framework_config.yaml

framework:
  name: QwenGR00T
  qwenvl:
    attn_implementation: flash_attention_2
    vl_hidden_dim: 2048
    base_vlm: ./playground/Pretrained_models/Qwen3-VL-4B-Instruct
  dino:
    dino_backbone: dinov2_vits14
  action_model:
    action_model_type: DiT-B
    action_hidden_dim: 2560
    hidden_size: 2560
    add_pos_embed: true
    max_seq_len: 1024
    action_dim: 12
    state_dim: 16
    future_action_window_size: 15
    action_horizon: 16
    past_action_window_size: 0
    repeated_diffusion_steps: 8
    noise_beta_alpha: 1.5
    noise_beta_beta: 1.0
    noise_s: 0.999
    num_timestep_buckets: 1000
    num_inference_timesteps: 4
    num_target_vision_tokens: 32
    diffusion_model_cfg:
      cross_attention_dim: 2560
      dropout: 0.2
      final_dropout: true
      interleave_self_attention: true
      norm_type: ada_norm
      num_layers: 16
      output_dim: 2560
      positional_embeddings: null
  reduce_in_full_precision: true
datasets:
  vlm_data:
    dataset_py: vlm_datasets
    dataformat: llava_json
    dataset_use: asv2_conversation_en,asv2_detailed_description_en,asv2_region_captioning_en,coco_internvl_longcap_en,coco_karpathy_train_567_en,coco_negative_gpt4o_en,coco_poetry_zh,coco_rem_en_zh,cocorem_exist_yorn_en,cocotextv2_en,cocotextv2_gpt4o_en,okvqa_en,refcoco_grounding_aug_en,refcoco_grounding_en,tallyqa_coco_en,toloka_grounding_aug_en,vqav2_en,vsr_en
    eval_dataset: aokvqa_cauldron_llava_format
    data_flatten: false
    base_interval: 2
    max_pixels: 50176
    min_pixels: 784
    model_max_length: 2048
    model_type: qwen3vl
    per_device_batch_size: 4
    dataloader_module: vlm_datasets
  vla_data:
    dataset_py: lerobot_datasets
    data_root_dir: /share/zhaoyiren/robocasa_dataset/v1.0
    data_mix: robocasa365_pretrain300
    action_mode: abs
    sequential_step_sampling: false
    CoT_prompt: Your task is {instruction}. To identify the key objects for your task.
      Locate their bounding boxes in [x1,y1,x2,y2] format.
    CoT_answer: bbox
    default_image_resolution:
    - 3
    - 224
    - 224
    per_device_batch_size: 16
    load_all_data_for_training: true
    obs:
    - robot0_agentview_left
    - robot0_agentview_right
    - robot0_eye_in_hand
    image_size:
    - 224
    - 224
    delete_pause_frame: false
    video_backend: torchvision_av
    dataset_mix: robocasa365_target_atomic
    dataloader_module: lerobot_datasets
trainer:
  enable_gradient_checkpointing: true
  enable_mixed_precision_training: true
  epochs: 100
  eval_interval: 1000000
  freeze_modules: ''
  gradient_accumulation_steps: 2
  gradient_clipping: 1.0
  is_resume: false
  learning_rate:
    action_model: 0.0001
    base: 3.0e-05
    qwen_vl_interface: 1.0e-05
  logging_frequency: 10
  loss_scale:
    vla: 1.0
    vlm: 0.1
  lr_scheduler_type: cosine_with_min_lr
  max_grad_norm: 1.0
  max_train_steps: 120000
  num_warmup_steps: 5000
  optimizer:
    betas:
    - 0.9
    - 0.95
    eps: 1.0e-08
    name: AdamW
    weight_decay: 1.0e-08
  resume_epoch: null
  resume_step: null
  save_interval: 10000
  scheduler_specific_kwargs:
    min_lr: 1.0e-06
  warmup_ratio: 0.1
  weight_decay: 0.0
environment:
  wandb_mode: disabled
  wandb_project: ''
  wandb_entity: ''
  num_gpus: 2
  main_process_port: 29503
  nccl:
    ib_hca: mlx5_2,mlx5_3
    blocking_wait: 1
    async_error_handling: 1
    timeout: 10000
    socket_timeout_ms: 360000
seed: 42
run_id: qwen_gr00t_robocasa365_target_atomic
output_root_dir: ./results/training
wandb_project: VLA
wandb_entity: trevorzhaozhihan-zyr
output_dir: ./results/training/qwen_gr00t_robocasa365_target_atomic