 +2026-05-10 20:49:10,521 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 20:49:28,473 - INFO - Step 15110 | mode=video loss=0.110840 diff=0.110840 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0336 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0860 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=12089 i_micro=3021 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4700 mask_tok=141.00 y_norm=48.0860±0.0061 +2026-05-10 20:49:47,661 - INFO - Step 15120 | mode=video loss=0.124023 diff=0.124023 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0598 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0885 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=12097 i_micro=3023 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3567 mask_tok=107.00 y_norm=48.0885±0.0059 +2026-05-10 20:50:05,660 - INFO - Step 15130 | mode=video loss=0.105957 diff=0.105957 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0376 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0864 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=12105 i_micro=3025 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3433 mask_tok=103.00 y_norm=48.0864±0.0062 +2026-05-10 20:50:24,702 - INFO - Step 15140 | mode=video loss=0.109863 diff=0.109863 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0586 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0907 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=12113 i_micro=3027 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.5500 mask_tok=165.00 y_norm=48.0907±0.0073 +2026-05-10 20:50:42,676 - INFO - Step 15150 | mode=video loss=0.071289 diff=0.071289 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0615 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0876 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=12121 i_micro=3029 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3567 mask_tok=107.00 y_norm=48.0876±0.0066 +2026-05-10 20:51:00,666 - INFO - Step 15160 | mode=video loss=0.067871 diff=0.067871 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0334 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0883 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=12129 i_micro=3031 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3267 mask_tok=98.00 y_norm=48.0883±0.0052 +2026-05-10 20:51:19,795 - INFO - Step 15170 | mode=video loss=0.114258 diff=0.114258 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0617 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0867 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=12137 i_micro=3033 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3367 mask_tok=101.00 y_norm=48.0867±0.0052 +2026-05-10 20:51:37,770 - INFO - Step 15180 | mode=video loss=0.115723 diff=0.115723 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0471 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0877 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=12145 i_micro=3035 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3433 mask_tok=103.00 y_norm=48.0877±0.0060 +2026-05-10 20:51:56,166 - INFO - Step 15190 | mode=video loss=0.087891 diff=0.087891 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1313 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0861 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=12153 i_micro=3037 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3933 mask_tok=118.00 y_norm=48.0861±0.0059 +2026-05-10 20:52:16,512 - INFO - Step 15200 | mode=video loss=0.079102 diff=0.079102 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0295 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0875 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=12161 i_micro=3039 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3433 mask_tok=103.00 y_norm=48.0875±0.0054 cond_uncond_dloss=0.312500 cond_grad=0.000132 cond_pred_l2=47.534073 cond_pred_ratio=0.029374 +2026-05-10 20:52:16,549 - INFO - Step 15200 | probe_semantic mcp_offdiag(mean/min/max)=0.645668/0.532481/0.707859 smol_offdiag=0.660557 prompts=4 mcp_tok=17.50 smol_tok=17.50 +2026-05-10 20:52:22,881 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 20:52:40,823 - INFO - Step 15210 | mode=video loss=0.245117 diff=0.245117 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0378 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0860 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=12169 i_micro=3041 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3900 mask_tok=117.00 y_norm=48.0860±0.0063 +2026-05-10 20:53:00,169 - INFO - Step 15220 | mode=video loss=0.092773 diff=0.092773 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0393 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0874 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=12177 i_micro=3043 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4167 mask_tok=125.00 y_norm=48.0874±0.0065 +2026-05-10 20:53:18,138 - INFO - Step 15230 | mode=video loss=0.103516 diff=0.103516 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0493 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0863 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=12185 i_micro=3045 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4667 mask_tok=140.00 y_norm=48.0863±0.0062 +2026-05-10 20:53:36,524 - INFO - Step 15240 | mode=video loss=0.098633 diff=0.098633 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1055 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0868 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=12193 i_micro=3047 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4667 mask_tok=140.00 y_norm=48.0868±0.0071 +2026-05-10 20:53:55,213 - INFO - Step 15250 | mode=video loss=0.156250 diff=0.156250 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0398 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0880 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=12201 i_micro=3049 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3300 mask_tok=99.00 y_norm=48.0880±0.0062 +2026-05-10 20:54:13,180 - INFO - Step 15260 | mode=video loss=0.095703 diff=0.095703 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0476 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0883 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=12209 i_micro=3051 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3700 mask_tok=111.00 y_norm=48.0883±0.0064 +2026-05-10 20:54:32,318 - INFO - Step 15270 | mode=video loss=0.050537 diff=0.050537 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0296 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0861 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=12217 i_micro=3053 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4267 mask_tok=128.00 y_norm=48.0861±0.0063 +2026-05-10 20:54:50,290 - INFO - Step 15280 | mode=video loss=0.069336 diff=0.069336 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1145 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0875 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=12225 i_micro=3055 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.2433 mask_tok=73.00 y_norm=48.0875±0.0054 +2026-05-10 20:55:08,262 - INFO - Step 15290 | mode=video loss=0.142578 diff=0.142578 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0468 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0864 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=12233 i_micro=3057 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4733 mask_tok=142.00 y_norm=48.0864±0.0073 +2026-05-10 20:55:27,420 - INFO - Step 15300 | mode=video loss=0.075684 diff=0.075684 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0304 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0869 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=12241 i_micro=3059 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3633 mask_tok=109.00 y_norm=48.0869±0.0052 +2026-05-10 20:55:34,365 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 20:55:52,315 - INFO - Step 15310 | mode=video loss=0.078613 diff=0.078613 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0341 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0879 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=12249 i_micro=3061 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3400 mask_tok=102.00 y_norm=48.0879±0.0054 +2026-05-10 20:56:11,471 - INFO - Step 15320 | mode=video loss=0.115234 diff=0.115234 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0542 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0878 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=12257 i_micro=3063 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3600 mask_tok=108.00 y_norm=48.0878±0.0055 +2026-05-10 20:56:29,440 - INFO - Step 15330 | mode=video loss=0.096680 diff=0.096680 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0470 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0873 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=12265 i_micro=3065 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3833 mask_tok=115.00 y_norm=48.0873±0.0052 +2026-05-10 20:56:47,414 - INFO - Step 15340 | mode=video loss=0.120605 diff=0.120605 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0692 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0880 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=12273 i_micro=3067 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3867 mask_tok=116.00 y_norm=48.0880±0.0070 +2026-05-10 20:57:06,636 - INFO - Step 15350 | mode=video loss=0.077148 diff=0.077148 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0624 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0875 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=12281 i_micro=3069 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3367 mask_tok=101.00 y_norm=48.0875±0.0059 +2026-05-10 20:57:24,616 - INFO - Step 15360 | mode=video loss=0.117188 diff=0.117188 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0432 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0863 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=12289 i_micro=3071 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3433 mask_tok=103.00 y_norm=48.0863±0.0051 +2026-05-10 20:57:43,011 - INFO - Step 15370 | mode=video loss=0.043457 diff=0.043457 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0629 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0866 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=12297 i_micro=3073 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4000 mask_tok=120.00 y_norm=48.0866±0.0071 +2026-05-10 20:58:01,843 - INFO - Step 15380 | mode=video loss=0.123535 diff=0.123535 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0526 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0878 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=12305 i_micro=3075 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3333 mask_tok=100.00 y_norm=48.0878±0.0052 +2026-05-10 20:58:19,804 - INFO - Step 15390 | mode=video loss=0.131836 diff=0.131836 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0528 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0867 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=12313 i_micro=3077 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3400 mask_tok=102.00 y_norm=48.0867±0.0052 +2026-05-10 20:58:40,638 - INFO - Step 15400 | mode=video loss=0.167969 diff=0.167969 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1114 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0878 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=12321 i_micro=3079 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4833 mask_tok=145.00 y_norm=48.0878±0.0063 cond_uncond_dloss=0.832031 cond_grad=0.001084 cond_pred_l2=50.880234 cond_pred_ratio=0.033238 +2026-05-10 20:58:40,675 - INFO - Step 15400 | probe_semantic mcp_offdiag(mean/min/max)=0.645792/0.532571/0.707851 smol_offdiag=0.660557 prompts=4 mcp_tok=17.50 smol_tok=17.50 +2026-05-10 20:58:47,449 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 20:59:05,415 - INFO - Step 15410 | mode=video loss=0.135742 diff=0.135742 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0290 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0868 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=12329 i_micro=3081 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3700 mask_tok=111.00 y_norm=48.0868±0.0058 +2026-05-10 20:59:23,394 - INFO - Step 15420 | mode=video loss=0.079102 diff=0.079102 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0327 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0883 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=12337 i_micro=3083 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3367 mask_tok=101.00 y_norm=48.0883±0.0054 +2026-05-10 20:59:42,651 - INFO - Step 15430 | mode=video loss=0.087402 diff=0.087402 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0544 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0874 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=12345 i_micro=3085 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4400 mask_tok=132.00 y_norm=48.0874±0.0063 +2026-05-10 21:00:00,627 - INFO - Step 15440 | mode=video loss=0.032715 diff=0.032715 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0405 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0867 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=12353 i_micro=3087 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4100 mask_tok=123.00 y_norm=48.0867±0.0056 +2026-05-10 21:00:19,851 - INFO - Step 15450 | mode=video loss=0.133789 diff=0.133789 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0392 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0884 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=12361 i_micro=3089 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3367 mask_tok=101.00 y_norm=48.0884±0.0054 +2026-05-10 21:00:37,815 - INFO - Step 15460 | mode=video loss=0.107422 diff=0.107422 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0408 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0870 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=12369 i_micro=3091 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3433 mask_tok=103.00 y_norm=48.0870±0.0063 +2026-05-10 21:00:55,781 - INFO - Step 15470 | mode=video loss=0.076172 diff=0.076172 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0953 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0866 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=12377 i_micro=3093 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3400 mask_tok=102.00 y_norm=48.0866±0.0070 +2026-05-10 21:01:14,541 - INFO - Step 15480 | mode=video loss=0.145508 diff=0.145508 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0510 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0874 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=12385 i_micro=3095 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3567 mask_tok=107.00 y_norm=48.0874±0.0060 +2026-05-10 21:01:32,534 - INFO - Step 15490 | mode=video loss=0.164062 diff=0.164062 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0795 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0872 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=12393 i_micro=3097 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3433 mask_tok=103.00 y_norm=48.0872±0.0057 +2026-05-10 21:01:50,501 - INFO - Step 15500 | mode=video loss=0.102051 diff=0.102051 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0477 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0891 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=12401 i_micro=3099 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3433 mask_tok=103.00 y_norm=48.0891±0.0065 +2026-05-10 21:01:57,783 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 21:02:16,204 - INFO - Step 15510 | mode=video loss=0.076172 diff=0.076172 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0633 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0887 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=12409 i_micro=3101 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3133 mask_tok=94.00 y_norm=48.0887±0.0054 +2026-05-10 21:02:34,177 - INFO - Step 15520 | mode=video loss=0.078125 diff=0.078125 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0413 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0872 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=12417 i_micro=3103 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3367 mask_tok=101.00 y_norm=48.0872±0.0059 +2026-05-10 21:02:53,281 - INFO - Step 15530 | mode=video loss=0.064453 diff=0.064453 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0512 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0876 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=12425 i_micro=3105 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3533 mask_tok=106.00 y_norm=48.0876±0.0059 +2026-05-10 21:03:11,240 - INFO - Step 15540 | mode=video loss=0.140625 diff=0.140625 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0884 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0878 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=12433 i_micro=3107 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4633 mask_tok=139.00 y_norm=48.0878±0.0069 +2026-05-10 21:03:29,199 - INFO - Step 15550 | mode=video loss=0.131836 diff=0.131836 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0514 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0873 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=12441 i_micro=3109 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3833 mask_tok=115.00 y_norm=48.0873±0.0073 +2026-05-10 21:03:48,348 - INFO - Step 15560 | mode=video loss=0.168945 diff=0.168945 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0379 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0865 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=12449 i_micro=3111 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4467 mask_tok=134.00 y_norm=48.0865±0.0060 +2026-05-10 21:04:06,302 - INFO - Step 15570 | mode=video loss=0.081055 diff=0.081055 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0261 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0889 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=12457 i_micro=3113 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3567 mask_tok=107.00 y_norm=48.0889±0.0056 +2026-05-10 21:04:25,202 - INFO - Step 15580 | mode=video loss=0.094238 diff=0.094238 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1235 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0889 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=12465 i_micro=3115 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3433 mask_tok=103.00 y_norm=48.0889±0.0060 +2026-05-10 21:04:43,169 - INFO - Step 15590 | mode=video loss=0.129883 diff=0.129883 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0509 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0875 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=12473 i_micro=3117 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3567 mask_tok=107.00 y_norm=48.0875±0.0060 +2026-05-10 21:05:02,765 - INFO - Step 15600 | mode=video loss=0.066895 diff=0.066895 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0458 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0882 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=12481 i_micro=3119 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3333 mask_tok=100.00 y_norm=48.0882±0.0053 cond_uncond_dloss=0.355469 cond_grad=0.000074 cond_pred_l2=36.056259 cond_pred_ratio=0.020237 +2026-05-10 21:05:02,802 - INFO - Step 15600 | probe_semantic mcp_offdiag(mean/min/max)=0.645279/0.531827/0.707612 smol_offdiag=0.660557 prompts=4 mcp_tok=17.50 smol_tok=17.50 +2026-05-10 21:05:09,866 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 21:05:28,917 - INFO - Step 15610 | mode=video loss=0.144531 diff=0.144531 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0621 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0861 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=12489 i_micro=3121 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4067 mask_tok=122.00 y_norm=48.0861±0.0068 +2026-05-10 21:05:46,878 - INFO - Step 15620 | mode=video loss=0.059082 diff=0.059082 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0607 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0850 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=12497 i_micro=3123 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3767 mask_tok=113.00 y_norm=48.0850±0.0057 +2026-05-10 21:06:05,961 - INFO - Step 15630 | mode=video loss=0.104980 diff=0.104980 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0410 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0857 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=12505 i_micro=3125 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4067 mask_tok=122.00 y_norm=48.0857±0.0059 +2026-05-10 21:06:23,923 - INFO - Step 15640 | mode=video loss=0.119141 diff=0.119141 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0888 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0880 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=12513 i_micro=3127 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3333 mask_tok=100.00 y_norm=48.0880±0.0054 +2026-05-10 21:06:41,898 - INFO - Step 15650 | mode=video loss=0.114746 diff=0.114746 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0660 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0889 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=12521 i_micro=3129 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3067 mask_tok=92.00 y_norm=48.0889±0.0057 +2026-05-10 21:07:01,114 - INFO - Step 15660 | mode=video loss=0.181641 diff=0.181641 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1724 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0894 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=12529 i_micro=3131 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3167 mask_tok=95.00 y_norm=48.0894±0.0061 +2026-05-10 21:07:19,077 - INFO - Step 15670 | mode=video loss=0.090820 diff=0.090820 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0374 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0890 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=12537 i_micro=3133 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3200 mask_tok=96.00 y_norm=48.0890±0.0063 +2026-05-10 21:07:37,029 - INFO - Step 15680 | mode=video loss=0.070312 diff=0.070312 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0555 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0874 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=12545 i_micro=3135 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3300 mask_tok=99.00 y_norm=48.0874±0.0055 +2026-05-10 21:07:56,017 - INFO - Step 15690 | mode=video loss=0.122559 diff=0.122559 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0370 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0867 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=12553 i_micro=3137 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4567 mask_tok=137.00 y_norm=48.0867±0.0068 +2026-05-10 21:08:13,977 - INFO - Step 15700 | mode=video loss=0.074707 diff=0.074707 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0350 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0886 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=12561 i_micro=3139 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3600 mask_tok=108.00 y_norm=48.0886±0.0059 +2026-05-10 21:08:20,972 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 21:08:39,904 - INFO - Step 15710 | mode=video loss=0.126953 diff=0.126953 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0415 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0864 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=12569 i_micro=3141 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3300 mask_tok=99.00 y_norm=48.0864±0.0061 +2026-05-10 21:08:57,866 - INFO - Step 15720 | mode=video loss=0.147461 diff=0.147461 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0731 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0866 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=12577 i_micro=3143 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3533 mask_tok=106.00 y_norm=48.0866±0.0054 +2026-05-10 21:09:15,829 - INFO - Step 15730 | mode=video loss=0.112305 diff=0.112305 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0775 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0877 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=12585 i_micro=3145 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3900 mask_tok=117.00 y_norm=48.0877±0.0059 +2026-05-10 21:09:35,042 - INFO - Step 15740 | mode=video loss=0.073730 diff=0.073730 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0391 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0879 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=12593 i_micro=3147 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3267 mask_tok=98.00 y_norm=48.0879±0.0061 +2026-05-10 21:09:53,012 - INFO - Step 15750 | mode=video loss=0.136719 diff=0.136719 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1108 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0876 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=12601 i_micro=3149 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3367 mask_tok=101.00 y_norm=48.0876±0.0054 +2026-05-10 21:10:11,713 - INFO - Step 15760 | mode=video loss=0.065430 diff=0.065430 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1005 emb_mean=0.0008 emb_std=1.0017 tok_norm=48.0840 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=12609 i_micro=3151 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4067 mask_tok=122.00 y_norm=48.0840±0.0051 +2026-05-10 21:10:30,164 - INFO - Step 15770 | mode=video loss=0.103027 diff=0.103027 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0448 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0855 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=12617 i_micro=3153 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4167 mask_tok=125.00 y_norm=48.0855±0.0059 +2026-05-10 21:10:48,124 - INFO - Step 15780 | mode=video loss=0.061523 diff=0.061523 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0361 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0878 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=12625 i_micro=3155 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4133 mask_tok=124.00 y_norm=48.0878±0.0071 +2026-05-10 21:11:07,159 - INFO - Step 15790 | mode=video loss=0.130859 diff=0.130859 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0410 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0857 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=12633 i_micro=3157 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3667 mask_tok=110.00 y_norm=48.0857±0.0057 +2026-05-10 21:11:26,734 - INFO - Step 15800 | mode=video loss=0.086426 diff=0.086426 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0982 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0871 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=12641 i_micro=3159 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.2533 mask_tok=76.00 y_norm=48.0871±0.0051 cond_uncond_dloss=0.259766 cond_grad=0.000062 cond_pred_l2=39.038601 cond_pred_ratio=0.022847 +2026-05-10 21:11:26,770 - INFO - Step 15800 | probe_semantic mcp_offdiag(mean/min/max)=0.645517/0.532139/0.707595 smol_offdiag=0.660557 prompts=4 mcp_tok=17.50 smol_tok=17.50 +2026-05-10 21:11:33,499 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 21:11:52,111 - INFO - Step 15810 | mode=video loss=0.132812 diff=0.132812 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0844 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0881 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=12649 i_micro=3161 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3367 mask_tok=101.00 y_norm=48.0881±0.0054 +2026-05-10 21:12:10,544 - INFO - Step 15820 | mode=video loss=0.154297 diff=0.154297 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0428 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0885 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=12657 i_micro=3163 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3433 mask_tok=103.00 y_norm=48.0885±0.0060 +2026-05-10 21:12:28,507 - INFO - Step 15830 | mode=video loss=0.121094 diff=0.121094 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0405 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0864 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=12665 i_micro=3165 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3367 mask_tok=101.00 y_norm=48.0864±0.0061 +2026-05-10 21:12:47,499 - INFO - Step 15840 | mode=video loss=0.090332 diff=0.090332 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0966 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0882 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=12673 i_micro=3167 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3333 mask_tok=100.00 y_norm=48.0882±0.0056 +2026-05-10 21:13:05,454 - INFO - Step 15850 | mode=video loss=0.075195 diff=0.075195 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0494 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0868 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=12681 i_micro=3169 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4067 mask_tok=122.00 y_norm=48.0868±0.0064 +2026-05-10 21:13:23,415 - INFO - Step 15860 | mode=video loss=0.122559 diff=0.122559 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0570 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0865 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=12689 i_micro=3171 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4467 mask_tok=134.00 y_norm=48.0865±0.0060 +2026-05-10 21:13:42,446 - INFO - Step 15870 | mode=video loss=0.059570 diff=0.059570 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0290 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0882 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=12697 i_micro=3173 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4033 mask_tok=121.00 y_norm=48.0882±0.0057 +2026-05-10 21:14:00,406 - INFO - Step 15880 | mode=video loss=0.203125 diff=0.203125 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.3415 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0864 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=12705 i_micro=3175 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.5167 mask_tok=155.00 y_norm=48.0864±0.0067 +2026-05-10 21:14:19,071 - INFO - Step 15890 | mode=video loss=0.144531 diff=0.144531 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0419 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0878 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=12713 i_micro=3177 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3333 mask_tok=100.00 y_norm=48.0878±0.0051 +2026-05-10 21:14:37,500 - INFO - Step 15900 | mode=video loss=0.072754 diff=0.072754 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0388 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0879 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=12721 i_micro=3179 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3467 mask_tok=104.00 y_norm=48.0879±0.0061 +2026-05-10 21:14:44,378 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 21:15:02,333 - INFO - Step 15910 | mode=video loss=0.102539 diff=0.102539 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0446 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0879 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=12729 i_micro=3181 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3767 mask_tok=113.00 y_norm=48.0879±0.0062 +2026-05-10 21:15:21,284 - INFO - Step 15920 | mode=video loss=0.083496 diff=0.083496 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.4401 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0867 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=12737 i_micro=3183 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4867 mask_tok=146.00 y_norm=48.0867±0.0067 +2026-05-10 21:15:39,254 - INFO - Step 15930 | mode=video loss=0.098633 diff=0.098633 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0316 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0893 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=12745 i_micro=3185 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3567 mask_tok=107.00 y_norm=48.0893±0.0058 +2026-05-10 21:15:57,745 - INFO - Step 15940 | mode=video loss=0.070801 diff=0.070801 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0492 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0859 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=12753 i_micro=3187 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4767 mask_tok=143.00 y_norm=48.0859±0.0069 +2026-05-10 21:16:16,127 - INFO - Step 15950 | mode=video loss=0.106934 diff=0.106934 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0341 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0866 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=12761 i_micro=3189 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3600 mask_tok=108.00 y_norm=48.0866±0.0048 +2026-05-10 21:16:34,098 - INFO - Step 15960 | mode=video loss=0.067383 diff=0.067383 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0357 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0882 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=12769 i_micro=3191 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3700 mask_tok=111.00 y_norm=48.0882±0.0054 +2026-05-10 21:16:52,981 - INFO - Step 15970 | mode=video loss=0.109375 diff=0.109375 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0451 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0876 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=12777 i_micro=3193 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3300 mask_tok=99.00 y_norm=48.0876±0.0056 +2026-05-10 21:17:10,931 - INFO - Step 15980 | mode=video loss=0.240234 diff=0.240234 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1839 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0855 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=12785 i_micro=3195 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4633 mask_tok=139.00 y_norm=48.0855±0.0057 +2026-05-10 21:17:29,580 - INFO - Step 15990 | mode=video loss=0.138672 diff=0.138672 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0522 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0855 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=12793 i_micro=3197 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3533 mask_tok=106.00 y_norm=48.0855±0.0064 +2026-05-10 21:17:49,636 - INFO - Step 16000 | mode=video loss=0.070312 diff=0.070312 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0502 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0885 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=12801 i_micro=3199 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3700 mask_tok=111.00 y_norm=48.0885±0.0067 cond_uncond_dloss=0.453125 cond_grad=0.000105 cond_pred_l2=36.362759 cond_pred_ratio=0.021054 +2026-05-10 21:17:49,673 - INFO - Step 16000 | probe_semantic mcp_offdiag(mean/min/max)=0.645623/0.532345/0.707655 smol_offdiag=0.660557 prompts=4 mcp_tok=17.50 smol_tok=17.50 +2026-05-10 21:17:56,663 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 21:18:14,612 - INFO - Step 16010 | mode=video loss=0.079102 diff=0.079102 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0509 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0865 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=12809 i_micro=3201 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.2767 mask_tok=83.00 y_norm=48.0865±0.0056 +2026-05-10 21:18:33,227 - INFO - Step 16020 | mode=video loss=0.065918 diff=0.065918 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0459 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0883 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=12817 i_micro=3203 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3267 mask_tok=98.00 y_norm=48.0883±0.0052 +2026-05-10 21:18:51,673 - INFO - Step 16030 | mode=video loss=0.117188 diff=0.117188 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0665 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0889 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=12825 i_micro=3205 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3900 mask_tok=117.00 y_norm=48.0889±0.0059 +2026-05-10 21:19:09,651 - INFO - Step 16040 | mode=video loss=0.097168 diff=0.097168 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0366 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0903 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=12833 i_micro=3207 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3800 mask_tok=114.00 y_norm=48.0903±0.0059 +2026-05-10 21:19:28,673 - INFO - Step 16050 | mode=video loss=0.099121 diff=0.099121 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0571 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0865 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=12841 i_micro=3209 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.5133 mask_tok=154.00 y_norm=48.0865±0.0070 +2026-05-10 21:19:46,628 - INFO - Step 16060 | mode=video loss=0.063477 diff=0.063477 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0439 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0875 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=12849 i_micro=3211 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3567 mask_tok=107.00 y_norm=48.0875±0.0055 +2026-05-10 21:20:05,354 - INFO - Step 16070 | mode=video loss=0.091797 diff=0.091797 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0379 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0877 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=12857 i_micro=3213 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3367 mask_tok=101.00 y_norm=48.0877±0.0051 +2026-05-10 21:20:23,744 - INFO - Step 16080 | mode=video loss=0.068359 diff=0.068359 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0679 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0864 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=12865 i_micro=3215 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.5133 mask_tok=154.00 y_norm=48.0864±0.0066 +2026-05-10 21:20:41,722 - INFO - Step 16090 | mode=video loss=0.083984 diff=0.083984 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0415 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0878 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=12873 i_micro=3217 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.2467 mask_tok=74.00 y_norm=48.0878±0.0051 +2026-05-10 21:21:00,768 - INFO - Step 16100 | mode=video loss=0.030518 diff=0.030518 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0446 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0863 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=12881 i_micro=3219 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4133 mask_tok=124.00 y_norm=48.0863±0.0057 +2026-05-10 21:21:07,675 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 21:21:25,622 - INFO - Step 16110 | mode=video loss=0.169922 diff=0.169922 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0997 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0884 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=12889 i_micro=3221 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.2600 mask_tok=78.00 y_norm=48.0884±0.0051 +2026-05-10 21:21:44,254 - INFO - Step 16120 | mode=video loss=0.082520 diff=0.082520 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0449 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0859 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=12897 i_micro=3223 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3967 mask_tok=119.00 y_norm=48.0859±0.0059 +2026-05-10 21:22:02,644 - INFO - Step 16130 | mode=video loss=0.080566 diff=0.080566 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0349 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0881 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=12905 i_micro=3225 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3533 mask_tok=106.00 y_norm=48.0881±0.0060 +2026-05-10 21:22:20,605 - INFO - Step 16140 | mode=video loss=0.132812 diff=0.132812 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0700 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0885 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=12913 i_micro=3227 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3533 mask_tok=106.00 y_norm=48.0885±0.0055 +2026-05-10 21:22:39,058 - INFO - Step 16150 | mode=video loss=0.097656 diff=0.097656 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0377 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0878 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=12921 i_micro=3229 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3367 mask_tok=101.00 y_norm=48.0878±0.0051 +2026-05-10 21:22:57,501 - INFO - Step 16160 | mode=video loss=0.053711 diff=0.053711 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0465 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0861 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=12929 i_micro=3231 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4333 mask_tok=130.00 y_norm=48.0861±0.0060 +2026-05-10 21:23:15,460 - INFO - Step 16170 | mode=video loss=0.082031 diff=0.082031 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0379 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0887 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=12937 i_micro=3233 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3333 mask_tok=100.00 y_norm=48.0887±0.0049 +2026-05-10 21:23:34,415 - INFO - Step 16180 | mode=video loss=0.084473 diff=0.084473 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0658 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0861 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=12945 i_micro=3235 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4133 mask_tok=124.00 y_norm=48.0861±0.0060 +2026-05-10 21:23:52,391 - INFO - Step 16190 | mode=video loss=0.113281 diff=0.113281 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0683 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0883 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=12953 i_micro=3237 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3367 mask_tok=101.00 y_norm=48.0883±0.0053 +2026-05-10 21:24:12,763 - INFO - Step 16200 | mode=video loss=0.128906 diff=0.128906 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0757 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0896 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=12961 i_micro=3239 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3400 mask_tok=102.00 y_norm=48.0896±0.0056 cond_uncond_dloss=0.667969 cond_grad=0.000650 cond_pred_l2=59.265701 cond_pred_ratio=0.037377 +2026-05-10 21:24:12,800 - INFO - Step 16200 | probe_semantic mcp_offdiag(mean/min/max)=0.645665/0.532202/0.707751 smol_offdiag=0.660557 prompts=4 mcp_tok=17.50 smol_tok=17.50 +2026-05-10 21:24:19,819 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 21:24:38,203 - INFO - Step 16210 | mode=video loss=0.115723 diff=0.115723 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0359 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0885 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=12969 i_micro=3241 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4633 mask_tok=139.00 y_norm=48.0885±0.0066 +2026-05-10 21:24:56,168 - INFO - Step 16220 | mode=video loss=0.077637 diff=0.077637 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0747 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0887 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=12977 i_micro=3243 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.2933 mask_tok=88.00 y_norm=48.0887±0.0057 +2026-05-10 21:25:15,306 - INFO - Step 16230 | mode=video loss=0.088867 diff=0.088867 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0428 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0891 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=12985 i_micro=3245 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3400 mask_tok=102.00 y_norm=48.0891±0.0065 +2026-05-10 21:25:33,268 - INFO - Step 16240 | mode=video loss=0.121582 diff=0.121582 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0484 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0885 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=12993 i_micro=3247 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3467 mask_tok=104.00 y_norm=48.0885±0.0057 +2026-05-10 21:25:51,961 - INFO - Step 16250 | mode=video loss=0.111816 diff=0.111816 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0485 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0876 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=13001 i_micro=3249 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.2900 mask_tok=87.00 y_norm=48.0876±0.0052 +2026-05-10 21:26:10,370 - INFO - Step 16260 | mode=video loss=0.100586 diff=0.100586 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0374 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0870 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=13009 i_micro=3251 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3667 mask_tok=110.00 y_norm=48.0870±0.0058 +2026-05-10 21:26:28,336 - INFO - Step 16270 | mode=video loss=0.116211 diff=0.116211 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0372 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0863 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=13017 i_micro=3253 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3867 mask_tok=116.00 y_norm=48.0863±0.0053 +2026-05-10 21:26:46,951 - INFO - Step 16280 | mode=video loss=0.139648 diff=0.139648 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0807 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0876 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=13025 i_micro=3255 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3500 mask_tok=105.00 y_norm=48.0876±0.0056 +2026-05-10 21:27:05,394 - INFO - Step 16290 | mode=video loss=0.115234 diff=0.115234 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1052 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0883 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=13033 i_micro=3257 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4100 mask_tok=123.00 y_norm=48.0883±0.0062 +2026-05-10 21:27:23,384 - INFO - Step 16300 | mode=video loss=0.073730 diff=0.073730 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0353 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0879 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=13041 i_micro=3259 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.2567 mask_tok=77.00 y_norm=48.0879±0.0054 +2026-05-10 21:27:30,748 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 21:27:49,862 - INFO - Step 16310 | mode=video loss=0.093750 diff=0.093750 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0772 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0864 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=13049 i_micro=3261 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3433 mask_tok=103.00 y_norm=48.0864±0.0066 +2026-05-10 21:28:07,842 - INFO - Step 16320 | mode=video loss=0.073730 diff=0.073730 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0945 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0877 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=13057 i_micro=3263 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.5367 mask_tok=161.00 y_norm=48.0877±0.0068 +2026-05-10 21:28:26,607 - INFO - Step 16330 | mode=video loss=0.075684 diff=0.075684 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0659 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0882 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=13065 i_micro=3265 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3267 mask_tok=98.00 y_norm=48.0882±0.0057 +2026-05-10 21:28:45,007 - INFO - Step 16340 | mode=video loss=0.104004 diff=0.104004 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0874 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0877 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=13073 i_micro=3267 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3433 mask_tok=103.00 y_norm=48.0877±0.0062 +2026-05-10 21:29:02,987 - INFO - Step 16350 | mode=video loss=0.091797 diff=0.091797 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0408 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0871 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=13081 i_micro=3269 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3600 mask_tok=108.00 y_norm=48.0871±0.0067 +2026-05-10 21:29:21,866 - INFO - Step 16360 | mode=video loss=0.081055 diff=0.081055 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0307 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0886 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=13089 i_micro=3271 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.2867 mask_tok=86.00 y_norm=48.0886±0.0054 +2026-05-10 21:29:39,834 - INFO - Step 16370 | mode=video loss=0.092773 diff=0.092773 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1858 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0871 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=13097 i_micro=3273 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3433 mask_tok=103.00 y_norm=48.0871±0.0065 +2026-05-10 21:29:58,488 - INFO - Step 16380 | mode=video loss=0.075195 diff=0.075195 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0490 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0863 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=13105 i_micro=3275 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3333 mask_tok=100.00 y_norm=48.0863±0.0060 +2026-05-10 21:30:16,867 - INFO - Step 16390 | mode=video loss=0.142578 diff=0.142578 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0495 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0870 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=13113 i_micro=3277 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3833 mask_tok=115.00 y_norm=48.0870±0.0060 +2026-05-10 21:30:36,455 - INFO - Step 16400 | mode=video loss=0.039062 diff=0.039062 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0333 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0855 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=13121 i_micro=3279 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.5033 mask_tok=151.00 y_norm=48.0855±0.0067 cond_uncond_dloss=0.203125 cond_grad=0.000061 cond_pred_l2=39.810192 cond_pred_ratio=0.021919 +2026-05-10 21:30:36,492 - INFO - Step 16400 | probe_semantic mcp_offdiag(mean/min/max)=0.645307/0.531786/0.707488 smol_offdiag=0.660557 prompts=4 mcp_tok=17.50 smol_tok=17.50 +2026-05-10 21:30:43,600 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 21:31:02,425 - INFO - Step 16410 | mode=video loss=0.059814 diff=0.059814 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0296 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0867 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=13129 i_micro=3281 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3633 mask_tok=109.00 y_norm=48.0867±0.0059 +2026-05-10 21:31:20,884 - INFO - Step 16420 | mode=video loss=0.043701 diff=0.043701 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0305 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0871 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=13137 i_micro=3283 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4433 mask_tok=133.00 y_norm=48.0871±0.0071 +2026-05-10 21:31:39,773 - INFO - Step 16430 | mode=video loss=0.143555 diff=0.143555 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0515 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0875 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=13145 i_micro=3285 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3433 mask_tok=103.00 y_norm=48.0875±0.0053 +2026-05-10 21:31:58,179 - INFO - Step 16440 | mode=video loss=0.091309 diff=0.091309 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0667 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0860 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=13153 i_micro=3287 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4533 mask_tok=136.00 y_norm=48.0860±0.0067 +2026-05-10 21:32:16,135 - INFO - Step 16450 | mode=video loss=0.065918 diff=0.065918 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0524 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0856 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=13161 i_micro=3289 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4667 mask_tok=140.00 y_norm=48.0856±0.0057 +2026-05-10 21:32:34,903 - INFO - Step 16460 | mode=video loss=0.136719 diff=0.136719 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0261 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0878 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=13169 i_micro=3291 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.5433 mask_tok=163.00 y_norm=48.0878±0.0078 +2026-05-10 21:32:53,319 - INFO - Step 16470 | mode=video loss=0.096191 diff=0.096191 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0817 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0871 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=13177 i_micro=3293 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4533 mask_tok=136.00 y_norm=48.0871±0.0069 +2026-05-10 21:33:11,277 - INFO - Step 16480 | mode=video loss=0.166016 diff=0.166016 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.3884 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0868 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=13185 i_micro=3295 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.5300 mask_tok=159.00 y_norm=48.0868±0.0073 +2026-05-10 21:33:30,345 - INFO - Step 16490 | mode=video loss=0.096680 diff=0.096680 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0494 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0880 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=13193 i_micro=3297 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3167 mask_tok=95.00 y_norm=48.0880±0.0054 +2026-05-10 21:33:48,320 - INFO - Step 16500 | mode=video loss=0.104004 diff=0.104004 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0234 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0859 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=13201 i_micro=3299 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3700 mask_tok=111.00 y_norm=48.0859±0.0071 +2026-05-10 21:33:55,763 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 21:34:14,428 - INFO - Step 16510 | mode=video loss=0.022095 diff=0.022095 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0616 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0862 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=13209 i_micro=3301 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4433 mask_tok=133.00 y_norm=48.0862±0.0062 +2026-05-10 21:34:32,804 - INFO - Step 16520 | mode=video loss=0.204102 diff=0.204102 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0366 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0863 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=13217 i_micro=3303 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3133 mask_tok=94.00 y_norm=48.0863±0.0060 +2026-05-10 21:34:50,771 - INFO - Step 16530 | mode=video loss=0.058838 diff=0.058838 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0507 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0878 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=13225 i_micro=3305 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3367 mask_tok=101.00 y_norm=48.0878±0.0058 +2026-05-10 21:35:09,503 - INFO - Step 16540 | mode=video loss=0.097168 diff=0.097168 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0310 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0886 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=13233 i_micro=3307 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.2900 mask_tok=87.00 y_norm=48.0886±0.0051 +2026-05-10 21:35:27,920 - INFO - Step 16550 | mode=video loss=0.077148 diff=0.077148 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0524 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0864 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=13241 i_micro=3309 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3267 mask_tok=98.00 y_norm=48.0864±0.0061 +2026-05-10 21:35:46,663 - INFO - Step 16560 | mode=video loss=0.076172 diff=0.076172 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0357 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0887 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=13249 i_micro=3311 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3767 mask_tok=113.00 y_norm=48.0887±0.0057 +2026-05-10 21:36:05,016 - INFO - Step 16570 | mode=video loss=0.091309 diff=0.091309 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0435 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0868 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=13257 i_micro=3313 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3333 mask_tok=100.00 y_norm=48.0868±0.0055 +2026-05-10 21:36:23,007 - INFO - Step 16580 | mode=video loss=0.083984 diff=0.083984 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0664 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0892 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=13265 i_micro=3315 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3467 mask_tok=104.00 y_norm=48.0892±0.0058 +2026-05-10 21:36:41,740 - INFO - Step 16590 | mode=video loss=0.086426 diff=0.086426 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0646 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0863 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=13273 i_micro=3317 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3600 mask_tok=108.00 y_norm=48.0863±0.0061 +2026-05-10 21:37:01,729 - INFO - Step 16600 | mode=video loss=0.056641 diff=0.056641 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0666 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0882 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=13281 i_micro=3319 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3733 mask_tok=112.00 y_norm=48.0882±0.0063 cond_uncond_dloss=0.408203 cond_grad=0.000094 cond_pred_l2=54.072464 cond_pred_ratio=0.032297 +2026-05-10 21:37:01,765 - INFO - Step 16600 | probe_semantic mcp_offdiag(mean/min/max)=0.645857/0.532683/0.707719 smol_offdiag=0.660557 prompts=4 mcp_tok=17.50 smol_tok=17.50 +2026-05-10 21:37:08,473 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 21:37:26,426 - INFO - Step 16610 | mode=video loss=0.091309 diff=0.091309 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0480 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0872 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=13289 i_micro=3321 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4000 mask_tok=120.00 y_norm=48.0872±0.0055 +2026-05-10 21:37:45,586 - INFO - Step 16620 | mode=video loss=0.141602 diff=0.141602 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.2895 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0894 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=13297 i_micro=3323 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3400 mask_tok=102.00 y_norm=48.0894±0.0056 +2026-05-10 21:38:03,552 - INFO - Step 16630 | mode=video loss=0.104492 diff=0.104492 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0674 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0891 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=13305 i_micro=3325 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3300 mask_tok=99.00 y_norm=48.0891±0.0054 +2026-05-10 21:38:22,276 - INFO - Step 16640 | mode=video loss=0.088379 diff=0.088379 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0443 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0865 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=13313 i_micro=3327 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3533 mask_tok=106.00 y_norm=48.0865±0.0060 +2026-05-10 21:38:40,642 - INFO - Step 16650 | mode=video loss=0.045898 diff=0.045898 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0464 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0881 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=13321 i_micro=3329 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3567 mask_tok=107.00 y_norm=48.0881±0.0054 +2026-05-10 21:38:58,598 - INFO - Step 16660 | mode=video loss=0.129883 diff=0.129883 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0405 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0880 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=13329 i_micro=3331 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3400 mask_tok=102.00 y_norm=48.0880±0.0056 +2026-05-10 21:39:17,256 - INFO - Step 16670 | mode=video loss=0.069824 diff=0.069824 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0391 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0871 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=13337 i_micro=3333 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.5033 mask_tok=151.00 y_norm=48.0871±0.0075 +2026-05-10 21:39:35,650 - INFO - Step 16680 | mode=video loss=0.081543 diff=0.081543 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0650 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0858 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=13345 i_micro=3335 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.5133 mask_tok=154.00 y_norm=48.0858±0.0070 +2026-05-10 21:39:54,385 - INFO - Step 16690 | mode=video loss=0.059326 diff=0.059326 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0957 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0859 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=13353 i_micro=3337 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4667 mask_tok=140.00 y_norm=48.0859±0.0064 +2026-05-10 21:40:12,723 - INFO - Step 16700 | mode=video loss=0.057129 diff=0.057129 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0622 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0867 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=13361 i_micro=3339 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4367 mask_tok=131.00 y_norm=48.0867±0.0067 +2026-05-10 21:40:19,523 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 21:40:37,471 - INFO - Step 16710 | mode=video loss=0.089844 diff=0.089844 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0357 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0879 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=13369 i_micro=3341 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3233 mask_tok=97.00 y_norm=48.0879±0.0055 +2026-05-10 21:40:56,188 - INFO - Step 16720 | mode=video loss=0.090332 diff=0.090332 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0358 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0854 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=13377 i_micro=3343 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3567 mask_tok=107.00 y_norm=48.0854±0.0065 +2026-05-10 21:41:14,609 - INFO - Step 16730 | mode=video loss=0.094727 diff=0.094727 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0784 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0844 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=13385 i_micro=3345 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4000 mask_tok=120.00 y_norm=48.0844±0.0061 +2026-05-10 21:41:32,567 - INFO - Step 16740 | mode=video loss=0.059082 diff=0.059082 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0722 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0884 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=13393 i_micro=3347 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3333 mask_tok=100.00 y_norm=48.0884±0.0049 +2026-05-10 21:41:51,626 - INFO - Step 16750 | mode=video loss=0.101074 diff=0.101074 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0524 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0868 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=13401 i_micro=3349 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3633 mask_tok=109.00 y_norm=48.0868±0.0064 +2026-05-10 21:42:09,601 - INFO - Step 16760 | mode=video loss=0.129883 diff=0.129883 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0470 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0851 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=13409 i_micro=3351 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3667 mask_tok=110.00 y_norm=48.0851±0.0055 +2026-05-10 21:42:28,270 - INFO - Step 16770 | mode=video loss=0.149414 diff=0.149414 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0262 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0852 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=13417 i_micro=3353 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4500 mask_tok=135.00 y_norm=48.0852±0.0066 +2026-05-10 21:42:46,671 - INFO - Step 16780 | mode=video loss=0.083008 diff=0.083008 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0580 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0864 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=13425 i_micro=3355 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3633 mask_tok=109.00 y_norm=48.0864±0.0058 +2026-05-10 21:43:04,623 - INFO - Step 16790 | mode=video loss=0.121582 diff=0.121582 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0865 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0867 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=13433 i_micro=3357 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3500 mask_tok=105.00 y_norm=48.0867±0.0066 +2026-05-10 21:43:24,776 - INFO - Step 16800 | mode=video loss=0.107422 diff=0.107422 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0314 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0873 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=13441 i_micro=3359 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4133 mask_tok=124.00 y_norm=48.0873±0.0062 cond_uncond_dloss=0.314453 cond_grad=0.000096 cond_pred_l2=54.931141 cond_pred_ratio=0.034744 +2026-05-10 21:43:24,813 - INFO - Step 16800 | probe_semantic mcp_offdiag(mean/min/max)=0.645440/0.532159/0.707479 smol_offdiag=0.660557 prompts=4 mcp_tok=17.50 smol_tok=17.50 +2026-05-10 21:43:31,756 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 21:43:50,212 - INFO - Step 16810 | mode=video loss=0.062256 diff=0.062256 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0381 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0880 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=13449 i_micro=3361 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.2467 mask_tok=74.00 y_norm=48.0880±0.0052 +2026-05-10 21:44:08,862 - INFO - Step 16820 | mode=video loss=0.091309 diff=0.091309 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1265 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0875 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=13457 i_micro=3363 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3633 mask_tok=109.00 y_norm=48.0875±0.0057 +2026-05-10 21:44:27,271 - INFO - Step 16830 | mode=video loss=0.120117 diff=0.120117 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0385 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0889 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=13465 i_micro=3365 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3733 mask_tok=112.00 y_norm=48.0889±0.0067 +2026-05-10 21:44:45,240 - INFO - Step 16840 | mode=video loss=0.115234 diff=0.115234 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0605 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0870 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=13473 i_micro=3367 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4167 mask_tok=125.00 y_norm=48.0870±0.0064 +2026-05-10 21:45:03,984 - INFO - Step 16850 | mode=video loss=0.084961 diff=0.084961 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0505 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0896 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=13481 i_micro=3369 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3433 mask_tok=103.00 y_norm=48.0896±0.0057 +2026-05-10 21:45:22,393 - INFO - Step 16860 | mode=video loss=0.086426 diff=0.086426 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0529 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0880 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=13489 i_micro=3371 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4267 mask_tok=128.00 y_norm=48.0880±0.0068 +2026-05-10 21:45:40,360 - INFO - Step 16870 | mode=video loss=0.269531 diff=0.269531 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.2668 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0855 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=13497 i_micro=3373 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4067 mask_tok=122.00 y_norm=48.0855±0.0066 +2026-05-10 21:45:59,531 - INFO - Step 16880 | mode=video loss=0.184570 diff=0.184570 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1280 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0886 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=13505 i_micro=3375 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3100 mask_tok=93.00 y_norm=48.0886±0.0052 +2026-05-10 21:46:17,500 - INFO - Step 16890 | mode=video loss=0.089355 diff=0.089355 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1480 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0869 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=13513 i_micro=3377 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3733 mask_tok=112.00 y_norm=48.0869±0.0063 +2026-05-10 21:46:36,141 - INFO - Step 16900 | mode=video loss=0.059570 diff=0.059570 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0461 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0868 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=13521 i_micro=3379 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3433 mask_tok=103.00 y_norm=48.0868±0.0058 +2026-05-10 21:46:42,986 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 21:47:01,399 - INFO - Step 16910 | mode=video loss=0.185547 diff=0.185547 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0631 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0892 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=13529 i_micro=3381 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3100 mask_tok=93.00 y_norm=48.0892±0.0049 +2026-05-10 21:47:19,367 - INFO - Step 16920 | mode=video loss=0.131836 diff=0.131836 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0657 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0896 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=13537 i_micro=3383 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3300 mask_tok=99.00 y_norm=48.0896±0.0059 +2026-05-10 21:47:38,625 - INFO - Step 16930 | mode=video loss=0.134766 diff=0.134766 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0645 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0865 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=13545 i_micro=3385 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3267 mask_tok=98.00 y_norm=48.0865±0.0067 +2026-05-10 21:47:56,583 - INFO - Step 16940 | mode=video loss=0.322266 diff=0.322266 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0831 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0868 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=13553 i_micro=3387 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3400 mask_tok=102.00 y_norm=48.0868±0.0056 +2026-05-10 21:48:15,336 - INFO - Step 16950 | mode=video loss=0.103516 diff=0.103516 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0584 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0844 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=13561 i_micro=3389 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3667 mask_tok=110.00 y_norm=48.0844±0.0057 +2026-05-10 21:48:33,754 - INFO - Step 16960 | mode=video loss=0.100098 diff=0.100098 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0749 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0903 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=13569 i_micro=3391 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3433 mask_tok=103.00 y_norm=48.0903±0.0065 +2026-05-10 21:48:51,707 - INFO - Step 16970 | mode=video loss=0.073242 diff=0.073242 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0446 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0865 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=13577 i_micro=3393 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4800 mask_tok=144.00 y_norm=48.0865±0.0060 +2026-05-10 21:49:10,377 - INFO - Step 16980 | mode=video loss=0.102539 diff=0.102539 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1038 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0883 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=13585 i_micro=3395 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.2767 mask_tok=83.00 y_norm=48.0883±0.0051 +2026-05-10 21:49:28,752 - INFO - Step 16990 | mode=video loss=0.104492 diff=0.104492 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0259 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0869 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=13593 i_micro=3397 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3767 mask_tok=113.00 y_norm=48.0869±0.0055 +2026-05-10 21:49:48,351 - INFO - Step 17000 | mode=video loss=0.094238 diff=0.094238 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0390 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0892 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=13601 i_micro=3399 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3400 mask_tok=102.00 y_norm=48.0892±0.0061 cond_uncond_dloss=0.253906 cond_grad=0.000035 cond_pred_l2=48.372738 cond_pred_ratio=0.028412 +2026-05-10 21:49:48,388 - INFO - Step 17000 | probe_semantic mcp_offdiag(mean/min/max)=0.645321/0.531978/0.707530 smol_offdiag=0.660557 prompts=4 mcp_tok=17.50 smol_tok=17.50 +2026-05-10 21:49:56,087 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 21:50:15,356 - INFO - Step 17010 | mode=video loss=0.107422 diff=0.107422 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0355 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0870 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=13609 i_micro=3401 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3433 mask_tok=103.00 y_norm=48.0870±0.0062 +2026-05-10 21:50:33,331 - INFO - Step 17020 | mode=video loss=0.080078 diff=0.080078 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0714 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0871 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=13617 i_micro=3403 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4467 mask_tok=134.00 y_norm=48.0871±0.0063 +2026-05-10 21:50:52,032 - INFO - Step 17030 | mode=video loss=0.125000 diff=0.125000 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0799 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0876 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=13625 i_micro=3405 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3467 mask_tok=104.00 y_norm=48.0876±0.0051 +2026-05-10 21:51:10,492 - INFO - Step 17040 | mode=video loss=0.094238 diff=0.094238 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0391 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0874 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=13633 i_micro=3407 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3500 mask_tok=105.00 y_norm=48.0874±0.0052 +2026-05-10 21:51:28,460 - INFO - Step 17050 | mode=video loss=0.125977 diff=0.125977 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0343 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0854 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=13641 i_micro=3409 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3967 mask_tok=119.00 y_norm=48.0854±0.0062 +2026-05-10 21:51:47,201 - INFO - Step 17060 | mode=video loss=0.099121 diff=0.099121 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0390 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0881 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=13649 i_micro=3411 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3333 mask_tok=100.00 y_norm=48.0881±0.0051 +2026-05-10 21:52:05,634 - INFO - Step 17070 | mode=video loss=0.076660 diff=0.076660 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0450 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0873 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=13657 i_micro=3413 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3400 mask_tok=102.00 y_norm=48.0873±0.0056 +2026-05-10 21:52:24,261 - INFO - Step 17080 | mode=video loss=0.069336 diff=0.069336 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0573 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0882 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=13665 i_micro=3415 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3500 mask_tok=105.00 y_norm=48.0882±0.0062 +2026-05-10 21:52:42,642 - INFO - Step 17090 | mode=video loss=0.071777 diff=0.071777 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0591 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0897 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=13673 i_micro=3417 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.5500 mask_tok=165.00 y_norm=48.0897±0.0070 +2026-05-10 21:53:00,594 - INFO - Step 17100 | mode=video loss=0.094238 diff=0.094238 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0307 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0890 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=13681 i_micro=3419 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3100 mask_tok=93.00 y_norm=48.0890±0.0056 +2026-05-10 21:53:07,405 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 21:53:26,171 - INFO - Step 17110 | mode=video loss=0.088867 diff=0.088867 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0286 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0875 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=13689 i_micro=3421 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3633 mask_tok=109.00 y_norm=48.0875±0.0068 +2026-05-10 21:53:44,563 - INFO - Step 17120 | mode=video loss=0.092773 diff=0.092773 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1112 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0858 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=13697 i_micro=3423 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3933 mask_tok=118.00 y_norm=48.0858±0.0057 +2026-05-10 21:54:02,530 - INFO - Step 17130 | mode=video loss=0.083008 diff=0.083008 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0334 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0879 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=13705 i_micro=3425 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3700 mask_tok=111.00 y_norm=48.0879±0.0060 +2026-05-10 21:54:21,717 - INFO - Step 17140 | mode=video loss=0.083496 diff=0.083496 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0343 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0874 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=13713 i_micro=3427 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3433 mask_tok=103.00 y_norm=48.0874±0.0049 +2026-05-10 21:54:39,671 - INFO - Step 17150 | mode=video loss=0.095215 diff=0.095215 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0544 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0919 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=13721 i_micro=3429 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.5467 mask_tok=164.00 y_norm=48.0919±0.0078 +2026-05-10 21:54:58,409 - INFO - Step 17160 | mode=video loss=0.088867 diff=0.088867 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0450 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0873 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=13729 i_micro=3431 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4667 mask_tok=140.00 y_norm=48.0873±0.0058 +2026-05-10 21:55:16,784 - INFO - Step 17170 | mode=video loss=0.083008 diff=0.083008 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0300 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0846 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=13737 i_micro=3433 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4133 mask_tok=124.00 y_norm=48.0846±0.0054 +2026-05-10 21:55:34,747 - INFO - Step 17180 | mode=video loss=0.092773 diff=0.092773 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0425 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0856 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=13745 i_micro=3435 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3833 mask_tok=115.00 y_norm=48.0856±0.0059 +2026-05-10 21:55:53,434 - INFO - Step 17190 | mode=video loss=0.078613 diff=0.078613 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0707 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0888 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=13753 i_micro=3437 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3700 mask_tok=111.00 y_norm=48.0888±0.0061 +2026-05-10 21:56:13,503 - INFO - Step 17200 | mode=video loss=0.089355 diff=0.089355 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0541 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0874 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=13761 i_micro=3439 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3400 mask_tok=102.00 y_norm=48.0874±0.0054 cond_uncond_dloss=0.664062 cond_grad=0.000256 cond_pred_l2=64.954903 cond_pred_ratio=0.039238 +2026-05-10 21:56:13,540 - INFO - Step 17200 | probe_semantic mcp_offdiag(mean/min/max)=0.645004/0.531566/0.707287 smol_offdiag=0.660557 prompts=4 mcp_tok=17.50 smol_tok=17.50 +2026-05-10 21:56:20,529 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 21:56:39,259 - INFO - Step 17210 | mode=video loss=0.091309 diff=0.091309 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0262 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0889 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=13769 i_micro=3441 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3333 mask_tok=100.00 y_norm=48.0889±0.0054 +2026-05-10 21:56:57,664 - INFO - Step 17220 | mode=video loss=0.085449 diff=0.085449 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0367 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0885 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=13777 i_micro=3443 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3467 mask_tok=104.00 y_norm=48.0885±0.0056 +2026-05-10 21:57:15,633 - INFO - Step 17230 | mode=video loss=0.078125 diff=0.078125 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0303 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0851 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=13785 i_micro=3445 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4800 mask_tok=144.00 y_norm=48.0851±0.0064 +2026-05-10 21:57:34,310 - INFO - Step 17240 | mode=video loss=0.103516 diff=0.103516 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0386 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0871 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=13793 i_micro=3447 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3500 mask_tok=105.00 y_norm=48.0871±0.0058 +2026-05-10 21:57:52,685 - INFO - Step 17250 | mode=video loss=0.069824 diff=0.069824 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0631 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0868 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=13801 i_micro=3449 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3033 mask_tok=91.00 y_norm=48.0868±0.0051 +2026-05-10 21:58:10,656 - INFO - Step 17260 | mode=video loss=0.077148 diff=0.077148 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0626 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0872 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=13809 i_micro=3451 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4533 mask_tok=136.00 y_norm=48.0872±0.0068 +2026-05-10 21:58:29,816 - INFO - Step 17270 | mode=video loss=0.112793 diff=0.112793 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0968 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0876 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=13817 i_micro=3453 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3133 mask_tok=94.00 y_norm=48.0876±0.0062 +2026-05-10 21:58:47,793 - INFO - Step 17280 | mode=video loss=0.057617 diff=0.057617 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0374 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0866 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=13825 i_micro=3455 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3633 mask_tok=109.00 y_norm=48.0866±0.0052 +2026-05-10 21:59:06,427 - INFO - Step 17290 | mode=video loss=0.125977 diff=0.125977 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0272 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0866 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=13833 i_micro=3457 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4533 mask_tok=136.00 y_norm=48.0866±0.0059 +2026-05-10 21:59:24,849 - INFO - Step 17300 | mode=video loss=0.071289 diff=0.071289 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0285 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0889 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=13841 i_micro=3459 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3333 mask_tok=100.00 y_norm=48.0889±0.0056 +2026-05-10 21:59:31,721 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 21:59:49,666 - INFO - Step 17310 | mode=video loss=0.166016 diff=0.166016 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.2151 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0874 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=13849 i_micro=3461 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3467 mask_tok=104.00 y_norm=48.0874±0.0058 +2026-05-10 22:00:08,288 - INFO - Step 17320 | mode=video loss=0.116699 diff=0.116699 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0363 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0881 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=13857 i_micro=3463 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3267 mask_tok=98.00 y_norm=48.0881±0.0051 +2026-05-10 22:00:26,736 - INFO - Step 17330 | mode=video loss=0.099121 diff=0.099121 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0809 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0858 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=13865 i_micro=3465 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3600 mask_tok=108.00 y_norm=48.0858±0.0063 +2026-05-10 22:00:45,481 - INFO - Step 17340 | mode=video loss=0.069824 diff=0.069824 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0552 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0887 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=13873 i_micro=3467 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3333 mask_tok=100.00 y_norm=48.0887±0.0057 +2026-05-10 22:01:03,849 - INFO - Step 17350 | mode=video loss=0.032959 diff=0.032959 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0590 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0857 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=13881 i_micro=3469 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4067 mask_tok=122.00 y_norm=48.0857±0.0061 +2026-05-10 22:01:21,797 - INFO - Step 17360 | mode=video loss=0.206055 diff=0.206055 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0540 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0866 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=13889 i_micro=3471 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3533 mask_tok=106.00 y_norm=48.0866±0.0067 +2026-05-10 22:01:40,499 - INFO - Step 17370 | mode=video loss=0.095215 diff=0.095215 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0434 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0878 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=13897 i_micro=3473 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.2867 mask_tok=86.00 y_norm=48.0878±0.0051 +2026-05-10 22:01:58,934 - INFO - Step 17380 | mode=video loss=0.115723 diff=0.115723 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0461 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0882 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=13905 i_micro=3475 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3633 mask_tok=109.00 y_norm=48.0882±0.0062 +2026-05-10 22:02:16,909 - INFO - Step 17390 | mode=video loss=0.088379 diff=0.088379 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0878 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0858 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=13913 i_micro=3477 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3500 mask_tok=105.00 y_norm=48.0858±0.0064 +2026-05-10 22:02:37,754 - INFO - Step 17400 | mode=video loss=0.119629 diff=0.119629 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0634 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0868 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=13921 i_micro=3479 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4133 mask_tok=124.00 y_norm=48.0868±0.0063 cond_uncond_dloss=0.404297 cond_grad=0.000117 cond_pred_l2=32.600746 cond_pred_ratio=0.018929 +2026-05-10 22:02:37,793 - INFO - Step 17400 | probe_semantic mcp_offdiag(mean/min/max)=0.645034/0.531749/0.707226 smol_offdiag=0.660557 prompts=4 mcp_tok=17.50 smol_tok=17.50 +2026-05-10 22:02:44,737 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 22:03:02,713 - INFO - Step 17410 | mode=video loss=0.068359 diff=0.068359 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0674 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0879 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=13929 i_micro=3481 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3367 mask_tok=101.00 y_norm=48.0879±0.0050 +2026-05-10 22:03:21,543 - INFO - Step 17420 | mode=video loss=0.155273 diff=0.155273 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0888 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0873 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=13937 i_micro=3483 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3600 mask_tok=108.00 y_norm=48.0873±0.0060 +2026-05-10 22:03:39,961 - INFO - Step 17430 | mode=video loss=0.103516 diff=0.103516 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0282 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0872 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=13945 i_micro=3485 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.2767 mask_tok=83.00 y_norm=48.0872±0.0055 +2026-05-10 22:03:57,932 - INFO - Step 17440 | mode=video loss=0.079590 diff=0.079590 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0863 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0871 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=13953 i_micro=3487 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3300 mask_tok=99.00 y_norm=48.0871±0.0062 +2026-05-10 22:04:16,671 - INFO - Step 17450 | mode=video loss=0.076172 diff=0.076172 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1567 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0880 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=13961 i_micro=3489 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3267 mask_tok=98.00 y_norm=48.0880±0.0050 +2026-05-10 22:04:35,113 - INFO - Step 17460 | mode=video loss=0.044922 diff=0.044922 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0760 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0859 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=13969 i_micro=3491 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4267 mask_tok=128.00 y_norm=48.0859±0.0058 +2026-05-10 22:04:53,863 - INFO - Step 17470 | mode=video loss=0.087891 diff=0.087891 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0431 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0857 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=13977 i_micro=3493 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4633 mask_tok=139.00 y_norm=48.0857±0.0062 +2026-05-10 22:05:12,255 - INFO - Step 17480 | mode=video loss=0.063965 diff=0.063965 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0388 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0865 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=13985 i_micro=3495 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4100 mask_tok=123.00 y_norm=48.0865±0.0067 +2026-05-10 22:05:30,268 - INFO - Step 17490 | mode=video loss=0.212891 diff=0.212891 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1261 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0869 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=13993 i_micro=3497 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4167 mask_tok=125.00 y_norm=48.0869±0.0071 +2026-05-10 22:05:48,844 - INFO - Step 17500 | mode=video loss=0.038818 diff=0.038818 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0310 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0887 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=14001 i_micro=3499 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3633 mask_tok=109.00 y_norm=48.0887±0.0054 +2026-05-10 22:05:55,993 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 22:06:14,354 - INFO - Step 17510 | mode=video loss=0.079102 diff=0.079102 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0543 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0863 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=14009 i_micro=3501 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.5100 mask_tok=153.00 y_norm=48.0863±0.0067 +2026-05-10 22:06:32,804 - INFO - Step 17520 | mode=video loss=0.073730 diff=0.073730 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0868 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0872 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=14017 i_micro=3503 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4467 mask_tok=134.00 y_norm=48.0872±0.0073 +2026-05-10 22:06:51,170 - INFO - Step 17530 | mode=video loss=0.081543 diff=0.081543 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0260 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0882 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=14025 i_micro=3505 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3433 mask_tok=103.00 y_norm=48.0882±0.0056 +2026-05-10 22:07:09,143 - INFO - Step 17540 | mode=video loss=0.132812 diff=0.132812 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0359 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0863 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=14033 i_micro=3507 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.2233 mask_tok=67.00 y_norm=48.0863±0.0050 +2026-05-10 22:07:27,609 - INFO - Step 17550 | mode=video loss=0.108887 diff=0.108887 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0372 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0885 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=14041 i_micro=3509 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3400 mask_tok=102.00 y_norm=48.0885±0.0048 +2026-05-10 22:07:45,989 - INFO - Step 17560 | mode=video loss=0.092773 diff=0.092773 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0464 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0874 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=14049 i_micro=3511 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3967 mask_tok=119.00 y_norm=48.0874±0.0063 +2026-05-10 22:08:03,982 - INFO - Step 17570 | mode=video loss=0.085449 diff=0.085449 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0782 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0860 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=14057 i_micro=3513 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4367 mask_tok=131.00 y_norm=48.0860±0.0057 +2026-05-10 22:08:22,563 - INFO - Step 17580 | mode=video loss=0.151367 diff=0.151367 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0418 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0869 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=14065 i_micro=3515 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4433 mask_tok=133.00 y_norm=48.0869±0.0065 +2026-05-10 22:08:40,993 - INFO - Step 17590 | mode=video loss=0.099609 diff=0.099609 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0405 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0881 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=14073 i_micro=3517 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.2833 mask_tok=85.00 y_norm=48.0881±0.0049 +2026-05-10 22:09:01,375 - INFO - Step 17600 | mode=video loss=0.094727 diff=0.094727 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0612 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0868 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=14081 i_micro=3519 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3800 mask_tok=114.00 y_norm=48.0868±0.0064 cond_uncond_dloss=1.171875 cond_grad=0.000085 cond_pred_l2=38.644562 cond_pred_ratio=0.023776 +2026-05-10 22:09:01,412 - INFO - Step 17600 | probe_semantic mcp_offdiag(mean/min/max)=0.645043/0.531811/0.707194 smol_offdiag=0.660557 prompts=4 mcp_tok=17.50 smol_tok=17.50 +2026-05-10 22:09:08,300 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 22:09:26,698 - INFO - Step 17610 | mode=video loss=0.041504 diff=0.041504 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1957 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0850 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=14089 i_micro=3521 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3833 mask_tok=115.00 y_norm=48.0850±0.0057 +2026-05-10 22:09:44,669 - INFO - Step 17620 | mode=video loss=0.098633 diff=0.098633 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0549 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0857 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=14097 i_micro=3523 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3500 mask_tok=105.00 y_norm=48.0857±0.0055 +2026-05-10 22:10:03,469 - INFO - Step 17630 | mode=video loss=0.081543 diff=0.081543 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0322 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0869 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=14105 i_micro=3525 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3533 mask_tok=106.00 y_norm=48.0869±0.0064 +2026-05-10 22:10:21,865 - INFO - Step 17640 | mode=video loss=0.109375 diff=0.109375 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0441 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0874 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=14113 i_micro=3527 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3333 mask_tok=100.00 y_norm=48.0874±0.0057 +2026-05-10 22:10:40,580 - INFO - Step 17650 | mode=video loss=0.110352 diff=0.110352 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0355 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0874 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=14121 i_micro=3529 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4000 mask_tok=120.00 y_norm=48.0874±0.0057 +2026-05-10 22:10:58,943 - INFO - Step 17660 | mode=video loss=0.092285 diff=0.092285 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0367 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0870 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=14129 i_micro=3531 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4267 mask_tok=128.00 y_norm=48.0870±0.0060 +2026-05-10 22:11:16,923 - INFO - Step 17670 | mode=video loss=0.111328 diff=0.111328 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1107 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0872 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=14137 i_micro=3533 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.2700 mask_tok=81.00 y_norm=48.0872±0.0054 +2026-05-10 22:11:35,603 - INFO - Step 17680 | mode=video loss=0.054199 diff=0.054199 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0424 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0872 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=14145 i_micro=3535 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3400 mask_tok=102.00 y_norm=48.0872±0.0058 +2026-05-10 22:11:53,983 - INFO - Step 17690 | mode=video loss=0.090332 diff=0.090332 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0354 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0898 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=14153 i_micro=3537 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3333 mask_tok=100.00 y_norm=48.0898±0.0059 +2026-05-10 22:12:11,951 - INFO - Step 17700 | mode=video loss=0.213867 diff=0.213867 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.2544 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0886 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=14161 i_micro=3539 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4133 mask_tok=124.00 y_norm=48.0886±0.0072 +2026-05-10 22:12:18,801 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 22:12:36,766 - INFO - Step 17710 | mode=video loss=0.050781 diff=0.050781 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0489 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0880 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=14169 i_micro=3541 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.2867 mask_tok=86.00 y_norm=48.0880±0.0047 +2026-05-10 22:12:55,344 - INFO - Step 17720 | mode=video loss=0.079590 diff=0.079590 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0312 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0875 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=14177 i_micro=3543 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4833 mask_tok=145.00 y_norm=48.0875±0.0066 +2026-05-10 22:13:14,058 - INFO - Step 17730 | mode=video loss=0.060059 diff=0.060059 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0412 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0879 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=14185 i_micro=3545 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3900 mask_tok=117.00 y_norm=48.0879±0.0053 +2026-05-10 22:13:32,461 - INFO - Step 17740 | mode=video loss=0.090332 diff=0.090332 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0349 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0873 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=14193 i_micro=3547 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.1933 mask_tok=58.00 y_norm=48.0873±0.0056 +2026-05-10 22:13:50,437 - INFO - Step 17750 | mode=video loss=0.114258 diff=0.114258 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0479 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0895 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=14201 i_micro=3549 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3867 mask_tok=116.00 y_norm=48.0895±0.0069 +2026-05-10 22:14:09,174 - INFO - Step 17760 | mode=video loss=0.100586 diff=0.100586 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0409 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0868 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=14209 i_micro=3551 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3433 mask_tok=103.00 y_norm=48.0868±0.0064 +2026-05-10 22:14:27,577 - INFO - Step 17770 | mode=video loss=0.090820 diff=0.090820 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0375 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0882 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=14217 i_micro=3553 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3200 mask_tok=96.00 y_norm=48.0882±0.0049 +2026-05-10 22:14:46,325 - INFO - Step 17780 | mode=video loss=0.125977 diff=0.125977 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0413 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0892 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=14225 i_micro=3555 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3333 mask_tok=100.00 y_norm=48.0892±0.0058 +2026-05-10 22:15:04,747 - INFO - Step 17790 | mode=video loss=0.149414 diff=0.149414 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0598 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0881 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=14233 i_micro=3557 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3433 mask_tok=103.00 y_norm=48.0881±0.0056 +2026-05-10 22:15:24,340 - INFO - Step 17800 | mode=video loss=0.116211 diff=0.116211 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0512 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0869 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=14241 i_micro=3559 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3533 mask_tok=106.00 y_norm=48.0869±0.0068 cond_uncond_dloss=0.300781 cond_grad=0.000047 cond_pred_l2=27.878283 cond_pred_ratio=0.016265 +2026-05-10 22:15:24,377 - INFO - Step 17800 | probe_semantic mcp_offdiag(mean/min/max)=0.644991/0.531798/0.707170 smol_offdiag=0.660557 prompts=4 mcp_tok=17.50 smol_tok=17.50 +2026-05-10 22:15:31,312 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 22:15:50,031 - INFO - Step 17810 | mode=video loss=0.074219 diff=0.074219 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0577 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0883 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=14249 i_micro=3561 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3033 mask_tok=91.00 y_norm=48.0883±0.0048 +2026-05-10 22:16:08,440 - INFO - Step 17820 | mode=video loss=0.086914 diff=0.086914 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0394 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0875 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=14257 i_micro=3563 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3633 mask_tok=109.00 y_norm=48.0875±0.0061 +2026-05-10 22:16:27,064 - INFO - Step 17830 | mode=video loss=0.061035 diff=0.061035 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0429 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0885 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=14265 i_micro=3565 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3633 mask_tok=109.00 y_norm=48.0885±0.0071 +2026-05-10 22:16:45,053 - INFO - Step 17840 | mode=video loss=0.178711 diff=0.178711 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0360 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0863 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=14273 i_micro=3567 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4100 mask_tok=123.00 y_norm=48.0863±0.0059 +2026-05-10 22:17:03,502 - INFO - Step 17850 | mode=video loss=0.047852 diff=0.047852 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0541 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0858 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=14281 i_micro=3569 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3967 mask_tok=119.00 y_norm=48.0858±0.0055 +2026-05-10 22:17:22,168 - INFO - Step 17860 | mode=video loss=0.127930 diff=0.127930 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0788 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0867 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=14289 i_micro=3571 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3400 mask_tok=102.00 y_norm=48.0867±0.0052 +2026-05-10 22:17:40,574 - INFO - Step 17870 | mode=video loss=0.062500 diff=0.062500 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0270 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0878 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=14297 i_micro=3573 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3500 mask_tok=105.00 y_norm=48.0878±0.0054 +2026-05-10 22:17:58,520 - INFO - Step 17880 | mode=video loss=0.117676 diff=0.117676 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0265 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0882 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=14305 i_micro=3575 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3433 mask_tok=103.00 y_norm=48.0882±0.0055 +2026-05-10 22:18:17,153 - INFO - Step 17890 | mode=video loss=0.067383 diff=0.067383 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0432 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0871 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=14313 i_micro=3577 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3500 mask_tok=105.00 y_norm=48.0871±0.0067 +2026-05-10 22:18:35,538 - INFO - Step 17900 | mode=video loss=0.085449 diff=0.085449 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0657 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0877 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=14321 i_micro=3579 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3467 mask_tok=104.00 y_norm=48.0877±0.0058 +2026-05-10 22:18:42,304 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 22:19:00,864 - INFO - Step 17910 | mode=video loss=0.102051 diff=0.102051 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0477 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0886 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=14329 i_micro=3581 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3500 mask_tok=105.00 y_norm=48.0886±0.0054 +2026-05-10 22:19:19,247 - INFO - Step 17920 | mode=video loss=0.093750 diff=0.093750 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0373 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0866 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=14337 i_micro=3583 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3867 mask_tok=116.00 y_norm=48.0866±0.0065 +2026-05-10 22:19:37,214 - INFO - Step 17930 | mode=video loss=0.070312 diff=0.070312 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0267 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0857 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=14345 i_micro=3585 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3867 mask_tok=116.00 y_norm=48.0857±0.0073 +2026-05-10 22:19:55,802 - INFO - Step 17940 | mode=video loss=0.109375 diff=0.109375 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0629 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0871 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=14353 i_micro=3587 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3733 mask_tok=112.00 y_norm=48.0871±0.0063 +2026-05-10 22:20:14,191 - INFO - Step 17950 | mode=video loss=0.218750 diff=0.218750 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0513 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0858 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=14361 i_micro=3589 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3233 mask_tok=97.00 y_norm=48.0858±0.0062 +2026-05-10 22:20:32,833 - INFO - Step 17960 | mode=video loss=0.067871 diff=0.067871 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0506 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0883 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=14369 i_micro=3591 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3333 mask_tok=100.00 y_norm=48.0883±0.0053 +2026-05-10 22:20:50,806 - INFO - Step 17970 | mode=video loss=0.085449 diff=0.085449 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0367 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0870 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=14377 i_micro=3593 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3500 mask_tok=105.00 y_norm=48.0870±0.0060 +2026-05-10 22:21:09,231 - INFO - Step 17980 | mode=video loss=0.083496 diff=0.083496 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0370 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0888 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=14385 i_micro=3595 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3100 mask_tok=93.00 y_norm=48.0888±0.0053 +2026-05-10 22:21:27,894 - INFO - Step 17990 | mode=video loss=0.180664 diff=0.180664 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0998 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0888 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=14393 i_micro=3597 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3533 mask_tok=106.00 y_norm=48.0888±0.0058 +2026-05-10 22:21:47,892 - INFO - Step 18000 | mode=video loss=0.122070 diff=0.122070 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0358 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0883 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=14401 i_micro=3599 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4633 mask_tok=139.00 y_norm=48.0883±0.0065 cond_uncond_dloss=0.332031 cond_grad=0.000236 cond_pred_l2=50.571117 cond_pred_ratio=0.032456 +2026-05-10 22:21:47,929 - INFO - Step 18000 | probe_semantic mcp_offdiag(mean/min/max)=0.645178/0.531941/0.707244 smol_offdiag=0.660557 prompts=4 mcp_tok=17.50 smol_tok=17.50 +2026-05-10 22:21:54,870 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 22:22:13,626 - INFO - Step 18010 | mode=video loss=0.074707 diff=0.074707 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0515 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0874 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=14409 i_micro=3601 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.2533 mask_tok=76.00 y_norm=48.0874±0.0051 +2026-05-10 22:22:31,625 - INFO - Step 18020 | mode=video loss=0.086914 diff=0.086914 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0750 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0876 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=14417 i_micro=3603 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4533 mask_tok=136.00 y_norm=48.0876±0.0064 +2026-05-10 22:22:50,011 - INFO - Step 18030 | mode=video loss=0.118164 diff=0.118164 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0385 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0885 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=14425 i_micro=3605 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3533 mask_tok=106.00 y_norm=48.0885±0.0055 +2026-05-10 22:23:08,702 - INFO - Step 18040 | mode=video loss=0.097168 diff=0.097168 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0374 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0876 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=14433 i_micro=3607 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3567 mask_tok=107.00 y_norm=48.0876±0.0053 +2026-05-10 22:23:27,123 - INFO - Step 18050 | mode=video loss=0.087402 diff=0.087402 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0686 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0888 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=14441 i_micro=3609 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3800 mask_tok=114.00 y_norm=48.0888±0.0056 +2026-05-10 22:23:45,100 - INFO - Step 18060 | mode=video loss=0.125977 diff=0.125977 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0449 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0869 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=14449 i_micro=3611 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.2800 mask_tok=84.00 y_norm=48.0869±0.0056 +2026-05-10 22:24:03,770 - INFO - Step 18070 | mode=video loss=0.089355 diff=0.089355 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0545 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0864 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=14457 i_micro=3613 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3467 mask_tok=104.00 y_norm=48.0864±0.0067 +2026-05-10 22:24:22,144 - INFO - Step 18080 | mode=video loss=0.099121 diff=0.099121 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0308 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0884 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=14465 i_micro=3615 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3667 mask_tok=110.00 y_norm=48.0884±0.0059 +2026-05-10 22:24:40,741 - INFO - Step 18090 | mode=video loss=0.083984 diff=0.083984 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0219 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0873 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=14473 i_micro=3617 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3433 mask_tok=103.00 y_norm=48.0873±0.0059 +2026-05-10 22:24:58,700 - INFO - Step 18100 | mode=video loss=0.100098 diff=0.100098 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0876 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0873 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=14481 i_micro=3619 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4467 mask_tok=134.00 y_norm=48.0873±0.0059 +2026-05-10 22:25:05,612 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 22:25:24,005 - INFO - Step 18110 | mode=video loss=0.082520 diff=0.082520 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0667 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0874 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=14489 i_micro=3621 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3633 mask_tok=109.00 y_norm=48.0874±0.0053 +2026-05-10 22:25:42,509 - INFO - Step 18120 | mode=video loss=0.117188 diff=0.117188 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1426 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0884 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=14497 i_micro=3623 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3100 mask_tok=93.00 y_norm=48.0884±0.0054 +2026-05-10 22:26:00,880 - INFO - Step 18130 | mode=video loss=0.080566 diff=0.080566 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0405 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0879 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=14505 i_micro=3625 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3033 mask_tok=91.00 y_norm=48.0879±0.0064 +2026-05-10 22:26:19,498 - INFO - Step 18140 | mode=video loss=0.079102 diff=0.079102 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0559 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0843 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=14513 i_micro=3627 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3400 mask_tok=102.00 y_norm=48.0843±0.0071 +2026-05-10 22:26:37,466 - INFO - Step 18150 | mode=video loss=0.107910 diff=0.107910 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0254 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0865 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=14521 i_micro=3629 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3600 mask_tok=108.00 y_norm=48.0865±0.0050 +2026-05-10 22:26:55,865 - INFO - Step 18160 | mode=video loss=0.119141 diff=0.119141 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0336 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0866 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=14529 i_micro=3631 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4167 mask_tok=125.00 y_norm=48.0866±0.0060 +2026-05-10 22:27:14,522 - INFO - Step 18170 | mode=video loss=0.053467 diff=0.053467 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0835 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0853 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=14537 i_micro=3633 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4200 mask_tok=126.00 y_norm=48.0853±0.0063 +2026-05-10 22:27:32,912 - INFO - Step 18180 | mode=video loss=0.142578 diff=0.142578 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1002 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0881 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=14545 i_micro=3635 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3467 mask_tok=104.00 y_norm=48.0881±0.0056 +2026-05-10 22:27:51,672 - INFO - Step 18190 | mode=video loss=0.082031 diff=0.082031 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1259 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0871 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=14553 i_micro=3637 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4267 mask_tok=128.00 y_norm=48.0871±0.0067 +2026-05-10 22:28:11,247 - INFO - Step 18200 | mode=video loss=0.091309 diff=0.091309 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1250 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0866 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=14561 i_micro=3639 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4733 mask_tok=142.00 y_norm=48.0866±0.0061 cond_uncond_dloss=1.257812 cond_grad=0.001310 cond_pred_l2=50.877861 cond_pred_ratio=0.032228 +2026-05-10 22:28:11,284 - INFO - Step 18200 | probe_semantic mcp_offdiag(mean/min/max)=0.645329/0.532100/0.707253 smol_offdiag=0.660557 prompts=4 mcp_tok=17.50 smol_tok=17.50 +2026-05-10 22:28:18,170 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 22:28:36,531 - INFO - Step 18210 | mode=video loss=0.062500 diff=0.062500 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0278 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0873 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=14569 i_micro=3641 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3433 mask_tok=103.00 y_norm=48.0873±0.0059 +2026-05-10 22:28:55,234 - INFO - Step 18220 | mode=video loss=0.084473 diff=0.084473 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0528 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0892 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=14577 i_micro=3643 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.2667 mask_tok=80.00 y_norm=48.0892±0.0055 +2026-05-10 22:29:13,204 - INFO - Step 18230 | mode=video loss=0.076660 diff=0.076660 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0263 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0880 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=14585 i_micro=3645 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.2700 mask_tok=81.00 y_norm=48.0880±0.0055 +2026-05-10 22:29:31,637 - INFO - Step 18240 | mode=video loss=0.088867 diff=0.088867 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0494 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0879 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=14593 i_micro=3647 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4633 mask_tok=139.00 y_norm=48.0879±0.0071 +2026-05-10 22:29:50,253 - INFO - Step 18250 | mode=video loss=0.129883 diff=0.129883 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0549 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0869 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=14601 i_micro=3649 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4067 mask_tok=122.00 y_norm=48.0869±0.0058 +2026-05-10 22:30:08,617 - INFO - Step 18260 | mode=video loss=0.141602 diff=0.141602 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0643 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0884 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=14609 i_micro=3651 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.2733 mask_tok=82.00 y_norm=48.0884±0.0054 +2026-05-10 22:30:27,192 - INFO - Step 18270 | mode=video loss=0.176758 diff=0.176758 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1067 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0872 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=14617 i_micro=3653 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3667 mask_tok=110.00 y_norm=48.0872±0.0057 +2026-05-10 22:30:45,166 - INFO - Step 18280 | mode=video loss=0.110352 diff=0.110352 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0676 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0863 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=14625 i_micro=3655 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3533 mask_tok=106.00 y_norm=48.0863±0.0061 +2026-05-10 22:31:03,540 - INFO - Step 18290 | mode=video loss=0.119629 diff=0.119629 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0307 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0885 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=14633 i_micro=3657 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3400 mask_tok=102.00 y_norm=48.0885±0.0064 +2026-05-10 22:31:22,122 - INFO - Step 18300 | mode=video loss=0.200195 diff=0.200195 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0939 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0868 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=14641 i_micro=3659 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3700 mask_tok=111.00 y_norm=48.0868±0.0057 +2026-05-10 22:31:28,854 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 22:31:47,220 - INFO - Step 18310 | mode=video loss=0.107910 diff=0.107910 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0574 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0874 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=14649 i_micro=3661 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3367 mask_tok=101.00 y_norm=48.0874±0.0057 +2026-05-10 22:32:05,914 - INFO - Step 18320 | mode=video loss=0.084473 diff=0.084473 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0708 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0882 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=14657 i_micro=3663 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.2933 mask_tok=88.00 y_norm=48.0882±0.0052 +2026-05-10 22:32:23,887 - INFO - Step 18330 | mode=video loss=0.089355 diff=0.089355 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0334 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0878 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=14665 i_micro=3665 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3667 mask_tok=110.00 y_norm=48.0878±0.0074 +2026-05-10 22:32:42,268 - INFO - Step 18340 | mode=video loss=0.466797 diff=0.466797 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0548 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0877 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=14673 i_micro=3667 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3500 mask_tok=105.00 y_norm=48.0877±0.0069 +2026-05-10 22:33:00,893 - INFO - Step 18350 | mode=video loss=0.169922 diff=0.169922 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1435 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0884 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=14681 i_micro=3669 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3200 mask_tok=96.00 y_norm=48.0884±0.0058 +2026-05-10 22:33:18,840 - INFO - Step 18360 | mode=video loss=0.095215 diff=0.095215 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0277 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0866 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=14689 i_micro=3671 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3533 mask_tok=106.00 y_norm=48.0866±0.0069 +2026-05-10 22:33:37,220 - INFO - Step 18370 | mode=video loss=0.101562 diff=0.101562 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1425 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0893 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=14697 i_micro=3673 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.2833 mask_tok=85.00 y_norm=48.0893±0.0054 +2026-05-10 22:33:55,920 - INFO - Step 18380 | mode=video loss=0.094727 diff=0.094727 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0925 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0874 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=14705 i_micro=3675 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3500 mask_tok=105.00 y_norm=48.0874±0.0063 +2026-05-10 22:34:14,294 - INFO - Step 18390 | mode=video loss=0.065918 diff=0.065918 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0673 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0886 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=14713 i_micro=3677 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3733 mask_tok=112.00 y_norm=48.0886±0.0059 +2026-05-10 22:34:34,485 - INFO - Step 18400 | mode=video loss=0.098633 diff=0.098633 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0314 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0873 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=14721 i_micro=3679 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3367 mask_tok=101.00 y_norm=48.0873±0.0060 cond_uncond_dloss=0.421875 cond_grad=0.000021 cond_pred_l2=22.571806 cond_pred_ratio=0.013256 +2026-05-10 22:34:34,522 - INFO - Step 18400 | probe_semantic mcp_offdiag(mean/min/max)=0.645265/0.532049/0.707179 smol_offdiag=0.660557 prompts=4 mcp_tok=17.50 smol_tok=17.50 +2026-05-10 22:34:41,327 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 22:34:59,267 - INFO - Step 18410 | mode=video loss=0.093750 diff=0.093750 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0343 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0890 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=14729 i_micro=3681 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3700 mask_tok=111.00 y_norm=48.0890±0.0053 +2026-05-10 22:35:17,635 - INFO - Step 18420 | mode=video loss=0.065430 diff=0.065430 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0489 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0867 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=14737 i_micro=3683 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4800 mask_tok=144.00 y_norm=48.0867±0.0067 +2026-05-10 22:35:36,383 - INFO - Step 18430 | mode=video loss=0.120605 diff=0.120605 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0435 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0866 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=14745 i_micro=3685 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4333 mask_tok=130.00 y_norm=48.0866±0.0064 +2026-05-10 22:35:54,811 - INFO - Step 18440 | mode=video loss=0.089355 diff=0.089355 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1401 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0872 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=14753 i_micro=3687 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3533 mask_tok=106.00 y_norm=48.0872±0.0060 +2026-05-10 22:36:13,436 - INFO - Step 18450 | mode=video loss=0.083984 diff=0.083984 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0321 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0886 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=14761 i_micro=3689 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.2400 mask_tok=72.00 y_norm=48.0886±0.0054 +2026-05-10 22:36:31,402 - INFO - Step 18460 | mode=video loss=0.073242 diff=0.073242 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0474 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0873 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=14769 i_micro=3691 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4133 mask_tok=124.00 y_norm=48.0873±0.0054 +2026-05-10 22:36:49,786 - INFO - Step 18470 | mode=video loss=0.092285 diff=0.092285 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0558 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0891 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=14777 i_micro=3693 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4100 mask_tok=123.00 y_norm=48.0891±0.0060 +2026-05-10 22:37:08,435 - INFO - Step 18480 | mode=video loss=0.092773 diff=0.092773 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0354 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0868 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=14785 i_micro=3695 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4533 mask_tok=136.00 y_norm=48.0868±0.0057 +2026-05-10 22:37:26,397 - INFO - Step 18490 | mode=video loss=0.078613 diff=0.078613 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0613 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0871 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=14793 i_micro=3697 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3567 mask_tok=107.00 y_norm=48.0871±0.0061 +2026-05-10 22:37:44,775 - INFO - Step 18500 | mode=video loss=0.051270 diff=0.051270 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0266 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0883 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=14801 i_micro=3699 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.2867 mask_tok=86.00 y_norm=48.0883±0.0055 +2026-05-10 22:37:51,731 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 22:38:10,385 - INFO - Step 18510 | mode=video loss=0.115234 diff=0.115234 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0868 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0873 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=14809 i_micro=3701 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3567 mask_tok=107.00 y_norm=48.0873±0.0061 +2026-05-10 22:38:28,774 - INFO - Step 18520 | mode=video loss=0.115723 diff=0.115723 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1705 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0884 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=14817 i_micro=3703 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3667 mask_tok=110.00 y_norm=48.0884±0.0064 +2026-05-10 22:38:47,424 - INFO - Step 18530 | mode=video loss=0.063477 diff=0.063477 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0472 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0858 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=14825 i_micro=3705 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4867 mask_tok=146.00 y_norm=48.0858±0.0062 +2026-05-10 22:39:05,392 - INFO - Step 18540 | mode=video loss=0.075195 diff=0.075195 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0532 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0869 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=14833 i_micro=3707 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.8867 mask_tok=266.00 y_norm=48.0869±0.0083 +2026-05-10 22:39:23,807 - INFO - Step 18550 | mode=video loss=0.095215 diff=0.095215 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0433 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0865 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=14841 i_micro=3709 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3333 mask_tok=100.00 y_norm=48.0865±0.0053 +2026-05-10 22:39:42,493 - INFO - Step 18560 | mode=video loss=0.079590 diff=0.079590 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0375 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0874 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=14849 i_micro=3711 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.5167 mask_tok=155.00 y_norm=48.0874±0.0070 +2026-05-10 22:40:00,914 - INFO - Step 18570 | mode=video loss=0.057129 diff=0.057129 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0360 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0861 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=14857 i_micro=3713 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4000 mask_tok=120.00 y_norm=48.0861±0.0059 +2026-05-10 22:40:19,434 - INFO - Step 18580 | mode=video loss=0.099609 diff=0.099609 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0233 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0880 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=14865 i_micro=3715 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.5467 mask_tok=164.00 y_norm=48.0880±0.0074 +2026-05-10 22:40:37,386 - INFO - Step 18590 | mode=video loss=0.175781 diff=0.175781 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0314 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0888 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=14873 i_micro=3717 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.1767 mask_tok=53.00 y_norm=48.0888±0.0047 +2026-05-10 22:40:57,498 - INFO - Step 18600 | mode=video loss=0.066406 diff=0.066406 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0937 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0860 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=14881 i_micro=3719 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3700 mask_tok=111.00 y_norm=48.0860±0.0063 cond_uncond_dloss=1.312500 cond_grad=0.000321 cond_pred_l2=46.115097 cond_pred_ratio=0.028823 +2026-05-10 22:40:57,536 - INFO - Step 18600 | probe_semantic mcp_offdiag(mean/min/max)=0.645725/0.532869/0.707558 smol_offdiag=0.660557 prompts=4 mcp_tok=17.50 smol_tok=17.50 +2026-05-10 22:41:04,415 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 22:41:22,883 - INFO - Step 18610 | mode=video loss=0.089355 diff=0.089355 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0260 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0863 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=14889 i_micro=3721 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3400 mask_tok=102.00 y_norm=48.0863±0.0056 +2026-05-10 22:41:41,343 - INFO - Step 18620 | mode=video loss=0.193359 diff=0.193359 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0776 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0856 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=14897 i_micro=3723 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3833 mask_tok=115.00 y_norm=48.0856±0.0059 +2026-05-10 22:42:00,035 - INFO - Step 18630 | mode=video loss=0.124023 diff=0.124023 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0267 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0861 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=14905 i_micro=3725 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3433 mask_tok=103.00 y_norm=48.0861±0.0061 +2026-05-10 22:42:17,994 - INFO - Step 18640 | mode=video loss=0.086426 diff=0.086426 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0553 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0873 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=14913 i_micro=3727 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3667 mask_tok=110.00 y_norm=48.0873±0.0066 +2026-05-10 22:42:36,374 - INFO - Step 18650 | mode=video loss=0.092773 diff=0.092773 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0877 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0863 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=14921 i_micro=3729 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3967 mask_tok=119.00 y_norm=48.0863±0.0060 +2026-05-10 22:42:55,095 - INFO - Step 18660 | mode=video loss=0.127930 diff=0.127930 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0658 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0886 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=14929 i_micro=3731 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3233 mask_tok=97.00 y_norm=48.0886±0.0052 +2026-05-10 22:43:13,055 - INFO - Step 18670 | mode=video loss=0.082031 diff=0.082031 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0472 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0890 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=14937 i_micro=3733 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3700 mask_tok=111.00 y_norm=48.0890±0.0058 +2026-05-10 22:43:31,473 - INFO - Step 18680 | mode=video loss=0.111328 diff=0.111328 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0598 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0889 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=14945 i_micro=3735 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3767 mask_tok=113.00 y_norm=48.0889±0.0055 +2026-05-10 22:43:50,048 - INFO - Step 18690 | mode=video loss=0.118164 diff=0.118164 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0388 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0870 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=14953 i_micro=3737 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4333 mask_tok=130.00 y_norm=48.0870±0.0062 +2026-05-10 22:44:08,418 - INFO - Step 18700 | mode=video loss=0.106934 diff=0.106934 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0441 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0884 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=14961 i_micro=3739 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3167 mask_tok=95.00 y_norm=48.0884±0.0062 +2026-05-10 22:44:15,429 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 22:44:33,968 - INFO - Step 18710 | mode=video loss=0.140625 diff=0.140625 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0824 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0866 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=14969 i_micro=3741 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3200 mask_tok=96.00 y_norm=48.0866±0.0050 +2026-05-10 22:44:51,930 - INFO - Step 18720 | mode=video loss=0.057861 diff=0.057861 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1093 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0861 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=14977 i_micro=3743 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3900 mask_tok=117.00 y_norm=48.0861±0.0067 +2026-05-10 22:45:10,344 - INFO - Step 18730 | mode=video loss=0.090332 diff=0.090332 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0266 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0867 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=14985 i_micro=3745 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3433 mask_tok=103.00 y_norm=48.0867±0.0057 +2026-05-10 22:45:28,860 - INFO - Step 18740 | mode=video loss=0.152344 diff=0.152344 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0786 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0868 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=14993 i_micro=3747 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3967 mask_tok=119.00 y_norm=48.0868±0.0069 +2026-05-10 22:45:47,226 - INFO - Step 18750 | mode=video loss=0.067871 diff=0.067871 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0489 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0871 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=15001 i_micro=3749 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4333 mask_tok=130.00 y_norm=48.0871±0.0056 +2026-05-10 22:46:05,954 - INFO - Step 18760 | mode=video loss=0.054688 diff=0.054688 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0588 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0869 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=15009 i_micro=3751 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3600 mask_tok=108.00 y_norm=48.0869±0.0053 +2026-05-10 22:46:23,947 - INFO - Step 18770 | mode=video loss=0.059570 diff=0.059570 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0490 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0855 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=15017 i_micro=3753 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4800 mask_tok=144.00 y_norm=48.0855±0.0061 +2026-05-10 22:46:42,340 - INFO - Step 18780 | mode=video loss=0.077148 diff=0.077148 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0305 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0884 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=15025 i_micro=3755 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3400 mask_tok=102.00 y_norm=48.0884±0.0058 +2026-05-10 22:47:00,836 - INFO - Step 18790 | mode=video loss=0.361328 diff=0.361328 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.4099 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0858 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=15033 i_micro=3757 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3800 mask_tok=114.00 y_norm=48.0858±0.0060 +2026-05-10 22:47:20,830 - INFO - Step 18800 | mode=video loss=0.084473 diff=0.084473 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0458 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0878 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=15041 i_micro=3759 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4433 mask_tok=133.00 y_norm=48.0878±0.0071 cond_uncond_dloss=0.269531 cond_grad=0.000076 cond_pred_l2=35.176548 cond_pred_ratio=0.020307 +2026-05-10 22:47:20,868 - INFO - Step 18800 | probe_semantic mcp_offdiag(mean/min/max)=0.644964/0.531863/0.706822 smol_offdiag=0.660557 prompts=4 mcp_tok=17.50 smol_tok=17.50 +2026-05-10 22:47:27,870 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 22:47:45,820 - INFO - Step 18810 | mode=video loss=0.213867 diff=0.213867 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0563 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0859 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=15049 i_micro=3761 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.5267 mask_tok=158.00 y_norm=48.0859±0.0063 +2026-05-10 22:48:04,520 - INFO - Step 18820 | mode=video loss=0.125977 diff=0.125977 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0966 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0877 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=15057 i_micro=3763 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3467 mask_tok=104.00 y_norm=48.0877±0.0057 +2026-05-10 22:48:23,020 - INFO - Step 18830 | mode=video loss=0.049805 diff=0.049805 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0469 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0862 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=15065 i_micro=3765 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4600 mask_tok=138.00 y_norm=48.0862±0.0067 +2026-05-10 22:48:41,756 - INFO - Step 18840 | mode=video loss=0.165039 diff=0.165039 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0848 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0857 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=15073 i_micro=3767 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3400 mask_tok=102.00 y_norm=48.0857±0.0055 +2026-05-10 22:48:59,713 - INFO - Step 18850 | mode=video loss=0.096680 diff=0.096680 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0392 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0882 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=15081 i_micro=3769 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.8433 mask_tok=253.00 y_norm=48.0882±0.0074 +2026-05-10 22:49:18,101 - INFO - Step 18860 | mode=video loss=0.091797 diff=0.091797 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0588 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0882 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=15089 i_micro=3771 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3967 mask_tok=119.00 y_norm=48.0882±0.0067 +2026-05-10 22:49:36,655 - INFO - Step 18870 | mode=video loss=0.125000 diff=0.125000 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0297 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0875 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=15097 i_micro=3773 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3800 mask_tok=114.00 y_norm=48.0875±0.0055 +2026-05-10 22:49:55,054 - INFO - Step 18880 | mode=video loss=0.068359 diff=0.068359 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0522 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0879 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=15105 i_micro=3775 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3567 mask_tok=107.00 y_norm=48.0879±0.0061 +2026-05-10 22:50:13,709 - INFO - Step 18890 | mode=video loss=0.122070 diff=0.122070 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0425 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0885 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=15113 i_micro=3777 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3700 mask_tok=111.00 y_norm=48.0885±0.0060 +2026-05-10 22:50:31,664 - INFO - Step 18900 | mode=video loss=0.109375 diff=0.109375 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0634 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0894 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=15121 i_micro=3779 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3533 mask_tok=106.00 y_norm=48.0894±0.0060 +2026-05-10 22:50:38,540 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 22:50:56,901 - INFO - Step 18910 | mode=video loss=0.056641 diff=0.056641 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0249 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0861 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=15129 i_micro=3781 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3433 mask_tok=103.00 y_norm=48.0861±0.0062 +2026-05-10 22:51:15,572 - INFO - Step 18920 | mode=video loss=0.101074 diff=0.101074 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0346 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0862 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=15137 i_micro=3783 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3533 mask_tok=106.00 y_norm=48.0862±0.0061 +2026-05-10 22:51:33,525 - INFO - Step 18930 | mode=video loss=0.110840 diff=0.110840 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1288 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0872 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=15145 i_micro=3785 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4933 mask_tok=148.00 y_norm=48.0872±0.0075 +2026-05-10 22:51:52,016 - INFO - Step 18940 | mode=video loss=0.075684 diff=0.075684 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0308 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0897 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=15153 i_micro=3787 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.2867 mask_tok=86.00 y_norm=48.0897±0.0053 +2026-05-10 22:52:10,469 - INFO - Step 18950 | mode=video loss=0.124023 diff=0.124023 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0983 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0875 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=15161 i_micro=3789 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3633 mask_tok=109.00 y_norm=48.0875±0.0052 +2026-05-10 22:52:28,895 - INFO - Step 18960 | mode=video loss=0.104980 diff=0.104980 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0987 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0877 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=15169 i_micro=3791 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3467 mask_tok=104.00 y_norm=48.0877±0.0047 +2026-05-10 22:52:47,568 - INFO - Step 18970 | mode=video loss=0.068848 diff=0.068848 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0680 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0865 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=15177 i_micro=3793 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4667 mask_tok=140.00 y_norm=48.0865±0.0061 +2026-05-10 22:53:05,530 - INFO - Step 18980 | mode=video loss=0.130859 diff=0.130859 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0507 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0882 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=15185 i_micro=3795 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3267 mask_tok=98.00 y_norm=48.0882±0.0060 +2026-05-10 22:53:23,952 - INFO - Step 18990 | mode=video loss=0.069336 diff=0.069336 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0263 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0877 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=15193 i_micro=3797 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4500 mask_tok=135.00 y_norm=48.0877±0.0068 +2026-05-10 22:53:44,271 - INFO - Step 19000 | mode=video loss=0.102539 diff=0.102539 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0799 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0880 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=15201 i_micro=3799 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3400 mask_tok=102.00 y_norm=48.0880±0.0050 cond_uncond_dloss=0.421875 cond_grad=0.000163 cond_pred_l2=37.944351 cond_pred_ratio=0.022902 +2026-05-10 22:53:44,308 - INFO - Step 19000 | probe_semantic mcp_offdiag(mean/min/max)=0.645218/0.532042/0.707219 smol_offdiag=0.660557 prompts=4 mcp_tok=17.50 smol_tok=17.50 +2026-05-10 22:53:51,891 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 22:54:10,268 - INFO - Step 19010 | mode=video loss=0.084961 diff=0.084961 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0616 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0878 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=15209 i_micro=3801 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3267 mask_tok=98.00 y_norm=48.0878±0.0047 +2026-05-10 22:54:28,796 - INFO - Step 19020 | mode=video loss=0.110352 diff=0.110352 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1108 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0864 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=15217 i_micro=3803 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3300 mask_tok=99.00 y_norm=48.0864±0.0061 +2026-05-10 22:54:46,758 - INFO - Step 19030 | mode=video loss=0.082031 diff=0.082031 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0672 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0881 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=15225 i_micro=3805 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3300 mask_tok=99.00 y_norm=48.0881±0.0057 +2026-05-10 22:55:05,143 - INFO - Step 19040 | mode=video loss=0.153320 diff=0.153320 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0659 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0855 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=15233 i_micro=3807 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4600 mask_tok=138.00 y_norm=48.0855±0.0064 +2026-05-10 22:55:23,788 - INFO - Step 19050 | mode=video loss=0.074219 diff=0.074219 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0234 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0857 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=15241 i_micro=3809 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4267 mask_tok=128.00 y_norm=48.0857±0.0059 +2026-05-10 22:55:41,741 - INFO - Step 19060 | mode=video loss=0.076660 diff=0.076660 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0516 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0868 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=15249 i_micro=3811 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3367 mask_tok=101.00 y_norm=48.0868±0.0058 +2026-05-10 22:56:00,141 - INFO - Step 19070 | mode=video loss=0.095215 diff=0.095215 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0368 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0888 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=15257 i_micro=3813 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3933 mask_tok=118.00 y_norm=48.0888±0.0056 +2026-05-10 22:56:18,682 - INFO - Step 19080 | mode=video loss=0.172852 diff=0.172852 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0833 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0885 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=15265 i_micro=3815 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3333 mask_tok=100.00 y_norm=48.0885±0.0053 +2026-05-10 22:56:37,096 - INFO - Step 19090 | mode=video loss=0.073730 diff=0.073730 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0506 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0890 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=15273 i_micro=3817 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3567 mask_tok=107.00 y_norm=48.0890±0.0062 +2026-05-10 22:56:55,680 - INFO - Step 19100 | mode=video loss=0.063477 diff=0.063477 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0384 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0870 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=15281 i_micro=3819 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3567 mask_tok=107.00 y_norm=48.0870±0.0059 +2026-05-10 22:57:02,768 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 22:57:20,716 - INFO - Step 19110 | mode=video loss=0.100098 diff=0.100098 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0445 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0887 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=15289 i_micro=3821 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3600 mask_tok=108.00 y_norm=48.0887±0.0053 +2026-05-10 22:57:39,090 - INFO - Step 19120 | mode=video loss=0.083984 diff=0.083984 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0670 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0856 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=15297 i_micro=3823 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4700 mask_tok=141.00 y_norm=48.0856±0.0066 +2026-05-10 22:57:57,563 - INFO - Step 19130 | mode=video loss=0.110840 diff=0.110840 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0337 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0878 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=15305 i_micro=3825 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3433 mask_tok=103.00 y_norm=48.0878±0.0056 +2026-05-10 22:58:15,938 - INFO - Step 19140 | mode=video loss=0.139648 diff=0.139648 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0777 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0888 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=15313 i_micro=3827 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.2733 mask_tok=82.00 y_norm=48.0888±0.0060 +2026-05-10 22:58:34,377 - INFO - Step 19150 | mode=video loss=0.185547 diff=0.185547 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0545 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0868 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=15321 i_micro=3829 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3967 mask_tok=119.00 y_norm=48.0868±0.0060 +2026-05-10 22:58:52,345 - INFO - Step 19160 | mode=video loss=0.092285 diff=0.092285 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0627 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0887 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=15329 i_micro=3831 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3567 mask_tok=107.00 y_norm=48.0887±0.0051 +2026-05-10 22:59:10,728 - INFO - Step 19170 | mode=video loss=0.090332 diff=0.090332 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0635 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0869 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=15337 i_micro=3833 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3433 mask_tok=103.00 y_norm=48.0869±0.0059 +2026-05-10 22:59:29,190 - INFO - Step 19180 | mode=video loss=0.160156 diff=0.160156 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0721 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0870 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=15345 i_micro=3835 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3633 mask_tok=109.00 y_norm=48.0870±0.0059 +2026-05-10 22:59:47,559 - INFO - Step 19190 | mode=video loss=0.071289 diff=0.071289 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0347 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0865 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=15353 i_micro=3837 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3567 mask_tok=107.00 y_norm=48.0865±0.0054 +2026-05-10 23:00:07,148 - INFO - Step 19200 | mode=video loss=0.099609 diff=0.099609 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0950 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0888 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=15361 i_micro=3839 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3333 mask_tok=100.00 y_norm=48.0888±0.0055 cond_uncond_dloss=0.929688 cond_grad=0.000710 cond_pred_l2=46.788204 cond_pred_ratio=0.029381 +2026-05-10 23:00:07,186 - INFO - Step 19200 | probe_semantic mcp_offdiag(mean/min/max)=0.645019/0.531776/0.707122 smol_offdiag=0.660557 prompts=4 mcp_tok=17.50 smol_tok=17.50 +2026-05-10 23:00:14,152 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 23:00:32,732 - INFO - Step 19210 | mode=video loss=0.068848 diff=0.068848 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0357 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0862 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=15369 i_micro=3841 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4667 mask_tok=140.00 y_norm=48.0862±0.0062 +2026-05-10 23:00:51,132 - INFO - Step 19220 | mode=video loss=0.096191 diff=0.096191 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0736 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0869 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=15377 i_micro=3843 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3667 mask_tok=110.00 y_norm=48.0869±0.0055 +2026-05-10 23:01:09,784 - INFO - Step 19230 | mode=video loss=0.114746 diff=0.114746 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0438 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0868 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=15385 i_micro=3845 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3567 mask_tok=107.00 y_norm=48.0868±0.0065 +2026-05-10 23:01:27,739 - INFO - Step 19240 | mode=video loss=0.077637 diff=0.077637 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0413 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0877 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=15393 i_micro=3847 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3600 mask_tok=108.00 y_norm=48.0877±0.0055 +2026-05-10 23:01:46,112 - INFO - Step 19250 | mode=video loss=0.215820 diff=0.215820 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1571 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0887 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=15401 i_micro=3849 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3333 mask_tok=100.00 y_norm=48.0887±0.0062 +2026-05-10 23:02:04,780 - INFO - Step 19260 | mode=video loss=0.096191 diff=0.096191 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0553 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0887 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=15409 i_micro=3851 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3567 mask_tok=107.00 y_norm=48.0887±0.0057 +2026-05-10 23:02:23,170 - INFO - Step 19270 | mode=video loss=0.076172 diff=0.076172 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0261 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0887 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=15417 i_micro=3853 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3300 mask_tok=99.00 y_norm=48.0887±0.0057 +2026-05-10 23:02:41,828 - INFO - Step 19280 | mode=video loss=0.072266 diff=0.072266 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0283 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0861 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=15425 i_micro=3855 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3267 mask_tok=98.00 y_norm=48.0861±0.0062 +2026-05-10 23:02:59,786 - INFO - Step 19290 | mode=video loss=0.137695 diff=0.137695 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0449 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0868 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=15433 i_micro=3857 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3500 mask_tok=105.00 y_norm=48.0868±0.0060 +2026-05-10 23:03:18,146 - INFO - Step 19300 | mode=video loss=0.070312 diff=0.070312 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0423 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0868 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=15441 i_micro=3859 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3433 mask_tok=103.00 y_norm=48.0868±0.0070 +2026-05-10 23:03:25,406 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 23:03:43,892 - INFO - Step 19310 | mode=video loss=0.066895 diff=0.066895 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0339 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0895 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=15449 i_micro=3861 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.5467 mask_tok=164.00 y_norm=48.0895±0.0074 +2026-05-10 23:04:02,249 - INFO - Step 19320 | mode=video loss=0.070801 diff=0.070801 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0355 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0871 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=15457 i_micro=3863 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.5100 mask_tok=153.00 y_norm=48.0871±0.0072 +2026-05-10 23:04:20,219 - INFO - Step 19330 | mode=video loss=0.145508 diff=0.145508 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0346 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0852 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=15465 i_micro=3865 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3700 mask_tok=111.00 y_norm=48.0852±0.0064 +2026-05-10 23:04:38,750 - INFO - Step 19340 | mode=video loss=0.076660 diff=0.076660 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0518 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0883 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=15473 i_micro=3867 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3533 mask_tok=106.00 y_norm=48.0883±0.0056 +2026-05-10 23:04:57,196 - INFO - Step 19350 | mode=video loss=0.133789 diff=0.133789 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.2174 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0887 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=15481 i_micro=3869 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3433 mask_tok=103.00 y_norm=48.0887±0.0056 +2026-05-10 23:05:15,864 - INFO - Step 19360 | mode=video loss=0.062500 diff=0.062500 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0919 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0872 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=15489 i_micro=3871 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4433 mask_tok=133.00 y_norm=48.0872±0.0053 +2026-05-10 23:05:33,833 - INFO - Step 19370 | mode=video loss=0.143555 diff=0.143555 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0629 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0893 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=15497 i_micro=3873 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3067 mask_tok=92.00 y_norm=48.0893±0.0052 +2026-05-10 23:05:52,241 - INFO - Step 19380 | mode=video loss=0.132812 diff=0.132812 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0376 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0863 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=15505 i_micro=3875 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4167 mask_tok=125.00 y_norm=48.0863±0.0060 +2026-05-10 23:06:10,856 - INFO - Step 19390 | mode=video loss=0.079102 diff=0.079102 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0417 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0866 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=15513 i_micro=3877 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3800 mask_tok=114.00 y_norm=48.0866±0.0064 +2026-05-10 23:06:30,862 - INFO - Step 19400 | mode=video loss=0.054688 diff=0.054688 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0491 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0880 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=15521 i_micro=3879 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3467 mask_tok=104.00 y_norm=48.0880±0.0059 cond_uncond_dloss=0.332031 cond_grad=0.000118 cond_pred_l2=37.856728 cond_pred_ratio=0.022196 +2026-05-10 23:06:30,899 - INFO - Step 19400 | probe_semantic mcp_offdiag(mean/min/max)=0.645129/0.531859/0.707344 smol_offdiag=0.660557 prompts=4 mcp_tok=17.50 smol_tok=17.50 +2026-05-10 23:06:40,396 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 23:06:59,262 - INFO - Step 19410 | mode=video loss=0.108887 diff=0.108887 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0357 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0862 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=15529 i_micro=3881 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3500 mask_tok=105.00 y_norm=48.0862±0.0060 +2026-05-10 23:07:17,239 - INFO - Step 19420 | mode=video loss=0.050537 diff=0.050537 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0324 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0852 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=15537 i_micro=3883 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.6000 mask_tok=180.00 y_norm=48.0852±0.0078 +2026-05-10 23:07:35,630 - INFO - Step 19430 | mode=video loss=0.088379 diff=0.088379 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0429 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0887 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=15545 i_micro=3885 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3533 mask_tok=106.00 y_norm=48.0887±0.0054 +2026-05-10 23:07:54,425 - INFO - Step 19440 | mode=video loss=0.065430 diff=0.065430 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0398 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0878 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=15553 i_micro=3887 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.5267 mask_tok=158.00 y_norm=48.0878±0.0065 +2026-05-10 23:08:12,801 - INFO - Step 19450 | mode=video loss=0.110352 diff=0.110352 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0376 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0882 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=15561 i_micro=3889 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3400 mask_tok=102.00 y_norm=48.0882±0.0058 +2026-05-10 23:08:30,756 - INFO - Step 19460 | mode=video loss=0.104004 diff=0.104004 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0230 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0896 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=15569 i_micro=3891 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4133 mask_tok=124.00 y_norm=48.0896±0.0061 +2026-05-10 23:08:49,374 - INFO - Step 19470 | mode=video loss=0.083008 diff=0.083008 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0353 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0880 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=15577 i_micro=3893 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3100 mask_tok=93.00 y_norm=48.0880±0.0054 +2026-05-10 23:09:07,771 - INFO - Step 19480 | mode=video loss=0.084961 diff=0.084961 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0418 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0875 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=15585 i_micro=3895 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3433 mask_tok=103.00 y_norm=48.0875±0.0066 +2026-05-10 23:09:26,443 - INFO - Step 19490 | mode=video loss=0.063477 diff=0.063477 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0358 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0884 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=15593 i_micro=3897 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3233 mask_tok=97.00 y_norm=48.0884±0.0057 +2026-05-10 23:09:44,417 - INFO - Step 19500 | mode=video loss=0.095703 diff=0.095703 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0432 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0863 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=15601 i_micro=3899 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4600 mask_tok=138.00 y_norm=48.0863±0.0073 +2026-05-10 23:09:51,047 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 23:10:09,428 - INFO - Step 19510 | mode=video loss=0.120605 diff=0.120605 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0439 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0891 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=15609 i_micro=3901 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.2967 mask_tok=89.00 y_norm=48.0891±0.0052 +2026-05-10 23:10:28,063 - INFO - Step 19520 | mode=video loss=0.084961 diff=0.084961 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0432 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0892 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=15617 i_micro=3903 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3600 mask_tok=108.00 y_norm=48.0892±0.0062 +2026-05-10 23:10:46,439 - INFO - Step 19530 | mode=video loss=0.089844 diff=0.089844 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0389 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0857 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=15625 i_micro=3905 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3800 mask_tok=114.00 y_norm=48.0857±0.0057 +2026-05-10 23:11:04,940 - INFO - Step 19540 | mode=video loss=0.041992 diff=0.041992 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0608 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0875 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=15633 i_micro=3907 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4867 mask_tok=146.00 y_norm=48.0875±0.0057 +2026-05-10 23:11:22,905 - INFO - Step 19550 | mode=video loss=0.094727 diff=0.094727 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0397 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0865 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=15641 i_micro=3909 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4800 mask_tok=144.00 y_norm=48.0865±0.0062 +2026-05-10 23:11:41,325 - INFO - Step 19560 | mode=video loss=0.059326 diff=0.059326 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0577 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0849 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=15649 i_micro=3911 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3833 mask_tok=115.00 y_norm=48.0849±0.0060 +2026-05-10 23:11:59,981 - INFO - Step 19570 | mode=video loss=0.101562 diff=0.101562 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0612 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0864 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=15657 i_micro=3913 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4133 mask_tok=124.00 y_norm=48.0864±0.0060 +2026-05-10 23:12:18,391 - INFO - Step 19580 | mode=video loss=0.095215 diff=0.095215 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0838 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0887 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=15665 i_micro=3915 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3300 mask_tok=99.00 y_norm=48.0887±0.0051 +2026-05-10 23:12:36,348 - INFO - Step 19590 | mode=video loss=0.079590 diff=0.079590 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0891 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0860 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=15673 i_micro=3917 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4133 mask_tok=124.00 y_norm=48.0860±0.0056 +2026-05-10 23:12:56,548 - INFO - Step 19600 | mode=video loss=0.036377 diff=0.036377 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0349 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0856 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=15681 i_micro=3919 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4100 mask_tok=123.00 y_norm=48.0856±0.0056 cond_uncond_dloss=0.130859 cond_grad=0.000187 cond_pred_l2=53.861900 cond_pred_ratio=0.031095 +2026-05-10 23:12:56,585 - INFO - Step 19600 | probe_semantic mcp_offdiag(mean/min/max)=0.644987/0.531591/0.707356 smol_offdiag=0.660557 prompts=4 mcp_tok=17.50 smol_tok=17.50 +2026-05-10 23:13:03,680 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 23:13:22,122 - INFO - Step 19610 | mode=video loss=0.102539 diff=0.102539 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0474 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0891 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=15689 i_micro=3921 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3633 mask_tok=109.00 y_norm=48.0891±0.0063 +2026-05-10 23:13:40,856 - INFO - Step 19620 | mode=video loss=0.091797 diff=0.091797 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0535 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0872 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=15697 i_micro=3923 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4067 mask_tok=122.00 y_norm=48.0872±0.0063 +2026-05-10 23:13:58,825 - INFO - Step 19630 | mode=video loss=0.166016 diff=0.166016 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0652 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0844 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=15705 i_micro=3925 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4567 mask_tok=137.00 y_norm=48.0844±0.0063 +2026-05-10 23:14:17,272 - INFO - Step 19640 | mode=video loss=0.062500 diff=0.062500 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0531 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0860 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=15713 i_micro=3927 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4200 mask_tok=126.00 y_norm=48.0860±0.0058 +2026-05-10 23:14:35,972 - INFO - Step 19650 | mode=video loss=0.194336 diff=0.194336 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0831 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0881 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=15721 i_micro=3929 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3600 mask_tok=108.00 y_norm=48.0881±0.0056 +2026-05-10 23:14:54,341 - INFO - Step 19660 | mode=video loss=0.078125 diff=0.078125 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0657 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0845 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=15729 i_micro=3931 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4633 mask_tok=139.00 y_norm=48.0845±0.0064 +2026-05-10 23:15:12,980 - INFO - Step 19670 | mode=video loss=0.106445 diff=0.106445 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0411 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0872 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=15737 i_micro=3933 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3700 mask_tok=111.00 y_norm=48.0872±0.0060 +2026-05-10 23:15:30,941 - INFO - Step 19680 | mode=video loss=0.082031 diff=0.082031 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0520 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0883 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=15745 i_micro=3935 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3100 mask_tok=93.00 y_norm=48.0883±0.0053 +2026-05-10 23:15:49,324 - INFO - Step 19690 | mode=video loss=0.115234 diff=0.115234 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0443 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0871 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=15753 i_micro=3937 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3767 mask_tok=113.00 y_norm=48.0871±0.0056 +2026-05-10 23:16:07,942 - INFO - Step 19700 | mode=video loss=0.047852 diff=0.047852 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0282 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0845 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=15761 i_micro=3939 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4567 mask_tok=137.00 y_norm=48.0845±0.0062 +2026-05-10 23:16:14,773 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 23:16:33,180 - INFO - Step 19710 | mode=video loss=0.085449 diff=0.085449 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0522 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0884 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=15769 i_micro=3941 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3633 mask_tok=109.00 y_norm=48.0884±0.0055 +2026-05-10 23:16:51,935 - INFO - Step 19720 | mode=video loss=0.109375 diff=0.109375 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0635 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0872 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=15777 i_micro=3943 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3767 mask_tok=113.00 y_norm=48.0872±0.0064 +2026-05-10 23:17:09,901 - INFO - Step 19730 | mode=video loss=0.073242 diff=0.073242 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0643 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0879 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=15785 i_micro=3945 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3500 mask_tok=105.00 y_norm=48.0879±0.0055 +2026-05-10 23:17:28,323 - INFO - Step 19740 | mode=video loss=0.085938 diff=0.085938 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0621 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0891 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=15793 i_micro=3947 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.2967 mask_tok=89.00 y_norm=48.0891±0.0053 +2026-05-10 23:17:46,999 - INFO - Step 19750 | mode=video loss=0.101562 diff=0.101562 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0559 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0857 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=15801 i_micro=3949 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4433 mask_tok=133.00 y_norm=48.0857±0.0063 +2026-05-10 23:18:04,980 - INFO - Step 19760 | mode=video loss=0.118164 diff=0.118164 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0491 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0885 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=15809 i_micro=3951 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3333 mask_tok=100.00 y_norm=48.0885±0.0057 +2026-05-10 23:18:23,379 - INFO - Step 19770 | mode=video loss=0.098145 diff=0.098145 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0546 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0887 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=15817 i_micro=3953 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3467 mask_tok=104.00 y_norm=48.0887±0.0056 +2026-05-10 23:18:42,162 - INFO - Step 19780 | mode=video loss=0.060547 diff=0.060547 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0461 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0867 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=15825 i_micro=3955 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4500 mask_tok=135.00 y_norm=48.0867±0.0054 +2026-05-10 23:19:00,540 - INFO - Step 19790 | mode=video loss=0.094727 diff=0.094727 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0377 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0885 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=15833 i_micro=3957 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3500 mask_tok=105.00 y_norm=48.0885±0.0066 +2026-05-10 23:19:20,961 - INFO - Step 19800 | mode=video loss=0.122070 diff=0.122070 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0599 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0879 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=15841 i_micro=3959 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3633 mask_tok=109.00 y_norm=48.0879±0.0065 cond_uncond_dloss=0.500000 cond_grad=0.000451 cond_pred_l2=78.227913 cond_pred_ratio=0.050033 +2026-05-10 23:19:20,998 - INFO - Step 19800 | probe_semantic mcp_offdiag(mean/min/max)=0.644830/0.531362/0.707202 smol_offdiag=0.660557 prompts=4 mcp_tok=17.50 smol_tok=17.50 +2026-05-10 23:19:27,884 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 23:19:45,834 - INFO - Step 19810 | mode=video loss=0.058350 diff=0.058350 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0836 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0856 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=15849 i_micro=3961 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3533 mask_tok=106.00 y_norm=48.0856±0.0057 +2026-05-10 23:20:04,250 - INFO - Step 19820 | mode=video loss=0.132812 diff=0.132812 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0421 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0861 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=15857 i_micro=3963 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3300 mask_tok=99.00 y_norm=48.0861±0.0057 +2026-05-10 23:20:23,053 - INFO - Step 19830 | mode=video loss=0.061279 diff=0.061279 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0969 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0864 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=15865 i_micro=3965 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.5367 mask_tok=161.00 y_norm=48.0864±0.0063 +2026-05-10 23:20:41,436 - INFO - Step 19840 | mode=video loss=0.263672 diff=0.263672 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.3718 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0869 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=15873 i_micro=3967 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3500 mask_tok=105.00 y_norm=48.0869±0.0055 +2026-05-10 23:21:00,165 - INFO - Step 19850 | mode=video loss=0.091797 diff=0.091797 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0459 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0866 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=15881 i_micro=3969 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4200 mask_tok=126.00 y_norm=48.0866±0.0063 +2026-05-10 23:21:18,133 - INFO - Step 19860 | mode=video loss=0.057861 diff=0.057861 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0617 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0887 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=15889 i_micro=3971 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3433 mask_tok=103.00 y_norm=48.0887±0.0053 +2026-05-10 23:21:36,611 - INFO - Step 19870 | mode=video loss=0.094727 diff=0.094727 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0344 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0872 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=15897 i_micro=3973 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4033 mask_tok=121.00 y_norm=48.0872±0.0057 +2026-05-10 23:21:55,454 - INFO - Step 19880 | mode=video loss=0.204102 diff=0.204102 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.3049 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0855 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=15905 i_micro=3975 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4667 mask_tok=140.00 y_norm=48.0855±0.0058 +2026-05-10 23:22:13,429 - INFO - Step 19890 | mode=video loss=0.072266 diff=0.072266 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0649 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0854 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=15913 i_micro=3977 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3700 mask_tok=111.00 y_norm=48.0854±0.0086 +2026-05-10 23:22:31,825 - INFO - Step 19900 | mode=video loss=0.095215 diff=0.095215 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0798 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0858 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=15921 i_micro=3979 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3400 mask_tok=102.00 y_norm=48.0858±0.0062 +2026-05-10 23:22:38,709 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 23:22:56,707 - INFO - Step 19910 | mode=video loss=0.098145 diff=0.098145 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0351 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0851 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=15929 i_micro=3981 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3800 mask_tok=114.00 y_norm=48.0851±0.0061 +2026-05-10 23:23:15,101 - INFO - Step 19920 | mode=video loss=0.129883 diff=0.129883 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0453 emb_mean=0.0008 emb_std=1.0017 tok_norm=48.0829 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=15937 i_micro=3983 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4767 mask_tok=143.00 y_norm=48.0829±0.0070 +2026-05-10 23:23:33,927 - INFO - Step 19930 | mode=video loss=0.130859 diff=0.130859 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0571 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0877 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=15945 i_micro=3985 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3800 mask_tok=114.00 y_norm=48.0877±0.0064 +2026-05-10 23:23:51,923 - INFO - Step 19940 | mode=video loss=0.072266 diff=0.072266 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0493 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0869 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=15953 i_micro=3987 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3367 mask_tok=101.00 y_norm=48.0869±0.0059 +2026-05-10 23:24:10,332 - INFO - Step 19950 | mode=video loss=0.093750 diff=0.093750 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0714 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0869 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=15961 i_micro=3989 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4700 mask_tok=141.00 y_norm=48.0869±0.0060 +2026-05-10 23:24:29,068 - INFO - Step 19960 | mode=video loss=0.113770 diff=0.113770 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0379 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0885 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=15969 i_micro=3991 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3367 mask_tok=101.00 y_norm=48.0885±0.0054 +2026-05-10 23:24:47,485 - INFO - Step 19970 | mode=video loss=0.096191 diff=0.096191 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0404 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0870 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=15977 i_micro=3993 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4667 mask_tok=140.00 y_norm=48.0870±0.0063 +2026-05-10 23:25:06,279 - INFO - Step 19980 | mode=video loss=0.125000 diff=0.125000 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.4575 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0872 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=15985 i_micro=3995 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3567 mask_tok=107.00 y_norm=48.0872±0.0065 +2026-05-10 23:25:24,264 - INFO - Step 19990 | mode=video loss=0.105469 diff=0.105469 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0446 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0869 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=15993 i_micro=3997 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3467 mask_tok=104.00 y_norm=48.0869±0.0060 +2026-05-10 23:25:44,274 - INFO - Step 20000 | mode=video loss=0.088379 diff=0.088379 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0928 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0852 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=16001 i_micro=3999 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4300 mask_tok=129.00 y_norm=48.0852±0.0062 cond_uncond_dloss=0.531250 cond_grad=0.000289 cond_pred_l2=44.211773 cond_pred_ratio=0.028553 +2026-05-10 23:25:44,311 - INFO - Step 20000 | probe_semantic mcp_offdiag(mean/min/max)=0.644651/0.531231/0.707190 smol_offdiag=0.660557 prompts=4 mcp_tok=17.50 smol_tok=17.50 +2026-05-10 23:25:51,072 - INFO - Saved checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_step20000.pt +2026-05-10 23:25:55,917 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 23:26:14,751 - INFO - Step 20010 | mode=video loss=0.049072 diff=0.049072 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0737 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0862 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=16009 i_micro=4001 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4867 mask_tok=146.00 y_norm=48.0862±0.0061 +2026-05-10 23:26:32,714 - INFO - Step 20020 | mode=video loss=0.124023 diff=0.124023 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0451 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0862 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=16017 i_micro=4003 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3133 mask_tok=94.00 y_norm=48.0862±0.0057 +2026-05-10 23:26:51,907 - INFO - Step 20030 | mode=video loss=0.094238 diff=0.094238 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0544 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0853 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=16025 i_micro=4005 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4000 mask_tok=120.00 y_norm=48.0853±0.0059 +2026-05-10 23:27:09,868 - INFO - Step 20040 | mode=video loss=0.238281 diff=0.238281 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1053 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0859 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=16033 i_micro=4007 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3767 mask_tok=113.00 y_norm=48.0859±0.0068 +2026-05-10 23:27:28,259 - INFO - Step 20050 | mode=video loss=0.129883 diff=0.129883 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0655 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0880 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=16041 i_micro=4009 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3667 mask_tok=110.00 y_norm=48.0880±0.0056 +2026-05-10 23:27:46,739 - INFO - Step 20060 | mode=video loss=0.096680 diff=0.096680 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0508 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0875 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=16049 i_micro=4011 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3333 mask_tok=100.00 y_norm=48.0875±0.0052 +2026-05-10 23:28:04,709 - INFO - Step 20070 | mode=video loss=0.085449 diff=0.085449 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0549 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0870 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=16057 i_micro=4013 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4733 mask_tok=142.00 y_norm=48.0870±0.0069 +2026-05-10 23:28:23,083 - INFO - Step 20080 | mode=video loss=0.093750 diff=0.093750 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0504 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0879 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=16065 i_micro=4015 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3667 mask_tok=110.00 y_norm=48.0879±0.0063 +2026-05-10 23:28:41,672 - INFO - Step 20090 | mode=video loss=0.099121 diff=0.099121 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0533 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0864 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=16073 i_micro=4017 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3267 mask_tok=98.00 y_norm=48.0864±0.0060 +2026-05-10 23:29:00,043 - INFO - Step 20100 | mode=video loss=0.088867 diff=0.088867 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0422 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0885 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=16081 i_micro=4019 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3367 mask_tok=101.00 y_norm=48.0885±0.0058 +2026-05-10 23:29:06,816 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 23:29:25,273 - INFO - Step 20110 | mode=video loss=0.112305 diff=0.112305 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0564 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0867 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=16089 i_micro=4021 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.5100 mask_tok=153.00 y_norm=48.0867±0.0068 +2026-05-10 23:29:43,234 - INFO - Step 20120 | mode=video loss=0.117676 diff=0.117676 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0641 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0880 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=16097 i_micro=4023 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3600 mask_tok=108.00 y_norm=48.0880±0.0055 +2026-05-10 23:30:01,641 - INFO - Step 20130 | mode=video loss=0.076172 diff=0.076172 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0654 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0864 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=16105 i_micro=4025 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3600 mask_tok=108.00 y_norm=48.0864±0.0055 +2026-05-10 23:30:20,106 - INFO - Step 20140 | mode=video loss=0.111816 diff=0.111816 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0415 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0868 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=16113 i_micro=4027 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.2867 mask_tok=86.00 y_norm=48.0868±0.0054 +2026-05-10 23:30:38,068 - INFO - Step 20150 | mode=video loss=0.069824 diff=0.069824 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0318 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0881 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=16121 i_micro=4029 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3533 mask_tok=106.00 y_norm=48.0881±0.0054 +2026-05-10 23:30:56,923 - INFO - Step 20160 | mode=video loss=0.183594 diff=0.183594 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0404 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0849 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=16129 i_micro=4031 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3433 mask_tok=103.00 y_norm=48.0849±0.0055 +2026-05-10 23:31:14,880 - INFO - Step 20170 | mode=video loss=0.063477 diff=0.063477 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0414 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0875 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=16137 i_micro=4033 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3767 mask_tok=113.00 y_norm=48.0875±0.0061 +2026-05-10 23:31:33,249 - INFO - Step 20180 | mode=video loss=0.068848 diff=0.068848 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1477 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0859 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=16145 i_micro=4035 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4433 mask_tok=133.00 y_norm=48.0859±0.0069 +2026-05-10 23:31:51,706 - INFO - Step 20190 | mode=video loss=0.117188 diff=0.117188 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0487 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0884 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=16153 i_micro=4037 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3367 mask_tok=101.00 y_norm=48.0884±0.0058 +2026-05-10 23:32:11,283 - INFO - Step 20200 | mode=video loss=0.083008 diff=0.083008 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0315 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0865 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=16161 i_micro=4039 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4300 mask_tok=129.00 y_norm=48.0865±0.0064 cond_uncond_dloss=0.308594 cond_grad=0.000092 cond_pred_l2=37.272690 cond_pred_ratio=0.022393 +2026-05-10 23:32:11,320 - INFO - Step 20200 | probe_semantic mcp_offdiag(mean/min/max)=0.644629/0.531180/0.707060 smol_offdiag=0.660557 prompts=4 mcp_tok=17.50 smol_tok=17.50 +2026-05-10 23:32:18,192 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 23:32:37,154 - INFO - Step 20210 | mode=video loss=0.098145 diff=0.098145 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0363 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0873 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=16169 i_micro=4041 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3267 mask_tok=98.00 y_norm=48.0873±0.0058 +2026-05-10 23:32:55,115 - INFO - Step 20220 | mode=video loss=0.127930 diff=0.127930 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0779 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0876 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=16177 i_micro=4043 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3833 mask_tok=115.00 y_norm=48.0876±0.0056 +2026-05-10 23:33:13,501 - INFO - Step 20230 | mode=video loss=0.069824 diff=0.069824 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0879 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0863 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=16185 i_micro=4045 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3433 mask_tok=103.00 y_norm=48.0863±0.0050 +2026-05-10 23:33:32,064 - INFO - Step 20240 | mode=video loss=0.194336 diff=0.194336 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1462 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0888 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=16193 i_micro=4047 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3467 mask_tok=104.00 y_norm=48.0888±0.0056 +2026-05-10 23:33:50,028 - INFO - Step 20250 | mode=video loss=0.098145 diff=0.098145 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0567 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0888 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=16201 i_micro=4049 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4100 mask_tok=123.00 y_norm=48.0888±0.0054 +2026-05-10 23:34:08,498 - INFO - Step 20260 | mode=video loss=0.192383 diff=0.192383 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1264 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0868 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=16209 i_micro=4051 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3433 mask_tok=103.00 y_norm=48.0868±0.0053 +2026-05-10 23:34:26,999 - INFO - Step 20270 | mode=video loss=0.112793 diff=0.112793 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0568 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0846 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=16217 i_micro=4053 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3600 mask_tok=108.00 y_norm=48.0846±0.0073 +2026-05-10 23:34:44,959 - INFO - Step 20280 | mode=video loss=0.096680 diff=0.096680 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0616 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0888 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=16225 i_micro=4055 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3567 mask_tok=107.00 y_norm=48.0888±0.0055 +2026-05-10 23:35:04,051 - INFO - Step 20290 | mode=video loss=0.083984 diff=0.083984 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0544 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0872 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=16233 i_micro=4057 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3567 mask_tok=107.00 y_norm=48.0872±0.0061 +2026-05-10 23:35:22,002 - INFO - Step 20300 | mode=video loss=0.141602 diff=0.141602 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0849 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0871 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=16241 i_micro=4059 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3267 mask_tok=98.00 y_norm=48.0871±0.0054 +2026-05-10 23:35:29,012 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 23:35:47,402 - INFO - Step 20310 | mode=video loss=0.078125 diff=0.078125 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0473 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0873 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=16249 i_micro=4061 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3700 mask_tok=111.00 y_norm=48.0873±0.0057 +2026-05-10 23:36:06,088 - INFO - Step 20320 | mode=video loss=0.073242 diff=0.073242 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1014 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0878 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=16257 i_micro=4063 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3467 mask_tok=104.00 y_norm=48.0878±0.0052 +2026-05-10 23:36:24,060 - INFO - Step 20330 | mode=video loss=0.083496 diff=0.083496 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0400 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0883 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=16265 i_micro=4065 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.2533 mask_tok=76.00 y_norm=48.0883±0.0053 +2026-05-10 23:36:43,034 - INFO - Step 20340 | mode=video loss=0.126953 diff=0.126953 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0350 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0883 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=16273 i_micro=4067 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3533 mask_tok=106.00 y_norm=48.0883±0.0061 +2026-05-10 23:37:01,000 - INFO - Step 20350 | mode=video loss=0.074707 diff=0.074707 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0454 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0884 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=16281 i_micro=4069 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3133 mask_tok=94.00 y_norm=48.0884±0.0054 +2026-05-10 23:37:19,432 - INFO - Step 20360 | mode=video loss=0.134766 diff=0.134766 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1769 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0872 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=16289 i_micro=4071 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4267 mask_tok=128.00 y_norm=48.0872±0.0056 +2026-05-10 23:37:38,000 - INFO - Step 20370 | mode=video loss=0.192383 diff=0.192383 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1408 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0855 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=16297 i_micro=4073 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3533 mask_tok=106.00 y_norm=48.0855±0.0068 +2026-05-10 23:37:55,951 - INFO - Step 20380 | mode=video loss=0.143555 diff=0.143555 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1216 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0865 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=16305 i_micro=4075 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3367 mask_tok=101.00 y_norm=48.0865±0.0062 +2026-05-10 23:38:14,861 - INFO - Step 20390 | mode=video loss=0.068848 diff=0.068848 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0372 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0869 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=16313 i_micro=4077 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3433 mask_tok=103.00 y_norm=48.0869±0.0055 +2026-05-10 23:38:34,437 - INFO - Step 20400 | mode=video loss=0.123047 diff=0.123047 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0343 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0857 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=16321 i_micro=4079 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3833 mask_tok=115.00 y_norm=48.0857±0.0063 cond_uncond_dloss=0.320312 cond_grad=0.000130 cond_pred_l2=36.537308 cond_pred_ratio=0.022803 +2026-05-10 23:38:34,474 - INFO - Step 20400 | probe_semantic mcp_offdiag(mean/min/max)=0.644537/0.531086/0.707043 smol_offdiag=0.660557 prompts=4 mcp_tok=17.50 smol_tok=17.50 +2026-05-10 23:38:41,290 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 23:38:59,236 - INFO - Step 20410 | mode=video loss=0.077637 diff=0.077637 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1054 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0869 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=16329 i_micro=4081 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3467 mask_tok=104.00 y_norm=48.0869±0.0063 +2026-05-10 23:39:18,342 - INFO - Step 20420 | mode=video loss=0.181641 diff=0.181641 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1324 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0870 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=16337 i_micro=4083 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4833 mask_tok=145.00 y_norm=48.0870±0.0065 +2026-05-10 23:39:36,302 - INFO - Step 20430 | mode=video loss=0.091309 diff=0.091309 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0260 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0858 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=16345 i_micro=4085 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4767 mask_tok=143.00 y_norm=48.0858±0.0066 +2026-05-10 23:39:54,699 - INFO - Step 20440 | mode=video loss=0.051025 diff=0.051025 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0374 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0881 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=16353 i_micro=4087 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3567 mask_tok=107.00 y_norm=48.0881±0.0056 +2026-05-10 23:40:13,166 - INFO - Step 20450 | mode=video loss=0.182617 diff=0.182617 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0899 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0854 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=16361 i_micro=4089 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3567 mask_tok=107.00 y_norm=48.0854±0.0059 +2026-05-10 23:40:31,123 - INFO - Step 20460 | mode=video loss=0.088379 diff=0.088379 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0422 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0850 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=16369 i_micro=4091 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3967 mask_tok=119.00 y_norm=48.0850±0.0054 +2026-05-10 23:40:50,455 - INFO - Step 20470 | mode=video loss=0.096680 diff=0.096680 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0505 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0850 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=16377 i_micro=4093 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3667 mask_tok=110.00 y_norm=48.0850±0.0069 +2026-05-10 23:41:08,418 - INFO - Step 20480 | mode=video loss=0.080078 diff=0.080078 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0470 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0879 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=16385 i_micro=4095 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3867 mask_tok=116.00 y_norm=48.0879±0.0060 +2026-05-10 23:41:26,805 - INFO - Step 20490 | mode=video loss=0.062988 diff=0.062988 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0372 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0863 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=16393 i_micro=4097 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4367 mask_tok=131.00 y_norm=48.0863±0.0065 +2026-05-10 23:41:45,557 - INFO - Step 20500 | mode=video loss=0.115234 diff=0.115234 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0555 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0871 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=16401 i_micro=4099 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3367 mask_tok=101.00 y_norm=48.0871±0.0055 +2026-05-10 23:41:52,549 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 23:42:10,495 - INFO - Step 20510 | mode=video loss=0.155273 diff=0.155273 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1422 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0884 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=16409 i_micro=4101 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3300 mask_tok=99.00 y_norm=48.0884±0.0056 +2026-05-10 23:42:29,647 - INFO - Step 20520 | mode=video loss=0.072754 diff=0.072754 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1076 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0850 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=16417 i_micro=4103 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3700 mask_tok=111.00 y_norm=48.0850±0.0063 +2026-05-10 23:42:47,626 - INFO - Step 20530 | mode=video loss=0.084961 diff=0.084961 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1102 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0884 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=16425 i_micro=4105 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3433 mask_tok=103.00 y_norm=48.0884±0.0052 +2026-05-10 23:43:05,604 - INFO - Step 20540 | mode=video loss=0.050049 diff=0.050049 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1542 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0866 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=16433 i_micro=4107 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4300 mask_tok=129.00 y_norm=48.0866±0.0060 +2026-05-10 23:43:24,825 - INFO - Step 20550 | mode=video loss=0.224609 diff=0.224609 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0369 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0864 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=16441 i_micro=4109 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4600 mask_tok=138.00 y_norm=48.0864±0.0064 +2026-05-10 23:43:42,819 - INFO - Step 20560 | mode=video loss=0.087402 diff=0.087402 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0370 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0872 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=16449 i_micro=4111 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3667 mask_tok=110.00 y_norm=48.0872±0.0061 +2026-05-10 23:44:01,264 - INFO - Step 20570 | mode=video loss=0.092773 diff=0.092773 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0423 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0868 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=16457 i_micro=4113 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3333 mask_tok=100.00 y_norm=48.0868±0.0056 +2026-05-10 23:44:19,835 - INFO - Step 20580 | mode=video loss=0.058350 diff=0.058350 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0505 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0884 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=16465 i_micro=4115 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3600 mask_tok=108.00 y_norm=48.0884±0.0055 +2026-05-10 23:44:37,813 - INFO - Step 20590 | mode=video loss=0.070801 diff=0.070801 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0948 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0872 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=16473 i_micro=4117 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4233 mask_tok=127.00 y_norm=48.0872±0.0066 +2026-05-10 23:44:58,374 - INFO - Step 20600 | mode=video loss=0.129883 diff=0.129883 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0621 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0866 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=16481 i_micro=4119 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4133 mask_tok=124.00 y_norm=48.0866±0.0061 cond_uncond_dloss=0.441406 cond_grad=0.000288 cond_pred_l2=56.825550 cond_pred_ratio=0.036618 +2026-05-10 23:44:58,411 - INFO - Step 20600 | probe_semantic mcp_offdiag(mean/min/max)=0.644830/0.531372/0.707192 smol_offdiag=0.660557 prompts=4 mcp_tok=17.50 smol_tok=17.50 +2026-05-10 23:45:05,415 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 23:45:23,368 - INFO - Step 20610 | mode=video loss=0.094238 diff=0.094238 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0326 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0862 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=16489 i_micro=4121 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4067 mask_tok=122.00 y_norm=48.0862±0.0061 +2026-05-10 23:45:41,805 - INFO - Step 20620 | mode=video loss=0.160156 diff=0.160156 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0872 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0860 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=16497 i_micro=4123 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3500 mask_tok=105.00 y_norm=48.0860±0.0072 +2026-05-10 23:46:00,578 - INFO - Step 20630 | mode=video loss=0.089844 diff=0.089844 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0444 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0866 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=16505 i_micro=4125 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3633 mask_tok=109.00 y_norm=48.0866±0.0058 +2026-05-10 23:46:18,554 - INFO - Step 20640 | mode=video loss=0.098633 diff=0.098633 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0674 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0887 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=16513 i_micro=4127 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.2867 mask_tok=86.00 y_norm=48.0887±0.0050 +2026-05-10 23:46:37,635 - INFO - Step 20650 | mode=video loss=0.122559 diff=0.122559 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0494 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0879 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=16521 i_micro=4129 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.5567 mask_tok=167.00 y_norm=48.0879±0.0067 +2026-05-10 23:46:55,612 - INFO - Step 20660 | mode=video loss=0.146484 diff=0.146484 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0346 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0857 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=16529 i_micro=4131 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4500 mask_tok=135.00 y_norm=48.0857±0.0055 +2026-05-10 23:47:13,582 - INFO - Step 20670 | mode=video loss=0.071289 diff=0.071289 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0510 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0887 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=16537 i_micro=4133 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3300 mask_tok=99.00 y_norm=48.0887±0.0055 +2026-05-10 23:47:32,710 - INFO - Step 20680 | mode=video loss=0.088379 diff=0.088379 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0503 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0871 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=16545 i_micro=4135 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4767 mask_tok=143.00 y_norm=48.0871±0.0069 +2026-05-10 23:47:50,675 - INFO - Step 20690 | mode=video loss=0.108887 diff=0.108887 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0512 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0888 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=16553 i_micro=4137 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3333 mask_tok=100.00 y_norm=48.0888±0.0065 +2026-05-10 23:48:09,040 - INFO - Step 20700 | mode=video loss=0.090332 diff=0.090332 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0330 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0872 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=16561 i_micro=4139 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3767 mask_tok=113.00 y_norm=48.0872±0.0059 +2026-05-10 23:48:15,850 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 23:48:34,612 - INFO - Step 20710 | mode=video loss=0.081055 diff=0.081055 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0376 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0895 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=16569 i_micro=4141 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3567 mask_tok=107.00 y_norm=48.0895±0.0055 +2026-05-10 23:48:52,577 - INFO - Step 20720 | mode=video loss=0.109375 diff=0.109375 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0821 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0885 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=16577 i_micro=4143 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3367 mask_tok=101.00 y_norm=48.0885±0.0055 +2026-05-10 23:49:11,657 - INFO - Step 20730 | mode=video loss=0.076172 diff=0.076172 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0436 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0880 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=16585 i_micro=4145 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4767 mask_tok=143.00 y_norm=48.0880±0.0068 +2026-05-10 23:49:29,605 - INFO - Step 20740 | mode=video loss=0.075684 diff=0.075684 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0777 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0861 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=16593 i_micro=4147 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.5133 mask_tok=154.00 y_norm=48.0861±0.0063 +2026-05-10 23:49:47,995 - INFO - Step 20750 | mode=video loss=0.095215 diff=0.095215 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0410 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0879 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=16601 i_micro=4149 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3567 mask_tok=107.00 y_norm=48.0879±0.0054 +2026-05-10 23:50:06,657 - INFO - Step 20760 | mode=video loss=0.109375 diff=0.109375 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0354 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0870 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=16609 i_micro=4151 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3700 mask_tok=111.00 y_norm=48.0870±0.0053 +2026-05-10 23:50:24,628 - INFO - Step 20770 | mode=video loss=0.057861 diff=0.057861 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0476 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0857 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=16617 i_micro=4153 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4467 mask_tok=134.00 y_norm=48.0857±0.0063 +2026-05-10 23:50:43,482 - INFO - Step 20780 | mode=video loss=0.114746 diff=0.114746 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0765 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0860 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=16625 i_micro=4155 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4767 mask_tok=143.00 y_norm=48.0860±0.0062 +2026-05-10 23:51:01,438 - INFO - Step 20790 | mode=video loss=0.102539 diff=0.102539 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0586 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0887 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=16633 i_micro=4157 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3467 mask_tok=104.00 y_norm=48.0887±0.0061 +2026-05-10 23:51:21,030 - INFO - Step 20800 | mode=video loss=0.102539 diff=0.102539 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0304 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0893 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=16641 i_micro=4159 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3267 mask_tok=98.00 y_norm=48.0893±0.0058 cond_uncond_dloss=0.250000 cond_grad=0.000054 cond_pred_l2=33.578156 cond_pred_ratio=0.018413 +2026-05-10 23:51:21,067 - INFO - Step 20800 | probe_semantic mcp_offdiag(mean/min/max)=0.644751/0.531384/0.706903 smol_offdiag=0.660557 prompts=4 mcp_tok=17.50 smol_tok=17.50 +2026-05-10 23:51:27,835 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 23:51:47,004 - INFO - Step 20810 | mode=video loss=0.074707 diff=0.074707 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0313 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0864 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=16649 i_micro=4161 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3800 mask_tok=114.00 y_norm=48.0864±0.0059 +2026-05-10 23:52:04,963 - INFO - Step 20820 | mode=video loss=0.124023 diff=0.124023 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0277 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0868 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=16657 i_micro=4163 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3667 mask_tok=110.00 y_norm=48.0868±0.0055 +2026-05-10 23:52:24,172 - INFO - Step 20830 | mode=video loss=0.175781 diff=0.175781 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1058 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0885 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=16665 i_micro=4165 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3733 mask_tok=112.00 y_norm=48.0885±0.0069 +2026-05-10 23:52:42,121 - INFO - Step 20840 | mode=video loss=0.079590 diff=0.079590 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0758 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0890 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=16673 i_micro=4167 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3800 mask_tok=114.00 y_norm=48.0890±0.0055 +2026-05-10 23:53:00,085 - INFO - Step 20850 | mode=video loss=0.071777 diff=0.071777 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0785 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0884 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=16681 i_micro=4169 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3133 mask_tok=94.00 y_norm=48.0884±0.0059 +2026-05-10 23:53:19,001 - INFO - Step 20860 | mode=video loss=0.118652 diff=0.118652 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0559 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0878 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=16689 i_micro=4171 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.2833 mask_tok=85.00 y_norm=48.0878±0.0059 +2026-05-10 23:53:36,948 - INFO - Step 20870 | mode=video loss=0.079102 diff=0.079102 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1045 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0846 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=16697 i_micro=4173 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4467 mask_tok=134.00 y_norm=48.0846±0.0056 +2026-05-10 23:53:54,941 - INFO - Step 20880 | mode=video loss=0.068848 diff=0.068848 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0675 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0896 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=16705 i_micro=4175 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3300 mask_tok=99.00 y_norm=48.0896±0.0057 +2026-05-10 23:54:14,051 - INFO - Step 20890 | mode=video loss=0.125977 diff=0.125977 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0541 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0881 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=16713 i_micro=4177 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3900 mask_tok=117.00 y_norm=48.0881±0.0061 +2026-05-10 23:54:32,006 - INFO - Step 20900 | mode=video loss=0.100098 diff=0.100098 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0392 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0878 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=16721 i_micro=4179 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3400 mask_tok=102.00 y_norm=48.0878±0.0057 +2026-05-10 23:54:38,788 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 23:54:58,003 - INFO - Step 20910 | mode=video loss=0.040527 diff=0.040527 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0389 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0892 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=16729 i_micro=4181 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4167 mask_tok=125.00 y_norm=48.0892±0.0061 +2026-05-10 23:55:15,960 - INFO - Step 20920 | mode=video loss=0.122559 diff=0.122559 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0808 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0856 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=16737 i_micro=4183 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4500 mask_tok=135.00 y_norm=48.0856±0.0060 +2026-05-10 23:55:33,924 - INFO - Step 20930 | mode=video loss=0.103516 diff=0.103516 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0382 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0895 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=16745 i_micro=4185 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3433 mask_tok=103.00 y_norm=48.0895±0.0052 +2026-05-10 23:55:53,022 - INFO - Step 20940 | mode=video loss=0.107422 diff=0.107422 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1245 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0871 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=16753 i_micro=4187 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3467 mask_tok=104.00 y_norm=48.0871±0.0054 +2026-05-10 23:56:10,980 - INFO - Step 20950 | mode=video loss=0.083008 diff=0.083008 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0783 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0855 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=16761 i_micro=4189 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3767 mask_tok=113.00 y_norm=48.0855±0.0055 +2026-05-10 23:56:30,039 - INFO - Step 20960 | mode=video loss=0.134766 diff=0.134766 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0510 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0867 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=16769 i_micro=4191 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4533 mask_tok=136.00 y_norm=48.0867±0.0061 +2026-05-10 23:56:47,994 - INFO - Step 20970 | mode=video loss=0.077637 diff=0.077637 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0254 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0887 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=16777 i_micro=4193 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3267 mask_tok=98.00 y_norm=48.0887±0.0052 +2026-05-10 23:57:05,956 - INFO - Step 20980 | mode=video loss=0.083008 diff=0.083008 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0339 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0865 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=16785 i_micro=4195 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3933 mask_tok=118.00 y_norm=48.0865±0.0058 +2026-05-10 23:57:25,121 - INFO - Step 20990 | mode=video loss=0.085938 diff=0.085938 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0681 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0877 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=16793 i_micro=4197 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3633 mask_tok=109.00 y_norm=48.0877±0.0062 +2026-05-10 23:57:44,726 - INFO - Step 21000 | mode=video loss=0.088379 diff=0.088379 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0296 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0897 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=16801 i_micro=4199 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.2800 mask_tok=84.00 y_norm=48.0897±0.0055 cond_uncond_dloss=0.558594 cond_grad=0.000031 cond_pred_l2=45.985046 cond_pred_ratio=0.026973 +2026-05-10 23:57:44,763 - INFO - Step 21000 | probe_semantic mcp_offdiag(mean/min/max)=0.644710/0.531320/0.706907 smol_offdiag=0.660557 prompts=4 mcp_tok=17.50 smol_tok=17.50 +2026-05-10 23:57:52,486 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-10 23:58:10,937 - INFO - Step 21010 | mode=video loss=0.104980 diff=0.104980 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0312 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0852 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=16809 i_micro=4201 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3633 mask_tok=109.00 y_norm=48.0852±0.0056 +2026-05-10 23:58:29,617 - INFO - Step 21020 | mode=video loss=0.086426 diff=0.086426 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0671 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0887 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=16817 i_micro=4203 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3567 mask_tok=107.00 y_norm=48.0887±0.0060 +2026-05-10 23:58:47,616 - INFO - Step 21030 | mode=video loss=0.150391 diff=0.150391 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0587 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0883 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=16825 i_micro=4205 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3600 mask_tok=108.00 y_norm=48.0883±0.0062 +2026-05-10 23:59:06,701 - INFO - Step 21040 | mode=video loss=0.155273 diff=0.155273 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0580 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0880 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=16833 i_micro=4207 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3267 mask_tok=98.00 y_norm=48.0880±0.0057 +2026-05-10 23:59:24,672 - INFO - Step 21050 | mode=video loss=0.128906 diff=0.128906 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0631 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0878 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=16841 i_micro=4209 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3500 mask_tok=105.00 y_norm=48.0878±0.0053 +2026-05-10 23:59:42,635 - INFO - Step 21060 | mode=video loss=0.055420 diff=0.055420 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0667 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0867 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=16849 i_micro=4211 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4200 mask_tok=126.00 y_norm=48.0867±0.0056 +2026-05-11 00:00:01,733 - INFO - Step 21070 | mode=video loss=0.099609 diff=0.099609 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0903 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0895 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=16857 i_micro=4213 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3467 mask_tok=104.00 y_norm=48.0895±0.0053 +2026-05-11 00:00:19,698 - INFO - Step 21080 | mode=video loss=0.094727 diff=0.094727 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0627 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0893 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=16865 i_micro=4215 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.2933 mask_tok=88.00 y_norm=48.0893±0.0054 +2026-05-11 00:00:38,753 - INFO - Step 21090 | mode=video loss=0.078613 diff=0.078613 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0342 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0874 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=16873 i_micro=4217 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3833 mask_tok=115.00 y_norm=48.0874±0.0061 +2026-05-11 00:00:56,708 - INFO - Step 21100 | mode=video loss=0.064453 diff=0.064453 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0484 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0872 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=16881 i_micro=4219 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.5000 mask_tok=150.00 y_norm=48.0872±0.0073 +2026-05-11 00:01:03,651 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-11 00:01:21,601 - INFO - Step 21110 | mode=video loss=0.116211 diff=0.116211 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0738 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0870 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=16889 i_micro=4221 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.2433 mask_tok=73.00 y_norm=48.0870±0.0054 +2026-05-11 00:01:40,779 - INFO - Step 21120 | mode=video loss=0.097656 diff=0.097656 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0645 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0864 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=16897 i_micro=4223 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3767 mask_tok=113.00 y_norm=48.0864±0.0058 +2026-05-11 00:01:58,744 - INFO - Step 21130 | mode=video loss=0.159180 diff=0.159180 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1034 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0902 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=16905 i_micro=4225 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3367 mask_tok=101.00 y_norm=48.0902±0.0056 +2026-05-11 00:02:17,127 - INFO - Step 21140 | mode=video loss=0.104980 diff=0.104980 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0737 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0856 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=16913 i_micro=4227 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4233 mask_tok=127.00 y_norm=48.0856±0.0061 +2026-05-11 00:02:35,794 - INFO - Step 21150 | mode=video loss=0.082520 diff=0.082520 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0675 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0877 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=16921 i_micro=4229 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4100 mask_tok=123.00 y_norm=48.0877±0.0063 +2026-05-11 00:02:53,763 - INFO - Step 21160 | mode=video loss=0.164062 diff=0.164062 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0774 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0867 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=16929 i_micro=4231 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3400 mask_tok=102.00 y_norm=48.0867±0.0058 +2026-05-11 00:03:12,737 - INFO - Step 21170 | mode=video loss=0.101562 diff=0.101562 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0293 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0874 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=16937 i_micro=4233 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.2800 mask_tok=84.00 y_norm=48.0874±0.0052 +2026-05-11 00:03:30,704 - INFO - Step 21180 | mode=video loss=0.128906 diff=0.128906 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0804 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0868 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=16945 i_micro=4235 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3300 mask_tok=99.00 y_norm=48.0868±0.0056 +2026-05-11 00:03:48,672 - INFO - Step 21190 | mode=video loss=0.113770 diff=0.113770 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0446 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0846 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=16953 i_micro=4237 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3767 mask_tok=113.00 y_norm=48.0846±0.0061 +2026-05-11 00:04:09,193 - INFO - Step 21200 | mode=video loss=0.062500 diff=0.062500 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0411 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0871 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=16961 i_micro=4239 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3567 mask_tok=107.00 y_norm=48.0871±0.0055 cond_uncond_dloss=0.361328 cond_grad=0.000120 cond_pred_l2=52.556919 cond_pred_ratio=0.032446 +2026-05-11 00:04:09,230 - INFO - Step 21200 | probe_semantic mcp_offdiag(mean/min/max)=0.644810/0.531360/0.706907 smol_offdiag=0.660557 prompts=4 mcp_tok=17.50 smol_tok=17.50 +2026-05-11 00:04:16,303 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-11 00:04:34,254 - INFO - Step 21210 | mode=video loss=0.089355 diff=0.089355 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0533 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0879 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=16969 i_micro=4241 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.5133 mask_tok=154.00 y_norm=48.0879±0.0069 +2026-05-11 00:04:53,392 - INFO - Step 21220 | mode=video loss=0.069336 diff=0.069336 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0852 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0880 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=16977 i_micro=4243 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.2633 mask_tok=79.00 y_norm=48.0880±0.0046 +2026-05-11 00:05:11,356 - INFO - Step 21230 | mode=video loss=0.130859 diff=0.130859 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0665 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0874 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=16985 i_micro=4245 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3033 mask_tok=91.00 y_norm=48.0874±0.0054 +2026-05-11 00:05:29,330 - INFO - Step 21240 | mode=video loss=0.051514 diff=0.051514 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0746 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0854 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=16993 i_micro=4247 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.5300 mask_tok=159.00 y_norm=48.0854±0.0061 +2026-05-11 00:05:47,999 - INFO - Step 21250 | mode=video loss=0.099121 diff=0.099121 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0299 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0894 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=17001 i_micro=4249 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3600 mask_tok=108.00 y_norm=48.0894±0.0053 +2026-05-11 00:06:05,964 - INFO - Step 21260 | mode=video loss=0.101074 diff=0.101074 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0617 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0873 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=17009 i_micro=4251 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4900 mask_tok=147.00 y_norm=48.0873±0.0060 +2026-05-11 00:06:24,365 - INFO - Step 21270 | mode=video loss=0.079102 diff=0.079102 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0454 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0877 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=17017 i_micro=4253 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3500 mask_tok=105.00 y_norm=48.0877±0.0054 +2026-05-11 00:06:43,027 - INFO - Step 21280 | mode=video loss=0.066895 diff=0.066895 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0218 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0879 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=17025 i_micro=4255 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3400 mask_tok=102.00 y_norm=48.0879±0.0058 +2026-05-11 00:07:00,990 - INFO - Step 21290 | mode=video loss=0.061768 diff=0.061768 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0441 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0889 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=17033 i_micro=4257 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3700 mask_tok=111.00 y_norm=48.0889±0.0050 +2026-05-11 00:07:20,257 - INFO - Step 21300 | mode=video loss=0.068848 diff=0.068848 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0455 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0859 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=17041 i_micro=4259 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4700 mask_tok=141.00 y_norm=48.0859±0.0067 +2026-05-11 00:07:27,446 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-11 00:07:45,393 - INFO - Step 21310 | mode=video loss=0.131836 diff=0.131836 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0551 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0856 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=17049 i_micro=4261 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3467 mask_tok=104.00 y_norm=48.0856±0.0059 +2026-05-11 00:08:03,352 - INFO - Step 21320 | mode=video loss=0.125000 diff=0.125000 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0711 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0860 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=17057 i_micro=4263 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3800 mask_tok=114.00 y_norm=48.0860±0.0054 +2026-05-11 00:08:22,508 - INFO - Step 21330 | mode=video loss=0.123047 diff=0.123047 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0845 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0871 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=17065 i_micro=4265 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4767 mask_tok=143.00 y_norm=48.0871±0.0066 +2026-05-11 00:08:40,461 - INFO - Step 21340 | mode=video loss=0.062988 diff=0.062988 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0284 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0885 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=17073 i_micro=4267 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3200 mask_tok=96.00 y_norm=48.0885±0.0056 +2026-05-11 00:08:59,086 - INFO - Step 21350 | mode=video loss=0.104004 diff=0.104004 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0327 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0858 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=17081 i_micro=4269 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.5367 mask_tok=161.00 y_norm=48.0858±0.0067 +2026-05-11 00:09:17,033 - INFO - Step 21360 | mode=video loss=0.092285 diff=0.092285 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0755 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0887 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=17089 i_micro=4271 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.2567 mask_tok=77.00 y_norm=48.0887±0.0058 +2026-05-11 00:09:34,982 - INFO - Step 21370 | mode=video loss=0.085938 diff=0.085938 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0563 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0872 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=17097 i_micro=4273 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.2533 mask_tok=76.00 y_norm=48.0872±0.0049 +2026-05-11 00:09:53,953 - INFO - Step 21380 | mode=video loss=0.081543 diff=0.081543 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0567 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0860 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=17105 i_micro=4275 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.5167 mask_tok=155.00 y_norm=48.0860±0.0062 +2026-05-11 00:10:11,928 - INFO - Step 21390 | mode=video loss=0.085449 diff=0.085449 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1177 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0875 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=17113 i_micro=4277 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3533 mask_tok=106.00 y_norm=48.0875±0.0051 +2026-05-11 00:10:31,916 - INFO - Step 21400 | mode=video loss=0.089355 diff=0.089355 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0662 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0882 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=17121 i_micro=4279 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3433 mask_tok=103.00 y_norm=48.0882±0.0057 cond_uncond_dloss=0.230469 cond_grad=0.000073 cond_pred_l2=36.083527 cond_pred_ratio=0.020649 +2026-05-11 00:10:31,953 - INFO - Step 21400 | probe_semantic mcp_offdiag(mean/min/max)=0.645076/0.531577/0.707196 smol_offdiag=0.660557 prompts=4 mcp_tok=17.50 smol_tok=17.50 +2026-05-11 00:10:39,015 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-11 00:10:57,767 - INFO - Step 21410 | mode=video loss=0.122559 diff=0.122559 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0482 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0877 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=17129 i_micro=4281 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3333 mask_tok=100.00 y_norm=48.0877±0.0052 +2026-05-11 00:11:15,732 - INFO - Step 21420 | mode=video loss=0.081055 diff=0.081055 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0435 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0888 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=17137 i_micro=4283 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3433 mask_tok=103.00 y_norm=48.0888±0.0057 +2026-05-11 00:11:34,945 - INFO - Step 21430 | mode=video loss=0.091309 diff=0.091309 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0754 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0866 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=17145 i_micro=4285 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.2367 mask_tok=71.00 y_norm=48.0866±0.0052 +2026-05-11 00:11:52,899 - INFO - Step 21440 | mode=video loss=0.086914 diff=0.086914 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0380 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0848 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=17153 i_micro=4287 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4667 mask_tok=140.00 y_norm=48.0848±0.0071 +2026-05-11 00:12:10,863 - INFO - Step 21450 | mode=video loss=0.049072 diff=0.049072 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0618 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0885 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=17161 i_micro=4289 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3333 mask_tok=100.00 y_norm=48.0885±0.0054 +2026-05-11 00:12:29,939 - INFO - Step 21460 | mode=video loss=0.186523 diff=0.186523 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1819 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0874 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=17169 i_micro=4291 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3700 mask_tok=111.00 y_norm=48.0874±0.0053 +2026-05-11 00:12:47,907 - INFO - Step 21470 | mode=video loss=0.118164 diff=0.118164 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0577 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0878 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=17177 i_micro=4293 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3133 mask_tok=94.00 y_norm=48.0878±0.0055 +2026-05-11 00:13:06,665 - INFO - Step 21480 | mode=video loss=0.135742 diff=0.135742 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0786 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0873 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=17185 i_micro=4295 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3533 mask_tok=106.00 y_norm=48.0873±0.0070 +2026-05-11 00:13:24,617 - INFO - Step 21490 | mode=video loss=0.063965 diff=0.063965 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0541 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0852 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=17193 i_micro=4297 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4133 mask_tok=124.00 y_norm=48.0852±0.0055 +2026-05-11 00:13:42,604 - INFO - Step 21500 | mode=video loss=0.060059 diff=0.060059 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0630 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0864 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=17201 i_micro=4299 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4700 mask_tok=141.00 y_norm=48.0864±0.0064 +2026-05-11 00:13:49,403 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-11 00:14:08,481 - INFO - Step 21510 | mode=video loss=0.179688 diff=0.179688 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0937 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0869 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=17209 i_micro=4301 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3633 mask_tok=109.00 y_norm=48.0869±0.0054 +2026-05-11 00:14:26,437 - INFO - Step 21520 | mode=video loss=0.075195 diff=0.075195 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0765 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0887 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=17217 i_micro=4303 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3300 mask_tok=99.00 y_norm=48.0887±0.0049 +2026-05-11 00:14:44,813 - INFO - Step 21530 | mode=video loss=0.083984 diff=0.083984 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0211 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0874 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=17225 i_micro=4305 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3233 mask_tok=97.00 y_norm=48.0874±0.0056 +2026-05-11 00:15:03,383 - INFO - Step 21540 | mode=video loss=0.099609 diff=0.099609 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0645 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0886 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=17233 i_micro=4307 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3533 mask_tok=106.00 y_norm=48.0886±0.0059 +2026-05-11 00:15:21,340 - INFO - Step 21550 | mode=video loss=0.072754 diff=0.072754 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0520 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0881 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=17241 i_micro=4309 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3533 mask_tok=106.00 y_norm=48.0881±0.0055 +2026-05-11 00:15:40,481 - INFO - Step 21560 | mode=video loss=0.106934 diff=0.106934 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0276 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0858 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=17249 i_micro=4311 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4167 mask_tok=125.00 y_norm=48.0858±0.0056 +2026-05-11 00:15:58,452 - INFO - Step 21570 | mode=video loss=0.138672 diff=0.138672 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0318 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0855 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=17257 i_micro=4313 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.5333 mask_tok=160.00 y_norm=48.0855±0.0071 +2026-05-11 00:16:16,417 - INFO - Step 21580 | mode=video loss=0.085938 diff=0.085938 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1102 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0874 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=17265 i_micro=4315 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3633 mask_tok=109.00 y_norm=48.0874±0.0054 +2026-05-11 00:16:35,172 - INFO - Step 21590 | mode=video loss=0.094238 diff=0.094238 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0615 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0875 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=17273 i_micro=4317 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3633 mask_tok=109.00 y_norm=48.0875±0.0061 +2026-05-11 00:16:54,745 - INFO - Step 21600 | mode=video loss=0.107910 diff=0.107910 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0511 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0864 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=17281 i_micro=4319 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4367 mask_tok=131.00 y_norm=48.0864±0.0059 cond_uncond_dloss=0.228516 cond_grad=0.000092 cond_pred_l2=38.975430 cond_pred_ratio=0.023619 +2026-05-11 00:16:54,782 - INFO - Step 21600 | probe_semantic mcp_offdiag(mean/min/max)=0.645080/0.531596/0.707355 smol_offdiag=0.660557 prompts=4 mcp_tok=17.50 smol_tok=17.50 +2026-05-11 00:17:01,561 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-11 00:17:20,788 - INFO - Step 21610 | mode=video loss=0.151367 diff=0.151367 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0528 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0888 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=17289 i_micro=4321 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4467 mask_tok=134.00 y_norm=48.0888±0.0071 +2026-05-11 00:17:38,761 - INFO - Step 21620 | mode=video loss=0.100098 diff=0.100098 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0282 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0854 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=17297 i_micro=4323 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3267 mask_tok=98.00 y_norm=48.0854±0.0058 +2026-05-11 00:17:56,728 - INFO - Step 21630 | mode=video loss=0.066406 diff=0.066406 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0676 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0857 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=17305 i_micro=4325 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3167 mask_tok=95.00 y_norm=48.0857±0.0064 +2026-05-11 00:18:15,729 - INFO - Step 21640 | mode=video loss=0.091309 diff=0.091309 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0385 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0873 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=17313 i_micro=4327 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4700 mask_tok=141.00 y_norm=48.0873±0.0077 +2026-05-11 00:18:33,683 - INFO - Step 21650 | mode=video loss=0.089844 diff=0.089844 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0401 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0848 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=17321 i_micro=4329 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3467 mask_tok=104.00 y_norm=48.0848±0.0057 +2026-05-11 00:18:52,071 - INFO - Step 21660 | mode=video loss=0.139648 diff=0.139648 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0686 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0860 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=17329 i_micro=4331 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3533 mask_tok=106.00 y_norm=48.0860±0.0056 +2026-05-11 00:19:10,812 - INFO - Step 21670 | mode=video loss=0.104004 diff=0.104004 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0309 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0868 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=17337 i_micro=4333 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3067 mask_tok=92.00 y_norm=48.0868±0.0062 +2026-05-11 00:19:28,780 - INFO - Step 21680 | mode=video loss=0.099609 diff=0.099609 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0597 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0886 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=17345 i_micro=4335 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3800 mask_tok=114.00 y_norm=48.0886±0.0063 +2026-05-11 00:19:47,934 - INFO - Step 21690 | mode=video loss=0.085938 diff=0.085938 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0729 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0872 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=17353 i_micro=4337 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3733 mask_tok=112.00 y_norm=48.0872±0.0066 +2026-05-11 00:20:05,900 - INFO - Step 21700 | mode=video loss=0.117188 diff=0.117188 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0369 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0887 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=17361 i_micro=4339 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3300 mask_tok=99.00 y_norm=48.0887±0.0060 +2026-05-11 00:20:12,822 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-11 00:20:30,767 - INFO - Step 21710 | mode=video loss=0.054932 diff=0.054932 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0308 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0868 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=17369 i_micro=4341 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4933 mask_tok=148.00 y_norm=48.0868±0.0062 +2026-05-11 00:20:49,886 - INFO - Step 21720 | mode=video loss=0.075684 diff=0.075684 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0446 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0879 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=17377 i_micro=4343 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.2667 mask_tok=80.00 y_norm=48.0879±0.0054 +2026-05-11 00:21:07,871 - INFO - Step 21730 | mode=video loss=0.093262 diff=0.093262 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0370 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0887 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=17385 i_micro=4345 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3167 mask_tok=95.00 y_norm=48.0887±0.0055 +2026-05-11 00:21:26,808 - INFO - Step 21740 | mode=video loss=0.118652 diff=0.118652 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0628 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0875 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=17393 i_micro=4347 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3333 mask_tok=100.00 y_norm=48.0875±0.0057 +2026-05-11 00:21:44,789 - INFO - Step 21750 | mode=video loss=0.085938 diff=0.085938 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0431 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0869 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=17401 i_micro=4349 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3267 mask_tok=98.00 y_norm=48.0869±0.0063 +2026-05-11 00:22:02,774 - INFO - Step 21760 | mode=video loss=0.098633 diff=0.098633 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0779 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0894 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=17409 i_micro=4351 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3267 mask_tok=98.00 y_norm=48.0894±0.0052 +2026-05-11 00:22:21,928 - INFO - Step 21770 | mode=video loss=0.068848 diff=0.068848 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0386 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0883 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=17417 i_micro=4353 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3267 mask_tok=98.00 y_norm=48.0883±0.0053 +2026-05-11 00:22:39,906 - INFO - Step 21780 | mode=video loss=0.124023 diff=0.124023 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0480 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0866 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=17425 i_micro=4355 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3333 mask_tok=100.00 y_norm=48.0866±0.0066 +2026-05-11 00:22:58,329 - INFO - Step 21790 | mode=video loss=0.076172 diff=0.076172 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0400 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0870 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=17433 i_micro=4357 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3267 mask_tok=98.00 y_norm=48.0870±0.0059 +2026-05-11 00:23:18,706 - INFO - Step 21800 | mode=video loss=0.064941 diff=0.064941 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0506 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0896 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=17441 i_micro=4359 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3400 mask_tok=102.00 y_norm=48.0896±0.0059 cond_uncond_dloss=0.248047 cond_grad=0.000074 cond_pred_l2=37.735981 cond_pred_ratio=0.020840 +2026-05-11 00:23:18,743 - INFO - Step 21800 | probe_semantic mcp_offdiag(mean/min/max)=0.645381/0.532011/0.707556 smol_offdiag=0.660557 prompts=4 mcp_tok=17.50 smol_tok=17.50 +2026-05-11 00:23:26,387 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-11 00:23:44,366 - INFO - Step 21810 | mode=video loss=0.096680 diff=0.096680 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0482 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0874 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=17449 i_micro=4361 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4900 mask_tok=147.00 y_norm=48.0874±0.0071 +2026-05-11 00:24:03,692 - INFO - Step 21820 | mode=video loss=0.255859 diff=0.255859 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.4820 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0858 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=17457 i_micro=4363 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4400 mask_tok=132.00 y_norm=48.0858±0.0064 +2026-05-11 00:24:21,664 - INFO - Step 21830 | mode=video loss=0.093262 diff=0.093262 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0846 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0858 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=17465 i_micro=4365 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4567 mask_tok=137.00 y_norm=48.0858±0.0064 +2026-05-11 00:24:39,638 - INFO - Step 21840 | mode=video loss=0.088867 diff=0.088867 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0311 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0874 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=17473 i_micro=4367 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3533 mask_tok=106.00 y_norm=48.0874±0.0057 +2026-05-11 00:24:58,891 - INFO - Step 21850 | mode=video loss=0.139648 diff=0.139648 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0545 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0866 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=17481 i_micro=4369 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4467 mask_tok=134.00 y_norm=48.0866±0.0061 +2026-05-11 00:25:16,865 - INFO - Step 21860 | mode=video loss=0.071777 diff=0.071777 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0300 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0874 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=17489 i_micro=4371 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3500 mask_tok=105.00 y_norm=48.0874±0.0056 +2026-05-11 00:25:35,537 - INFO - Step 21870 | mode=video loss=0.037842 diff=0.037842 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0658 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0864 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=17497 i_micro=4373 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3567 mask_tok=107.00 y_norm=48.0864±0.0057 +2026-05-11 00:25:53,511 - INFO - Step 21880 | mode=video loss=0.071289 diff=0.071289 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0741 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0879 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=17505 i_micro=4375 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3567 mask_tok=107.00 y_norm=48.0879±0.0056 +2026-05-11 00:26:11,483 - INFO - Step 21890 | mode=video loss=0.118652 diff=0.118652 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0876 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0871 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=17513 i_micro=4377 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3400 mask_tok=102.00 y_norm=48.0871±0.0062 +2026-05-11 00:26:30,174 - INFO - Step 21900 | mode=video loss=0.051514 diff=0.051514 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0922 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0886 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=17521 i_micro=4379 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.2767 mask_tok=83.00 y_norm=48.0886±0.0049 +2026-05-11 00:26:37,172 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-11 00:26:55,125 - INFO - Step 21910 | mode=video loss=0.146484 diff=0.146484 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0788 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0883 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=17529 i_micro=4381 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3533 mask_tok=106.00 y_norm=48.0883±0.0053 +2026-05-11 00:27:13,750 - INFO - Step 21920 | mode=video loss=0.102051 diff=0.102051 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0381 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0890 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=17537 i_micro=4383 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.8467 mask_tok=254.00 y_norm=48.0890±0.0079 +2026-05-11 00:27:31,747 - INFO - Step 21930 | mode=video loss=0.075684 diff=0.075684 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0264 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0891 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=17545 i_micro=4385 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.2967 mask_tok=89.00 y_norm=48.0891±0.0051 +2026-05-11 00:27:49,724 - INFO - Step 21940 | mode=video loss=0.085449 diff=0.085449 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0469 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0875 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=17553 i_micro=4387 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3167 mask_tok=95.00 y_norm=48.0875±0.0057 +2026-05-11 00:28:08,434 - INFO - Step 21950 | mode=video loss=0.100586 diff=0.100586 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0775 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0863 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=17561 i_micro=4389 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4400 mask_tok=132.00 y_norm=48.0863±0.0063 +2026-05-11 00:28:26,410 - INFO - Step 21960 | mode=video loss=0.085938 diff=0.085938 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0280 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0864 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=17569 i_micro=4391 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4867 mask_tok=146.00 y_norm=48.0864±0.0062 +2026-05-11 00:28:44,384 - INFO - Step 21970 | mode=video loss=0.083008 diff=0.083008 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0288 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0897 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=17577 i_micro=4393 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3767 mask_tok=113.00 y_norm=48.0897±0.0061 +2026-05-11 00:29:03,090 - INFO - Step 21980 | mode=video loss=0.109863 diff=0.109863 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0724 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0869 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=17585 i_micro=4395 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3267 mask_tok=98.00 y_norm=48.0869±0.0058 +2026-05-11 00:29:21,070 - INFO - Step 21990 | mode=video loss=0.058105 diff=0.058105 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0686 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0859 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=17593 i_micro=4397 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3767 mask_tok=113.00 y_norm=48.0859±0.0058 +2026-05-11 00:29:41,264 - INFO - Step 22000 | mode=video loss=0.083008 diff=0.083008 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0245 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0880 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=17601 i_micro=4399 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.2967 mask_tok=89.00 y_norm=48.0880±0.0051 cond_uncond_dloss=0.273438 cond_grad=0.000061 cond_pred_l2=36.264870 cond_pred_ratio=0.020216 +2026-05-11 00:29:41,301 - INFO - Step 22000 | probe_semantic mcp_offdiag(mean/min/max)=0.645290/0.531858/0.707472 smol_offdiag=0.660557 prompts=4 mcp_tok=17.50 smol_tok=17.50 +2026-05-11 00:29:48,453 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-11 00:30:06,410 - INFO - Step 22010 | mode=video loss=0.108398 diff=0.108398 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0505 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0879 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=17609 i_micro=4401 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4100 mask_tok=123.00 y_norm=48.0879±0.0064 +2026-05-11 00:30:24,386 - INFO - Step 22020 | mode=video loss=0.109863 diff=0.109863 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0655 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0870 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=17617 i_micro=4403 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4833 mask_tok=145.00 y_norm=48.0870±0.0070 +2026-05-11 00:30:43,492 - INFO - Step 22030 | mode=video loss=0.184570 diff=0.184570 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0746 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0895 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=17625 i_micro=4405 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3300 mask_tok=99.00 y_norm=48.0895±0.0057 +2026-05-11 00:31:01,492 - INFO - Step 22040 | mode=video loss=0.083984 diff=0.083984 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0928 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0869 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=17633 i_micro=4407 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3633 mask_tok=109.00 y_norm=48.0869±0.0056 +2026-05-11 00:31:20,159 - INFO - Step 22050 | mode=video loss=0.103027 diff=0.103027 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0399 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0877 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=17641 i_micro=4409 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3467 mask_tok=104.00 y_norm=48.0877±0.0062 +2026-05-11 00:31:38,130 - INFO - Step 22060 | mode=video loss=0.092285 diff=0.092285 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1127 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0867 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=17649 i_micro=4411 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3667 mask_tok=110.00 y_norm=48.0867±0.0058 +2026-05-11 00:31:56,114 - INFO - Step 22070 | mode=video loss=0.083984 diff=0.083984 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0653 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0902 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=17657 i_micro=4413 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3400 mask_tok=102.00 y_norm=48.0902±0.0058 +2026-05-11 00:32:15,282 - INFO - Step 22080 | mode=video loss=0.133789 diff=0.133789 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1313 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0856 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=17665 i_micro=4415 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3900 mask_tok=117.00 y_norm=48.0856±0.0081 +2026-05-11 00:32:33,237 - INFO - Step 22090 | mode=video loss=0.099609 diff=0.099609 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0660 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0898 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=17673 i_micro=4417 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.8667 mask_tok=260.00 y_norm=48.0898±0.0075 +2026-05-11 00:32:51,225 - INFO - Step 22100 | mode=video loss=0.154297 diff=0.154297 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0458 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0857 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=17681 i_micro=4419 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3467 mask_tok=104.00 y_norm=48.0857±0.0061 +2026-05-11 00:32:59,033 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-11 00:33:17,407 - INFO - Step 22110 | mode=video loss=0.087402 diff=0.087402 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0380 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0892 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=17689 i_micro=4421 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3200 mask_tok=96.00 y_norm=48.0892±0.0047 +2026-05-11 00:33:35,375 - INFO - Step 22120 | mode=video loss=0.062256 diff=0.062256 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0515 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0860 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=17697 i_micro=4423 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4367 mask_tok=131.00 y_norm=48.0860±0.0060 +2026-05-11 00:33:54,502 - INFO - Step 22130 | mode=video loss=0.089355 diff=0.089355 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0284 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0868 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=17705 i_micro=4425 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.5333 mask_tok=160.00 y_norm=48.0868±0.0070 +2026-05-11 00:34:12,453 - INFO - Step 22140 | mode=video loss=0.097168 diff=0.097168 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0488 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0871 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=17713 i_micro=4427 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3600 mask_tok=108.00 y_norm=48.0871±0.0058 +2026-05-11 00:34:30,411 - INFO - Step 22150 | mode=video loss=0.119629 diff=0.119629 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0371 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0867 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=17721 i_micro=4429 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.2867 mask_tok=86.00 y_norm=48.0867±0.0057 +2026-05-11 00:34:49,515 - INFO - Step 22160 | mode=video loss=0.065918 diff=0.065918 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0459 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0882 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=17729 i_micro=4431 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3700 mask_tok=111.00 y_norm=48.0882±0.0055 +2026-05-11 00:35:07,467 - INFO - Step 22170 | mode=video loss=0.042480 diff=0.042480 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0371 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0885 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=17737 i_micro=4433 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3800 mask_tok=114.00 y_norm=48.0885±0.0061 +2026-05-11 00:35:26,103 - INFO - Step 22180 | mode=video loss=0.150391 diff=0.150391 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1253 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0864 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=17745 i_micro=4435 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4233 mask_tok=127.00 y_norm=48.0864±0.0062 +2026-05-11 00:35:44,493 - INFO - Step 22190 | mode=video loss=0.079102 diff=0.079102 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0360 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0865 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=17753 i_micro=4437 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3600 mask_tok=108.00 y_norm=48.0865±0.0059 +2026-05-11 00:36:04,084 - INFO - Step 22200 | mode=video loss=0.074219 diff=0.074219 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0465 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0876 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=17761 i_micro=4439 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3533 mask_tok=106.00 y_norm=48.0876±0.0057 cond_uncond_dloss=0.175781 cond_grad=0.000036 cond_pred_l2=35.977787 cond_pred_ratio=0.020664 +2026-05-11 00:36:04,121 - INFO - Step 22200 | probe_semantic mcp_offdiag(mean/min/max)=0.644834/0.531326/0.707020 smol_offdiag=0.660557 prompts=4 mcp_tok=17.50 smol_tok=17.50 +2026-05-11 00:36:11,160 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-11 00:36:30,318 - INFO - Step 22210 | mode=video loss=0.049805 diff=0.049805 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0258 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0879 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=17769 i_micro=4441 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3600 mask_tok=108.00 y_norm=48.0879±0.0051 +2026-05-11 00:36:48,303 - INFO - Step 22220 | mode=video loss=0.168945 diff=0.168945 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0765 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0876 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=17777 i_micro=4443 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3467 mask_tok=104.00 y_norm=48.0876±0.0059 +2026-05-11 00:37:07,044 - INFO - Step 22230 | mode=video loss=0.085938 diff=0.085938 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0281 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0876 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=17785 i_micro=4445 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.8567 mask_tok=257.00 y_norm=48.0876±0.0075 +2026-05-11 00:37:25,492 - INFO - Step 22240 | mode=video loss=0.096191 diff=0.096191 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0867 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0876 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=17793 i_micro=4447 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.5000 mask_tok=150.00 y_norm=48.0876±0.0078 +2026-05-11 00:37:43,453 - INFO - Step 22250 | mode=video loss=0.080078 diff=0.080078 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1657 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0876 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=17801 i_micro=4449 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3200 mask_tok=96.00 y_norm=48.0876±0.0057 +2026-05-11 00:38:02,579 - INFO - Step 22260 | mode=video loss=0.127930 diff=0.127930 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0544 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0881 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=17809 i_micro=4451 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3467 mask_tok=104.00 y_norm=48.0881±0.0063 +2026-05-11 00:38:20,538 - INFO - Step 22270 | mode=video loss=0.130859 diff=0.130859 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.2379 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0879 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=17817 i_micro=4453 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3733 mask_tok=112.00 y_norm=48.0879±0.0060 +2026-05-11 00:38:38,529 - INFO - Step 22280 | mode=video loss=0.063965 diff=0.063965 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0653 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0861 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=17825 i_micro=4455 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4100 mask_tok=123.00 y_norm=48.0861±0.0065 +2026-05-11 00:38:57,626 - INFO - Step 22290 | mode=video loss=0.051025 diff=0.051025 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0937 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0859 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=17833 i_micro=4457 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4733 mask_tok=142.00 y_norm=48.0859±0.0063 +2026-05-11 00:39:15,580 - INFO - Step 22300 | mode=video loss=0.123535 diff=0.123535 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0286 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0868 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=17841 i_micro=4459 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3600 mask_tok=108.00 y_norm=48.0868±0.0058 +2026-05-11 00:39:22,541 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-11 00:39:41,545 - INFO - Step 22310 | mode=video loss=0.162109 diff=0.162109 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0950 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0865 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=17849 i_micro=4461 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4033 mask_tok=121.00 y_norm=48.0865±0.0060 +2026-05-11 00:39:59,506 - INFO - Step 22320 | mode=video loss=0.083984 diff=0.083984 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0564 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0880 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=17857 i_micro=4463 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3600 mask_tok=108.00 y_norm=48.0880±0.0056 +2026-05-11 00:40:17,474 - INFO - Step 22330 | mode=video loss=0.097168 diff=0.097168 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0506 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0884 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=17865 i_micro=4465 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3767 mask_tok=113.00 y_norm=48.0884±0.0059 +2026-05-11 00:40:36,416 - INFO - Step 22340 | mode=video loss=0.081543 diff=0.081543 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1069 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0868 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=17873 i_micro=4467 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4233 mask_tok=127.00 y_norm=48.0868±0.0064 +2026-05-11 00:40:54,390 - INFO - Step 22350 | mode=video loss=0.073242 diff=0.073242 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0358 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0872 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=17881 i_micro=4469 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3333 mask_tok=100.00 y_norm=48.0872±0.0058 +2026-05-11 00:41:12,856 - INFO - Step 22360 | mode=video loss=0.095703 diff=0.095703 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0477 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0895 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=17889 i_micro=4471 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3567 mask_tok=107.00 y_norm=48.0895±0.0060 +2026-05-11 00:41:31,232 - INFO - Step 22370 | mode=video loss=0.114258 diff=0.114258 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0901 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0876 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=17897 i_micro=4473 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3400 mask_tok=102.00 y_norm=48.0876±0.0052 +2026-05-11 00:41:49,224 - INFO - Step 22380 | mode=video loss=0.091309 diff=0.091309 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0407 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0854 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=17905 i_micro=4475 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4433 mask_tok=133.00 y_norm=48.0854±0.0060 +2026-05-11 00:42:08,342 - INFO - Step 22390 | mode=video loss=0.198242 diff=0.198242 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0372 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0879 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=17913 i_micro=4477 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3733 mask_tok=112.00 y_norm=48.0879±0.0052 +2026-05-11 00:42:27,952 - INFO - Step 22400 | mode=video loss=0.156250 diff=0.156250 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1428 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0866 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=17921 i_micro=4479 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3500 mask_tok=105.00 y_norm=48.0866±0.0056 cond_uncond_dloss=0.820312 cond_grad=0.000920 cond_pred_l2=49.890854 cond_pred_ratio=0.032555 +2026-05-11 00:42:27,989 - INFO - Step 22400 | probe_semantic mcp_offdiag(mean/min/max)=0.644545/0.530976/0.706715 smol_offdiag=0.660557 prompts=4 mcp_tok=17.50 smol_tok=17.50 +2026-05-11 00:42:34,770 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-11 00:42:53,554 - INFO - Step 22410 | mode=video loss=0.170898 diff=0.170898 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1489 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0879 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=17929 i_micro=4481 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3467 mask_tok=104.00 y_norm=48.0879±0.0058 +2026-05-11 00:43:12,000 - INFO - Step 22420 | mode=video loss=0.060791 diff=0.060791 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0727 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0875 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=17937 i_micro=4483 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3500 mask_tok=105.00 y_norm=48.0875±0.0054 +2026-05-11 00:43:30,022 - INFO - Step 22430 | mode=video loss=0.145508 diff=0.145508 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0376 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0878 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=17945 i_micro=4485 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3500 mask_tok=105.00 y_norm=48.0878±0.0053 +2026-05-11 00:43:48,800 - INFO - Step 22440 | mode=video loss=0.072266 diff=0.072266 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0641 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0888 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=17953 i_micro=4487 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3367 mask_tok=101.00 y_norm=48.0888±0.0057 +2026-05-11 00:44:07,271 - INFO - Step 22450 | mode=video loss=0.070801 diff=0.070801 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0330 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0891 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=17961 i_micro=4489 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3300 mask_tok=99.00 y_norm=48.0891±0.0061 +2026-05-11 00:44:25,241 - INFO - Step 22460 | mode=video loss=0.075684 diff=0.075684 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0721 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0888 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=17969 i_micro=4491 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3567 mask_tok=107.00 y_norm=48.0888±0.0056 +2026-05-11 00:44:44,433 - INFO - Step 22470 | mode=video loss=0.055176 diff=0.055176 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0291 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0844 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=17977 i_micro=4493 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4033 mask_tok=121.00 y_norm=48.0844±0.0052 +2026-05-11 00:45:02,378 - INFO - Step 22480 | mode=video loss=0.170898 diff=0.170898 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0329 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0862 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=17985 i_micro=4495 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3367 mask_tok=101.00 y_norm=48.0862±0.0055 +2026-05-11 00:45:21,160 - INFO - Step 22490 | mode=video loss=0.087891 diff=0.087891 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1116 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0880 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=17993 i_micro=4497 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3467 mask_tok=104.00 y_norm=48.0880±0.0053 +2026-05-11 00:45:39,564 - INFO - Step 22500 | mode=video loss=0.097656 diff=0.097656 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0418 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0866 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=18001 i_micro=4499 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3800 mask_tok=114.00 y_norm=48.0866±0.0068 +2026-05-11 00:45:46,786 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-11 00:46:04,737 - INFO - Step 22510 | mode=video loss=0.083984 diff=0.083984 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0423 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0879 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=18009 i_micro=4501 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3167 mask_tok=95.00 y_norm=48.0879±0.0055 +2026-05-11 00:46:23,875 - INFO - Step 22520 | mode=video loss=0.137695 diff=0.137695 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0520 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0871 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=18017 i_micro=4503 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3700 mask_tok=111.00 y_norm=48.0871±0.0057 +2026-05-11 00:46:41,855 - INFO - Step 22530 | mode=video loss=0.130859 diff=0.130859 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0680 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0880 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=18025 i_micro=4505 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3633 mask_tok=109.00 y_norm=48.0880±0.0060 +2026-05-11 00:47:00,380 - INFO - Step 22540 | mode=video loss=0.131836 diff=0.131836 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0347 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0881 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=18033 i_micro=4507 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3433 mask_tok=103.00 y_norm=48.0881±0.0050 +2026-05-11 00:47:18,734 - INFO - Step 22550 | mode=video loss=0.072754 diff=0.072754 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0581 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0858 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=18041 i_micro=4509 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3267 mask_tok=98.00 y_norm=48.0858±0.0059 +2026-05-11 00:47:36,686 - INFO - Step 22560 | mode=video loss=0.106445 diff=0.106445 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0334 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0889 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=18049 i_micro=4511 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3700 mask_tok=111.00 y_norm=48.0889±0.0063 +2026-05-11 00:47:55,674 - INFO - Step 22570 | mode=video loss=0.077148 diff=0.077148 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0837 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0877 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=18057 i_micro=4513 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4967 mask_tok=149.00 y_norm=48.0877±0.0070 +2026-05-11 00:48:13,642 - INFO - Step 22580 | mode=video loss=0.131836 diff=0.131836 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0813 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0880 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=18065 i_micro=4515 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3400 mask_tok=102.00 y_norm=48.0880±0.0058 +2026-05-11 00:48:32,087 - INFO - Step 22590 | mode=video loss=0.110352 diff=0.110352 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0763 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0886 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=18073 i_micro=4517 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3167 mask_tok=95.00 y_norm=48.0886±0.0054 +2026-05-11 00:48:52,164 - INFO - Step 22600 | mode=video loss=0.123047 diff=0.123047 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0346 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0864 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=18081 i_micro=4519 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3700 mask_tok=111.00 y_norm=48.0864±0.0050 cond_uncond_dloss=0.384766 cond_grad=0.000184 cond_pred_l2=39.460564 cond_pred_ratio=0.023104 +2026-05-11 00:48:52,201 - INFO - Step 22600 | probe_semantic mcp_offdiag(mean/min/max)=0.644609/0.531106/0.706673 smol_offdiag=0.660557 prompts=4 mcp_tok=17.50 smol_tok=17.50 +2026-05-11 00:48:59,397 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-11 00:49:17,340 - INFO - Step 22610 | mode=video loss=0.109863 diff=0.109863 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.2669 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0882 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=18089 i_micro=4521 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3300 mask_tok=99.00 y_norm=48.0882±0.0052 +2026-05-11 00:49:36,114 - INFO - Step 22620 | mode=video loss=0.087891 diff=0.087891 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0606 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0871 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=18097 i_micro=4523 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3900 mask_tok=117.00 y_norm=48.0871±0.0056 +2026-05-11 00:49:54,567 - INFO - Step 22630 | mode=video loss=0.091797 diff=0.091797 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0336 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0867 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=18105 i_micro=4525 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4767 mask_tok=143.00 y_norm=48.0867±0.0065 +2026-05-11 00:50:12,537 - INFO - Step 22640 | mode=video loss=0.081055 diff=0.081055 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0614 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0868 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=18113 i_micro=4527 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3333 mask_tok=100.00 y_norm=48.0868±0.0051 +2026-05-11 00:50:31,676 - INFO - Step 22650 | mode=video loss=0.097656 diff=0.097656 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0819 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0888 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=18121 i_micro=4529 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3467 mask_tok=104.00 y_norm=48.0888±0.0050 +2026-05-11 00:50:49,670 - INFO - Step 22660 | mode=video loss=0.071289 diff=0.071289 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0259 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0861 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=18129 i_micro=4531 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4133 mask_tok=124.00 y_norm=48.0861±0.0050 +2026-05-11 00:51:08,342 - INFO - Step 22670 | mode=video loss=0.100098 diff=0.100098 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0359 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0882 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=18137 i_micro=4533 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3667 mask_tok=110.00 y_norm=48.0882±0.0060 +2026-05-11 00:51:26,751 - INFO - Step 22680 | mode=video loss=0.095215 diff=0.095215 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0391 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0876 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=18145 i_micro=4535 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3600 mask_tok=108.00 y_norm=48.0876±0.0056 +2026-05-11 00:51:44,703 - INFO - Step 22690 | mode=video loss=0.073242 diff=0.073242 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0540 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0867 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=18153 i_micro=4537 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3233 mask_tok=97.00 y_norm=48.0867±0.0054 +2026-05-11 00:52:03,634 - INFO - Step 22700 | mode=video loss=0.129883 diff=0.129883 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0385 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0866 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=18161 i_micro=4539 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3500 mask_tok=105.00 y_norm=48.0866±0.0057 +2026-05-11 00:52:10,301 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-11 00:52:28,245 - INFO - Step 22710 | mode=video loss=0.111816 diff=0.111816 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0508 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0874 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=18169 i_micro=4541 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3967 mask_tok=119.00 y_norm=48.0874±0.0067 +2026-05-11 00:52:47,046 - INFO - Step 22720 | mode=video loss=0.089844 diff=0.089844 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1756 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0865 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=18177 i_micro=4543 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4533 mask_tok=136.00 y_norm=48.0865±0.0060 +2026-05-11 00:53:05,440 - INFO - Step 22730 | mode=video loss=0.084961 diff=0.084961 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0884 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0858 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=18185 i_micro=4545 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3633 mask_tok=109.00 y_norm=48.0858±0.0069 +2026-05-11 00:53:23,410 - INFO - Step 22740 | mode=video loss=0.080566 diff=0.080566 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0410 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0884 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=18193 i_micro=4547 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3267 mask_tok=98.00 y_norm=48.0884±0.0058 +2026-05-11 00:53:41,969 - INFO - Step 22750 | mode=video loss=0.087891 diff=0.087891 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0579 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0882 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=18201 i_micro=4549 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3267 mask_tok=98.00 y_norm=48.0882±0.0051 +2026-05-11 00:54:00,404 - INFO - Step 22760 | mode=video loss=0.088379 diff=0.088379 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0704 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0857 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=18209 i_micro=4551 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4733 mask_tok=142.00 y_norm=48.0857±0.0069 +2026-05-11 00:54:18,356 - INFO - Step 22770 | mode=video loss=0.147461 diff=0.147461 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0828 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0866 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=18217 i_micro=4553 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.9033 mask_tok=271.00 y_norm=48.0866±0.0075 +2026-05-11 00:54:37,238 - INFO - Step 22780 | mode=video loss=0.101074 diff=0.101074 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0390 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0895 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=18225 i_micro=4555 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3567 mask_tok=107.00 y_norm=48.0895±0.0059 +2026-05-11 00:54:55,224 - INFO - Step 22790 | mode=video loss=0.131836 diff=0.131836 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0410 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0853 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=18233 i_micro=4557 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4200 mask_tok=126.00 y_norm=48.0853±0.0054 +2026-05-11 00:55:15,310 - INFO - Step 22800 | mode=video loss=0.093262 diff=0.093262 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0409 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0860 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=18241 i_micro=4559 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3433 mask_tok=103.00 y_norm=48.0860±0.0068 cond_uncond_dloss=0.267578 cond_grad=0.000103 cond_pred_l2=36.539364 cond_pred_ratio=0.021425 +2026-05-11 00:55:15,347 - INFO - Step 22800 | probe_semantic mcp_offdiag(mean/min/max)=0.644480/0.530899/0.706674 smol_offdiag=0.660557 prompts=4 mcp_tok=17.50 smol_tok=17.50 +2026-05-11 00:55:22,154 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-11 00:55:40,553 - INFO - Step 22810 | mode=video loss=0.155273 diff=0.155273 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1564 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0895 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=18249 i_micro=4561 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3433 mask_tok=103.00 y_norm=48.0895±0.0061 +2026-05-11 00:55:58,533 - INFO - Step 22820 | mode=video loss=0.078125 diff=0.078125 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0443 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0886 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=18257 i_micro=4563 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3500 mask_tok=105.00 y_norm=48.0886±0.0054 +2026-05-11 00:56:17,723 - INFO - Step 22830 | mode=video loss=0.071777 diff=0.071777 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0444 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0857 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=18265 i_micro=4565 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4500 mask_tok=135.00 y_norm=48.0857±0.0065 +2026-05-11 00:56:35,714 - INFO - Step 22840 | mode=video loss=0.073242 diff=0.073242 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0323 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0885 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=18273 i_micro=4567 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3433 mask_tok=103.00 y_norm=48.0885±0.0055 +2026-05-11 00:56:54,454 - INFO - Step 22850 | mode=video loss=0.068848 diff=0.068848 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1035 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0876 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=18281 i_micro=4569 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3300 mask_tok=99.00 y_norm=48.0876±0.0052 +2026-05-11 00:57:12,827 - INFO - Step 22860 | mode=video loss=0.094727 diff=0.094727 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0675 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0865 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=18289 i_micro=4571 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4367 mask_tok=131.00 y_norm=48.0865±0.0064 +2026-05-11 00:57:30,792 - INFO - Step 22870 | mode=video loss=0.104980 diff=0.104980 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0775 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0873 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=18297 i_micro=4573 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3300 mask_tok=99.00 y_norm=48.0873±0.0053 +2026-05-11 00:57:49,885 - INFO - Step 22880 | mode=video loss=0.092285 diff=0.092285 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0200 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0856 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=18305 i_micro=4575 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3500 mask_tok=105.00 y_norm=48.0856±0.0066 +2026-05-11 00:58:07,866 - INFO - Step 22890 | mode=video loss=0.104004 diff=0.104004 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0414 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0882 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=18313 i_micro=4577 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3367 mask_tok=101.00 y_norm=48.0882±0.0059 +2026-05-11 00:58:25,857 - INFO - Step 22900 | mode=video loss=0.103516 diff=0.103516 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.1236 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0879 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=18321 i_micro=4579 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3500 mask_tok=105.00 y_norm=48.0879±0.0060 +2026-05-11 00:58:32,844 - INFO - Saved latest checkpoint: output/stage1_joint_openvid_fullmobile_o_fulldit_diffonly_initlatest_bs64_v2_20260429_8gpu/20260510_124431/checkpoint_latest.pt +2026-05-11 00:58:52,014 - INFO - Step 22910 | mode=video loss=0.094238 diff=0.094238 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0342 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0879 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=18329 i_micro=4581 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3367 mask_tok=101.00 y_norm=48.0879±0.0055 +2026-05-11 00:59:09,977 - INFO - Step 22920 | mode=video loss=0.116699 diff=0.116699 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0366 emb_mean=0.0008 emb_std=1.0019 tok_norm=48.0894 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=18337 i_micro=4583 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.2800 mask_tok=84.00 y_norm=48.0894±0.0055 +2026-05-11 00:59:28,517 - INFO - Step 22930 | mode=video loss=0.083984 diff=0.083984 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0778 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0877 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=18345 i_micro=4585 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.2400 mask_tok=72.00 y_norm=48.0877±0.0055 +2026-05-11 00:59:46,873 - INFO - Step 22940 | mode=video loss=0.090820 diff=0.090820 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0353 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0848 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=18353 i_micro=4587 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4767 mask_tok=143.00 y_norm=48.0848±0.0061 +2026-05-11 01:00:04,821 - INFO - Step 22950 | mode=video loss=0.071777 diff=0.071777 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0296 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0858 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=18361 i_micro=4589 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4567 mask_tok=137.00 y_norm=48.0858±0.0064 +2026-05-11 01:00:23,840 - INFO - Step 22960 | mode=video loss=0.076660 diff=0.076660 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0326 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0857 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=18369 i_micro=4591 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4400 mask_tok=132.00 y_norm=48.0857±0.0065 +2026-05-11 01:00:41,791 - INFO - Step 22970 | mode=video loss=0.090820 diff=0.090820 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0665 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0872 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=18377 i_micro=4593 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.4333 mask_tok=130.00 y_norm=48.0872±0.0056 +2026-05-11 01:01:00,399 - INFO - Step 22980 | mode=video loss=0.081543 diff=0.081543 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0488 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0882 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=18385 i_micro=4595 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3533 mask_tok=106.00 y_norm=48.0882±0.0059 +2026-05-11 01:01:18,824 - INFO - Step 22990 | mode=video loss=0.143555 diff=0.143555 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0447 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0859 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=18393 i_micro=4597 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3500 mask_tok=105.00 y_norm=48.0859±0.0063 +2026-05-11 01:01:38,400 - INFO - Step 23000 | mode=video loss=0.079590 diff=0.079590 d_mse=0.000000 d_cos=0.000000 d_pool=0.000000 d_nce=0.000000 d_h0geom=0.000000 d_postnorm=0.000000 d_packnorm=0.000000 d_packpoolnorm=0.000000 d_kvv_mse=0.000000 d_kvv_cos=0.000000 d_kvv_norm=0.000000 norm=0.000000 offdiag_cos=nan grad=0.0695 emb_mean=0.0008 emb_std=1.0018 tok_norm=48.0870 gate=1.0000 cfg_drop=0.000 lex_gate=0.2006 v_micro=18401 i_micro=4599 y_shape=1x1x300x2304 mask_shape=1x1x1x300 mask_nonpad=0.3767 mask_tok=113.00 y_norm=48.0870±0.0059 cond_uncond_dloss=0.265625 cond_grad=0.000034 cond_pred_l2=38.396240 cond_pred_ratio=0.021171 +2026-05-11 01:01:38,438 - INFO - Step 23000 | probe_semantic mcp_offdiag(mean/min/max)=0.644422/0.530840/0.706525 smol_offdiag=0.660557 prompts=4 mcp_tok=17.50 smol_tok=17.50