{
  "file": "InternW0-RoboTwin-clean.pt",
  "source_checkpoint": "/data/shared/wamrsi/jisong/w0/release/robotwin/c2r/step_10545/step_010545.pt",
  "size_bytes": 14705453767,
  "sha256": "b718314f57c2febb159deb03d92cf9ebba830f7c75920c0b6e34c4903b91f5a2",
  "model_only": true,
  "optimizer_state_included": false,
  "step": 10545,
  "torch_dtype": "torch.bfloat16",
  "tensor_count": 1977,
  "parameter_elements": 7352528643,
  "tensor_bytes": 14705057286,
  "behavior_config": {
    "schema_version": 4,
    "num_soft_prompts": 8,
    "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"
      }
    ],
    "soft_prompt_domains": [
      "robotwin_wam/aloha"
    ],
    "soft_prompt_layout": "per_chunk_prefix",
    "soft_prompt_position_mode": "identity_rope",
    "soft_prompt_timestep_mode": "follow_chunk",
    "observation_encoder": "dino",
    "dino_encoder": {
      "model_name": "dinov3-vitb16-hf",
      "weights_sha256": "1bde6e2cbdb2973afa32f8b000da32c7686c5be6d8e82399ac581f9bb35a1e50",
      "input_resolution": [
        384,
        320
      ],
      "feature_dim": 768,
      "patch_size": 16,
      "use_cls_token": false,
      "normalize_features": false,
      "preprocessing": "uint8_or_minus_one_one_to_imagenet_float32_no_resize",
      "weights_hash_format": "hf_config_and_safetensors_v1"
    },
    "architecture": "ahawam_one_way_ovcr",
    "action_video_read_mode": "future",
    "video_cache_mode": "generated_future_last_denoise",
    "action_position_unit": "action_step",
    "video_rope_position_unit": "latent_frame",
    "action_horizon": 64,
    "action_chunk_size": 16,
    "num_history_frames": 0,
    "video_rope_frame_stride": null,
    "video_attention_mask_mode": "first_frame_causal",
    "video_action_conditioned": false,
    "autoregressive_teacher_forcing": true,
    "action_teacher_forcing_mask_mode": "stage1_chunkwise",
    "action_obs_downsample_factor": 1,
    "video_plan_horizon": 96,
    "video_action_stride": 8,
    "vae_temporal_downsample_factor": 4
  },
  "training_recipe": {
    "version": 2,
    "future_training_mode": "idm",
    "max_action_offset": 0,
    "data_offset_config": {
      "train": {
        "max_action_offset": 0,
        "action_horizon": 64,
        "action_chunk_size": 16,
        "action_video_freq_ratio": 8,
        "action_offset_enabled": false,
        "domains": {
          "version": 1,
          "source_fps": 50,
          "target_fps": 25.0,
          "stride": 2,
          "phases": [
            0,
            1
          ],
          "action_horizon": 64,
          "action_chunk_size": 16,
          "max_action_offset": 0,
          "video_plan_horizon": 96,
          "num_history_frames": 0,
          "short_episodes": "pad_mask",
          "slots": [
            0,
            1,
            2,
            3,
            4,
            5,
            16,
            17,
            18,
            19,
            20,
            21,
            22,
            33
          ],
          "domain": {
            "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"
          },
          "normalization": "bounds_q99_minmax_degenerate_clip5",
          "sigma": 1e-08,
          "camera_layout": "cover_center_crop_main_top_wrists_bottom_384x320",
          "text": "episode_first_instruction_default_prompt_128_wan22ti2v5b",
          "metadata_sha256": "b79ffeee79313dabad593d571cddb0bcf9acd4ee88da2030a39b487afb60e5d8",
          "split_seed": 42,
          "val_set_proportion": 0.01,
          "stats_sha256": "31362a24460803c5080ef925b807acbc7af3e08d847c9db222bc2a33c0d96653",
          "anchor_sampling": "all_valid_native_anchors_two_phases_v1",
          "samples_per_epoch": 539287,
          "color_jitter": {
            "enabled": true,
            "brightness": 0.2,
            "contrast": 0.2,
            "saturation": 0.2,
            "hue": 0.0,
            "version": 1,
            "implementation": "torchvision_float32_rgb_v1",
            "order": [
              "brightness",
              "contrast",
              "saturation",
              "hue"
            ],
            "sharing": "sample_all_frames_and_camera_slots",
            "rng": "numpy_seed_epoch_index_cjit_v1",
            "exclusion": "explicit_missing_camera_mask"
          }
        }
      },
      "val": {
        "max_action_offset": 0,
        "action_horizon": 64,
        "action_chunk_size": 16,
        "action_video_freq_ratio": 8,
        "action_offset_enabled": false,
        "domains": {
          "version": 1,
          "source_fps": 50,
          "target_fps": 25.0,
          "stride": 2,
          "phases": [
            0,
            1
          ],
          "action_horizon": 64,
          "action_chunk_size": 16,
          "max_action_offset": 0,
          "video_plan_horizon": 96,
          "num_history_frames": 0,
          "short_episodes": "pad_mask",
          "slots": [
            0,
            1,
            2,
            3,
            4,
            5,
            16,
            17,
            18,
            19,
            20,
            21,
            22,
            33
          ],
          "domain": {
            "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"
          },
          "normalization": "bounds_q99_minmax_degenerate_clip5",
          "sigma": 1e-08,
          "camera_layout": "cover_center_crop_main_top_wrists_bottom_384x320",
          "text": "episode_first_instruction_default_prompt_128_wan22ti2v5b",
          "metadata_sha256": "b79ffeee79313dabad593d571cddb0bcf9acd4ee88da2030a39b487afb60e5d8",
          "split_seed": 42,
          "val_set_proportion": 0.01,
          "stats_sha256": "31362a24460803c5080ef925b807acbc7af3e08d847c9db222bc2a33c0d96653",
          "anchor_sampling": "prefer_full_horizon_v1",
          "samples_per_epoch": 128
        }
      }
    },
    "idm_video_cond_noise_prob": 0.5,
    "idm_checkpoint_full_layer": true,
    "idm_checkpoint_layer_stride": 1,
    "video_scheduler": {
      "type": "ahawam.models.wan22.schedulers.scheduler_continuous.WanContinuousFlowMatchScheduler",
      "shift": 5.0,
      "num_train_timesteps": 1000
    },
    "action_scheduler": {
      "type": "ahawam.models.wan22.schedulers.scheduler_continuous.WanContinuousFlowMatchScheduler",
      "shift": 5.0,
      "num_train_timesteps": 1000
    },
    "loss_lambda_video": 1.0,
    "loss_lambda_action_prior": 1.0,
    "action_timestep_mode": "per_chunk",
    "soft_prompt_training": {
      "lr_scale": 1.0,
      "schedule": "per_group_relative_v1",
      "init_std": 0.02
    }
  },
  "posttrain_statistics": {
    "recipe": {
      "version": 1,
      "source_fps": 50,
      "target_fps": 12.5,
      "stride": 4,
      "phases": [
        0,
        1,
        2,
        3
      ],
      "action_horizon": 64,
      "action_chunk_size": 16,
      "max_action_offset": 15,
      "video_plan_horizon": 96,
      "num_history_frames": 0,
      "short_episodes": "pad_mask",
      "slots": [
        0,
        1,
        2,
        3,
        4,
        5,
        16,
        17,
        18,
        19,
        20,
        21,
        22,
        33
      ],
      "domain": {
        "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"
      },
      "normalization": "bounds_q99_minmax_degenerate_clip5",
      "sigma": 1e-08,
      "camera_layout": "cover_center_crop_main_top_wrists_bottom_384x320",
      "text": "episode_first_instruction_default_prompt_128_wan22ti2v5b"
    },
    "metadata_sha256": "b79ffeee79313dabad593d571cddb0bcf9acd4ee88da2030a39b487afb60e5d8",
    "split_seed": 42,
    "val_set_proportion": 0.01,
    "train_episodes": 2472,
    "validation_episodes": 28,
    "position_selection": "union_all_reachable_phase_tick_no_padding",
    "quantile_method": "exact_numpy_linear",
    "action": {
      "q01": [
        -0.9637558460235596,
        -5.257390398583084e-07,
        -2.296771708643064e-05,
        -1.5930664610862733,
        -0.44254571497440337,
        -2.2050026798248292,
        0.0,
        -0.1566295325756073,
        0.0,
        -2.81171942333458e-05,
        -1.7002750635147095,
        -1.0270036816596986,
        -1.6663912868499755,
        0.0
      ],
      "q99": [
        0.17338556051254272,
        2.582751655578613,
        2.4753708839416504,
        1.325472354888916,
        1.2205402231216373,
        1.4509482383728027,
        1.0,
        1.0792511796951278,
        2.602708811759947,
        2.4975505542754988,
        1.3120307922363281,
        1.0698820352554321,
        2.101373972892743,
        1.0
      ],
      "minmax_bounds": {},
      "valid_count": [
        534343,
        534343,
        534343,
        534343,
        534343,
        534343,
        534343,
        534343,
        534343,
        534343,
        534343,
        534343,
        534343,
        534343
      ]
    },
    "proprio": {
      "q01": [
        -0.9626896333694458,
        -5.257390398583084e-07,
        -2.296771708643064e-05,
        -1.587372796535492,
        -0.43772934675216674,
        -2.146702766418457,
        0.0,
        -0.1528666913509369,
        0.0,
        -2.81171942333458e-05,
        -1.6975934505462646,
        -1.0255959033966064,
        -1.6527643203735352,
        0.0
      ],
      "q99": [
        0.17338168621063232,
        2.581955127716058,
        2.469698667526245,
        1.3165214133262622,
        1.2082784199714538,
        1.4458667039871216,
        1.0,
        1.0600751638412476,
        2.594983887672424,
        2.489859433174133,
        1.3023490905761719,
        1.0461957573890615,
        2.079715509414667,
        1.0
      ],
      "minmax_bounds": {},
      "valid_count": [
        534343,
        534343,
        534343,
        534343,
        534343,
        534343,
        534343,
        534343,
        534343,
        534343,
        534343,
        534343,
        534343,
        534343
      ]
    }
  },
  "posttrain_provenance": {
    "source": "/data/shared/caijisong/ckpt/robotwin_posttrain/assets/rlh_step30000",
    "global_step": 30000,
    "domain_names": [
      "interndata/franka_panda",
      "interndata/franka_robotiq85",
      "interndata/genie1",
      "interndata/lift2",
      "interndata/split_aloha",
      "interndata/acone",
      "agibotworld/g1_gripper",
      "galaxea/r1lite",
      "ebench/lift2_long_horizon",
      "robodojo/dual_arx_x5",
      "robomind/agilex",
      "robomind/franka",
      "robomind/franka_fr3_dual",
      "robomind/ur",
      "robocoin/agibot_g1_profile5",
      "robocoin/agibot_g1_profile6",
      "robocoin/cobot_magic_profile7",
      "robocoin/galbot_g1",
      "robocoin/r1lite",
      "robocoin/rmc_aida_l",
      "robocoin/split_aloha_profile20",
      "robocoin/tianqin_a2",
      "agibotworld/g1_tactile",
      "ebench/lift2_simple_pnp",
      "ebench/lift2_teleop_tasks",
      "robocoin/cobot_magic_profile8",
      "robocoin/cobot_magic_profile9",
      "robocoin/cobot_magic_profile10",
      "robocoin/cobot_magic_profile11",
      "robocoin/split_aloha_profile21"
    ],
    "source_format": "rlh-fastwam-canonical-2",
    "source_checkpoint": "/nvme/datahub/tmp/caozhe/aha-wam-pretrain/w0-37d-sixsrc-b16-s43-384gpu-bf16-a40-20260917-r5-proprio-sharded/checkpoints/dcp/30000",
    "strategy": "new_domain",
    "seed": 42,
    "proprio_init": "new_domain",
    "proprio_source_domain": null
  }
}