datasets:
  vla_data:
    CoT_prompt: Your task is {instruction}. To identify the key objects for your task.
      Locate their bounding boxes in [x1,y1,x2,y2] format.
    data_mix: fourier_gr1_unified_1000
    data_root_dir: /gpfs/wangzixuan/visual_prompting/starVLA_robocasa/playground/Datasets/nvidia/PhysicalAI-Robotics-GR00T-X-Embodiment-Sim
    dataset_py: lerobot_datasets
    delete_pause_frame: false
    episode_subset_fraction: 0.3333333333
    episode_subset_seed: 42
    obs_image_size:
    - 224
    - 224
    per_device_batch_size: 32
    sequential_step_sampling: false
framework:
  action_model:
    action_dim: 29
    action_hidden_dim: 2560
    action_horizon: 16
    action_model_type: DiT-B
  name: QwenOFT
  qwenvl:
    attn_implementation: flash_attention_2
    base_vlm: /gpfs/wangzixuan/shared_project/sq/5-1/starVLA/playground/Pretrained_models/Qwen3-VL-4B-Instruct-Action
output_dir: ./playground/Checkpoints/starvla_qwen3oft_fourier_gr1_unified_1000_franka_epfrac_1_3
run_id: starvla_qwen3oft_fourier_gr1_unified_1000_franka_epfrac_1_3
run_root_dir: ./playground/Checkpoints
seed: 42
trainer:
  eval_interval: 100
  freeze_modules: true
  gradient_clipping: 1.0
  learning_rate:
    action_model: 0.0001
    base: 3.0e-05
    qwen_vl_interface: 1.0e-05
  logging_frequency: 10
  lr_scheduler_type: cosine_with_min_lr
  max_train_steps: 100000
  num_warmup_steps: 5000
  optimizer:
    betas:
    - 0.9
    - 0.95
    eps: 1.0e-08
    weight_decay: 1.0e-08
  pretrained_checkpoint: /gpfs/wangzixuan/shared_project/sq/5-1/starVLA/playground/Pretrained_models/0426_droid_pretrain_qwen3GR00T_triple_oft_pi_50k_16BS/final_model/pytorch_model.pt
  reload_modules: qwen_vl_interface
  save_interval: 10000
  scheduler_specific_kwargs:
    min_lr: 5.0e-07
wandb_entity: zwanggk
wandb_project: sq_robocasa