TheBurgstall commited on
Commit
2d9cc91
·
1 Parent(s): b387190

Remove training config yaml

Browse files
Files changed (1) hide show
  1. ltx2_tsantsalize_h100.yaml +0 -111
ltx2_tsantsalize_h100.yaml DELETED
@@ -1,111 +0,0 @@
1
- # tsantsalize IC-LoRA (Phase 3). Same recipe as the other two. 98 pairs (49 img + 49 vid@97f).
2
- # Buckets reflect this dataset: images 768x1024/1024x768/1440x1920, videos 768x1024x97/1024x768x97.
3
- model:
4
- model_path: /workspace/models/ltx-2.3-22b-dev.safetensors
5
- text_encoder_path: /workspace/models/gemma-3-12b-it/
6
- training_mode: lora
7
- load_checkpoint: /workspace/outputs/ltx2_tsantsalize/checkpoints # fresh -> step 0
8
-
9
- lora:
10
- rank: 64
11
- alpha: 64
12
- dropout: 0.0
13
- target_modules:
14
- - attn1.to_q
15
- - attn1.to_k
16
- - attn1.to_v
17
- - attn1.to_out.0
18
- - attn2.to_q
19
- - attn2.to_k
20
- - attn2.to_v
21
- - attn2.to_out.0
22
- - ff.net.0.proj
23
- - ff.net.2
24
-
25
- training_strategy:
26
- name: video_to_video
27
- reference_latents_dir: reference_latents
28
- first_frame_conditioning_p: 0.0
29
-
30
- data:
31
- preprocessed_data_root: /workspace/dataset_tsantsalize/.precomputed
32
- num_dataloader_workers: 4
33
-
34
- optimization:
35
- batch_size: 1
36
- gradient_accumulation_steps: 4
37
- enable_gradient_checkpointing: true # REQUIRED for 22B even on 94GB
38
- max_grad_norm: 1.0
39
- optimizer_type: prodigy
40
- optimizer_params:
41
- decouple: true
42
- d_coef: 1.0
43
- safeguard_warmup: true
44
- use_bias_correction: true
45
- weight_decay: 0.01
46
- learning_rate: 1.0
47
- scheduler_type: constant
48
- scheduler_params: {}
49
- steps: 3000
50
-
51
- acceleration:
52
- mixed_precision_mode: bf16
53
- quantization: null
54
- load_text_encoder_in_8bit: false
55
-
56
- flow_matching:
57
- timestep_sampling_mode: shifted_logit_normal
58
- timestep_sampling_params: {}
59
-
60
- validation:
61
- interval: 1000
62
- skip_initial_validation: true
63
- inference_steps: 30
64
- guidance_scale: 4.0
65
- negative_prompt: "worst quality, blurry, distorted, low quality, artifacts, deformed face, asymmetric features, generic face"
66
- prompts:
67
- - tsantsalize
68
- - tsantsalize
69
- - tsantsalize
70
- reference_videos:
71
- - /workspace/dataset_tsantsalize/reference/pair_00005.png # natural input (768x1024)
72
- - /workspace/dataset_tsantsalize/reference/pair_00007.png # natural input (768x1024)
73
- - /workspace/dataset_tsantsalize/reference/pair_00008.png # natural input (768x1024)
74
- reference_downscale_factor: 1
75
- video_dims: # 81-frame video val (was F=1 image); 576x768 keeps 0.75 AR + safe VRAM
76
- - 576
77
- - 768
78
- - 81
79
- videos_per_prompt: 1
80
- frame_rate: 24.0
81
- include_reference_in_output: true
82
- generate_audio: false
83
- seed: 42
84
- stg_blocks:
85
- - 29
86
- stg_mode: stg_v
87
- stg_scale: 1.0
88
-
89
- checkpoints:
90
- interval: 500
91
- keep_last_n: -1
92
- keep_training_states_last_n: 1
93
- save_training_state: full
94
- no_resume: false
95
- precision: bfloat16
96
- filename_prefix: tsantsalize
97
-
98
- wandb:
99
- enabled: true
100
- entity: null
101
- project: ltx2-tsantsalize
102
- tags:
103
- - ltx2
104
- - ic-lora
105
- - tsantsalize
106
- - image+video
107
- - h100-fullres
108
- log_validation_videos: true
109
-
110
- output_dir: /workspace/outputs/ltx2_tsantsalize
111
- seed: 42