| |
| |
| |
|
|
| output_dir: /data/shared/caijisong/ckpt/robotwin_posttrain/runs/${now:%Y-%m-%d}_${now:%H-%M-%S} |
| batch_size: 2 |
| num_workers: 4 |
| gc_collect_every: 1 |
| lr_scheduler_type: cosine |
| learning_rate: 1.0e-05 |
| soft_prompt_lr_scale: 1.0 |
| num_epochs: 10 |
| max_steps: null |
| log_every: 10 |
| save_every: 1000 |
| eval_every: 500 |
| eval_num_inference_steps: 10 |
| gradient_accumulation_steps: 1 |
| mixed_precision: bf16 |
| seed: 42 |
| max_grad_norm: 1.0 |
| weight_decay: 0.0 |
| freeze_video_dit: false |
| resume: null |
| init_checkpoint: null |
| init_soft_prompts: false |
| resume_global_batch_size: null |
| wandb: |
| enabled: false |
| workspace: null |
| project: ahawam |
| name: robotwin_posttrain_new_domain_12p5hz |
| group: null |
| mode: online |
| data: |
| train: |
| _target_: w0.datasets.robotwin_posttrain.RobotwinPosttrainDataset |
| dataset_root: ${oc.env:W0_ROBOTWIN_DATA,data/robotwin} |
| is_training_set: true |
| samples_per_epoch: null |
| seed: 42 |
| val_set_proportion: 0.01 |
| pretrained_norm_stats: ${oc.env:W0_ROBOTWIN_STATS,assets/robotwin/normalization.json} |
| max_action_offset: 0 |
| action_horizon: 64 |
| action_chunk_size: 16 |
| video_plan_horizon: 96 |
| action_video_freq_ratio: 8 |
| sampling_stride: 2 |
| color_jitter: |
| enabled: true |
| brightness: 0.2 |
| contrast: 0.2 |
| saturation: 0.2 |
| hue: 0.0 |
| processor: |
| _target_: w0.datasets.robotwin_posttrain.RobotwinPosttrainProcessor |
| stats_path: ${oc.env:W0_ROBOTWIN_STATS,assets/robotwin/normalization.json} |
| epoch_sampling: all_valid_windows |
| val: |
| _target_: w0.datasets.robotwin_posttrain.RobotwinPosttrainDataset |
| dataset_root: ${oc.env:W0_ROBOTWIN_DATA,data/robotwin} |
| is_training_set: false |
| samples_per_epoch: 128 |
| seed: 42 |
| val_set_proportion: 0.01 |
| pretrained_norm_stats: ${oc.env:W0_ROBOTWIN_STATS,assets/robotwin/normalization.json} |
| max_action_offset: 0 |
| action_horizon: 64 |
| action_chunk_size: 16 |
| video_plan_horizon: 96 |
| action_video_freq_ratio: 8 |
| sampling_stride: 2 |
| processor: |
| _target_: w0.datasets.robotwin_posttrain.RobotwinPosttrainProcessor |
| stats_path: ${oc.env:W0_ROBOTWIN_STATS,assets/robotwin/normalization.json} |
| color_jitter: |
| enabled: true |
| brightness: 0.2 |
| contrast: 0.2 |
| saturation: 0.2 |
| hue: 0.0 |
| model: |
| _target_: w0.runtime.create_w0 |
| model_id: Wan-AI/Wan2.2-TI2V-5B |
| tokenizer_model_id: Wan-AI/Wan2.1-T2V-1.3B |
| tokenizer_max_len: 128 |
| load_text_encoder: false |
| proprio_dim: 37 |
| redirect_common_files: true |
| mot_checkpoint_mixed_attn: true |
| action_dit_pretrained_path: null |
| skip_dit_load_from_pretrain: true |
| legacy_checkpoint_config: null |
| observation_encoder: dino |
| dino_config: |
| model_name: dinov3-vitb16-hf |
| weights_path: ${oc.env:W0_DINO_WEIGHTS,assets/dinov3-vitb16-pretrain-lvd1689m} |
| input_resolution: |
| - 384 |
| - 320 |
| feature_dim: 768 |
| patch_size: 16 |
| use_cls_token: false |
| normalize_features: false |
| encode_microbatch_size: 16 |
| checkpoint_observation_encoder_adapt: false |
| action_horizon: 64 |
| action_chunk_size: 16 |
| max_action_offset: 0 |
| action_train_timestep_mode: per_chunk |
| future_training_mode: idm |
| idm_video_cond_noise_prob: 0.5 |
| idm_checkpoint_full_layer: true |
| idm_checkpoint_layer_stride: 1 |
| action_obs_downsample_factor: 1 |
| chunk_kv_editor_num_queries: 32 |
| num_history_frames: 0 |
| action_video_read_mode: future |
| video_rope_frame_stride: 1 |
| video_dit_config: |
| compact_timestep_embedding: true |
| has_image_input: false |
| patch_size: |
| - 1 |
| - 2 |
| - 2 |
| in_dim: 48 |
| hidden_dim: 3072 |
| ffn_dim: 14336 |
| freq_dim: 256 |
| text_dim: 4096 |
| out_dim: 48 |
| num_heads: 24 |
| attn_head_dim: 128 |
| num_layers: 30 |
| eps: 1.0e-06 |
| seperated_timestep: true |
| require_clip_embedding: false |
| require_vae_embedding: false |
| fuse_vae_embedding_in_latents: true |
| use_gradient_checkpointing: true |
| video_attention_mask_mode: first_frame_causal |
| action_conditioned: false |
| action_dim: 37 |
| action_group_causal_mask_mode: group_diagonal |
| action_dit_config: |
| num_soft_prompts: 8 |
| soft_prompt_domains: |
| - robotwin_wam/aloha |
| soft_prompt_init_std: 0.02 |
| domain_configs: |
| - name: robotwin_wam/aloha |
| dataset: robotwin_wam |
| embodiment: aloha |
| action_dim: 37 |
| proprio_dim: 37 |
| action_valid_indices: |
| - 0 |
| - 1 |
| - 2 |
| - 3 |
| - 4 |
| - 5 |
| - 16 |
| - 17 |
| - 18 |
| - 19 |
| - 20 |
| - 21 |
| - 22 |
| - 33 |
| proprio_valid_indices: |
| - 0 |
| - 1 |
| - 2 |
| - 3 |
| - 4 |
| - 5 |
| - 16 |
| - 17 |
| - 18 |
| - 19 |
| - 20 |
| - 21 |
| - 22 |
| - 33 |
| raw_joint_names: |
| - left_waist |
| - left_shoulder |
| - left_elbow |
| - left_forearm_roll |
| - left_wrist_angle |
| - left_wrist_rotate |
| - left_gripper |
| - right_waist |
| - right_shoulder |
| - right_elbow |
| - right_forearm_roll |
| - right_wrist_angle |
| - right_wrist_rotate |
| - right_gripper |
| action_semantics: native_absolute_joint_command_interval_end |
| action_dim: 37 |
| hidden_dim: 1024 |
| ffn_dim: 4096 |
| num_heads: 24 |
| attn_head_dim: 128 |
| num_layers: 30 |
| text_dim: 4096 |
| freq_dim: 256 |
| eps: 1.0e-06 |
| autoregressive_teacher_forcing: true |
| action_chunk_size: 16 |
| action_teacher_forcing_mask_mode: stage1_chunkwise |
| use_gradient_checkpointing: true |
| video_scheduler: |
| train_shift: 5.0 |
| infer_shift: 5.0 |
| num_train_timesteps: 1000 |
| action_scheduler: |
| train_shift: 5.0 |
| infer_shift: 5.0 |
| num_train_timesteps: 1000 |
| loss: |
| lambda_action: 1.0 |
| lambda_action_prior: 1.0 |
| video_plan_horizon: 96 |
| video_action_stride: 8 |
| video_plan_steps: 10 |
| domain_registry_path: null |
| posttrain_required: true |
| posttrain_stats: ${oc.env:W0_ROBOTWIN_STATS,assets/robotwin/normalization.json} |
| ckpt: ${oc.env:W0_ROBOTWIN_CHECKPOINT,InternW0-RoboTwin-clean.pt} |
| gpu_id: 0 |
| EVALUATION: |
| robotwin_root: ${oc.env:ROBOTWIN_ROOT,external/RoboTwin} |
| policy_name: ahawam_policy |
| task_name: null |
| task_config: demo_randomized |
| instruction_type: unseen |
| eval_num_episodes: 15 |
| output_dir: ${oc.env:W0_OUTPUT_DIR,outputs}/robotwin/evaluation/source_step_10545 |
| action_horizon: 64 |
| chunks_per_video_prefill: 4 |
| num_inference_steps: 10 |
| sigma_shift: null |
| text_cfg_scale: 1.0 |
| negative_prompt: '' |
| rand_device: cpu |
| tiled: false |
| timing_enabled: false |
| training_config: training_config.yaml |
| control_mode: fixed_period |
| target_fps: 25.0 |
| step_limit_reference_fps: 12.5 |
| dataset_stats_path: ${oc.env:W0_ROBOTWIN_STATS,assets/robotwin/normalization.json} |
| prompt_template: 'A video recorded from a robot''s point of view executing the following |
| instruction: {task}' |
| dino_weights_path: ${oc.env:W0_DINO_WEIGHTS,assets/dinov3-vitb16-pretrain-lvd1689m} |
| model_id: Wan-AI/Wan2.2-TI2V-5B |
| tokenizer_model_id: Wan-AI/Wan2.1-T2V-1.3B |
| device: cuda |
| training_config_sha256: 6b8c8129cfea185d9b2938f629b4384829b27d22fb6ba7026c6e1b0fed36529d |
| sampling_stride: 2 |
| text_cache_dir: ${oc.env:W0_ROBOTWIN_DATA,data/robotwin}/text_embeds_cache |
| MULTIRUN: |
| enabled: false |
| num_gpus: 4 |
| max_tasks_per_gpu: 1 |
|
|
