Upload folder using huggingface_hub

dreamzero_real_teleop_g1_full_finetune_absolute_Pick_bottle_and_turn_and_pour_into_cup/checkpoint-20000/config.json

@@ -0,0 +1,113 @@
+{
+  "action_dim": 36,
+  "action_head_cfg": {
+    "_convert_": "object",
+    "_target_": "groot.vla.model.dreamzero.action_head.wan_flow_matching_action_tf.WANPolicyHead",
+    "config": {
+      "_recursive_": false,
+      "_target_": "groot.vla.model.dreamzero.action_head.wan_flow_matching_action_tf.WANPolicyHeadConfig",
+      "action_dim": 36,
+      "action_horizon": 24,
+      "action_loss_embodiment_ids": [
+        26,
+        17,
+        32
+      ],
+      "add_pos_embed": true,
+      "backbone_embedding_dim": 0,
+      "backbone_features_projector_cfg": null,
+      "decouple_video_action_noise": false,
+      "diffusion_model_cfg": {
+        "_convert_": "object",
+        "_target_": "groot.vla.model.dreamzero.modules.wan_video_dit_action_casual_chunk.CausalWanModel",
+        "action_dim": 36,
+        "diffusion_model_pretrained_path": "/hfm/boqian/liboqian_data/checkpoints/Wan2.1-I2V-14B-480P",
+        "dim": 5120,
+        "eps": 1e-06,
+        "ffn_dim": 13824,
+        "frame_seqlen": 220,
+        "freq_dim": 256,
+        "in_dim": 36,
+        "max_chunk_size": 4,
+        "model_type": "i2v",
+        "num_action_per_block": 24,
+        "num_frame_per_block": 2,
+        "num_heads": 40,
+        "num_layers": 40,
+        "num_state_per_block": 1,
+        "out_dim": 16
+      },
+      "expand_batch": null,
+      "freeze_decode_layer": false,
+      "hidden_size": 64,
+      "image_encoder_cfg": {
+        "_convert_": "object",
+        "_target_": "groot.vla.model.dreamzero.modules.wan_video_image_encoder.WanImageEncoder",
+        "image_encoder_pretrained_path": "/hfm/boqian/liboqian_data/checkpoints/Wan2.1-I2V-14B-480P/models_clip_open-clip-xlm-roberta-large-vit-huge-14.pth"
+      },
+      "init_lora_weights": "kaiming",
+      "input_embedding_dim": 1536,
+      "load_pretrained_det_decode_layer_path": null,
+      "lora_alpha": 4,
+      "lora_rank": 4,
+      "lora_target_modules": "q,k,v,o,ffn.0,ffn.2",
+      "max_action_dim": 36,
+      "max_state_dim": 64,
+      "model_dtype": "float32",
+      "noise_beta_alpha": 1.5,
+      "noise_beta_beta": 1.0,
+      "noise_s": 0.999,
+      "num_frame_per_block": 2,
+      "num_frames": 33,
+      "num_inference_timesteps": 4,
+      "num_timestep_buckets": 1000,
+      "repa_coeff": 1.0,
+      "repa_layer": 8,
+      "skip_component_loading": true,
+      "text_encoder_cfg": {
+        "_convert_": "object",
+        "_target_": "groot.vla.model.dreamzero.modules.wan_video_text_encoder.WanTextEncoder",
+        "text_encoder_pretrained_path": "/hfm/boqian/liboqian_data/checkpoints/Wan2.1-I2V-14B-480P/models_t5_umt5-xxl-enc-bf16.pth"
+      },
+      "tile_size_height": 34,
+      "tile_size_width": 34,
+      "tile_stride_height": 18,
+      "tile_stride_width": 16,
+      "tiled": false,
+      "train_architecture": "full",
+      "tune_diffusion_model": true,
+      "tune_projector": true,
+      "use_gradient_checkpointing": true,
+      "use_vlln": true,
+      "vae_cfg": {
+        "_convert_": "object",
+        "_target_": "groot.vla.model.dreamzero.modules.wan_video_vae.WanVideoVAE",
+        "vae_pretrained_path": "/hfm/boqian/liboqian_data/checkpoints/Wan2.1-I2V-14B-480P/Wan2.1_VAE.pth"
+      },
+      "video_noise_beta_alpha": 3.0,
+      "video_noise_beta_beta": 1.0,
+      "vl_self_attention_cfg": {
+        "_target_": "groot.vla.model.n1_5.modules.cross_attention_dit.SelfAttentionTransformer",
+        "attention_head_dim": 64,
+        "dropout": 0.2,
+        "final_dropout": true,
+        "num_attention_heads": 24,
+        "num_layers": 4,
+        "positional_embeddings": null
+      }
+    }
+  },
+  "action_horizon": 24,
+  "architectures": [
+    "VLA"
+  ],
+  "backbone_cfg": {
+    "_target_": "groot.vla.model.dreamzero.backbone.identity.IdentityBackbone"
+  },
+  "hidden_size": 0,
+  "model_dtype": "float32",
+  "model_type": "vla",
+  "resume_path": "./checkpoints/dreamzero_real_teleop_g1_full_finetune_absolute",
+  "torch_dtype": "bfloat16",
+  "transformers_version": "4.51.3"
+}

dreamzero_real_teleop_g1_full_finetune_absolute_Pick_bottle_and_turn_and_pour_into_cup/checkpoint-20000/experiment_cfg/conf.yaml

@@ -0,0 +1,3306 @@
+model:
+  _target_: groot.vla.model.dreamzero.base_vla.VLA
+  _convert_: object
+  config:
+    _target_: groot.vla.model.dreamzero.base_vla.VLAConfig
+    _recursive_: false
+    model_dtype: float32
+    hidden_size: 0
+    action_horizon: 24
+    action_dim: 36
+    backbone_cfg:
+      _target_: groot.vla.model.dreamzero.backbone.identity.IdentityBackbone
+    action_head_cfg:
+      config:
+        backbone_features_projector_cfg: null
+        _target_: groot.vla.model.dreamzero.action_head.wan_flow_matching_action_tf.WANPolicyHeadConfig
+        _recursive_: false
+        tiled: false
+        tile_size_height: 34
+        tile_size_width: 34
+        tile_stride_height: 18
+        tile_stride_width: 16
+        lora_rank: 4
+        lora_alpha: 4
+        num_frames: 33
+        num_frame_per_block: 2
+        lora_target_modules: q,k,v,o,ffn.0,ffn.2
+        init_lora_weights: kaiming
+        train_architecture: full
+        use_gradient_checkpointing: true
+        add_pos_embed: true
+        model_dtype: float32
+        max_state_dim: 64
+        max_action_dim: 36
+        action_loss_embodiment_ids:
+        - 26
+        - 17
+        - 32
+        hidden_size: 64
+        input_embedding_dim: 1536
+        backbone_embedding_dim: 0
+        repa_layer: 8
+        repa_coeff: 1.0
+        load_pretrained_det_decode_layer_path: null
+        freeze_decode_layer: false
+        expand_batch: null
+        use_vlln: true
+        vl_self_attention_cfg:
+          _target_: groot.vla.model.n1_5.modules.cross_attention_dit.SelfAttentionTransformer
+          positional_embeddings: null
+          num_layers: 4
+          num_attention_heads: 24
+          attention_head_dim: 64
+          dropout: 0.2
+          final_dropout: true
+        diffusion_model_cfg:
+          _target_: groot.vla.model.dreamzero.modules.wan_video_dit_action_casual_chunk.CausalWanModel
+          _convert_: object
+          diffusion_model_pretrained_path: /hfm/boqian/liboqian_data/checkpoints/Wan2.1-I2V-14B-480P
+          model_type: i2v
+          frame_seqlen: 220
+          dim: 5120
+          in_dim: 36
+          ffn_dim: 13824
+          out_dim: 16
+          freq_dim: 256
+          eps: 1.0e-06
+          num_heads: 40
+          num_layers: 40
+          max_chunk_size: 4
+          num_frame_per_block: 2
+          num_action_per_block: 24
+          num_state_per_block: 1
+          action_dim: 36
+        text_encoder_cfg:
+          _target_: groot.vla.model.dreamzero.modules.wan_video_text_encoder.WanTextEncoder
+          _convert_: object
+          text_encoder_pretrained_path: /hfm/boqian/liboqian_data/checkpoints/Wan2.1-I2V-14B-480P/models_t5_umt5-xxl-enc-bf16.pth
+        image_encoder_cfg:
+          _target_: groot.vla.model.dreamzero.modules.wan_video_image_encoder.WanImageEncoder
+          _convert_: object
+          image_encoder_pretrained_path: /hfm/boqian/liboqian_data/checkpoints/Wan2.1-I2V-14B-480P/models_clip_open-clip-xlm-roberta-large-vit-huge-14.pth
+        vae_cfg:
+          _target_: groot.vla.model.dreamzero.modules.wan_video_vae.WanVideoVAE
+          _convert_: object
+          vae_pretrained_path: /hfm/boqian/liboqian_data/checkpoints/Wan2.1-I2V-14B-480P/Wan2.1_VAE.pth
+        action_dim: 36
+        action_horizon: 24
+        num_inference_timesteps: 4
+        noise_beta_alpha: 1.5
+        noise_beta_beta: 1.0
+        noise_s: 0.999
+        num_timestep_buckets: 1000
+        decouple_video_action_noise: false
+        video_noise_beta_alpha: 3.0
+        video_noise_beta_beta: 1.0
+        tune_projector: true
+        tune_diffusion_model: true
+        skip_component_loading: true
+      _target_: groot.vla.model.dreamzero.action_head.wan_flow_matching_action_tf.WANPolicyHead
+      _convert_: object
+train_dataset:
+  _target_: groot.vla.data.dataset.lerobot_sharded.ShardedLeRobotMixtureDataset.from_mixture_spec
+  _convert_: object
+  mixture_spec:
+  - dataset_path:
+      real_teleop_g1:
+      - /hfm/boqian/liboqian_data/data/real_data/gear/g1/Pick_bottle_and_turn_and_pour_into_cup_absolute
+    dataset_weight: 1.0
+    distribute_weights: true
+  dataset_class: groot.vla.data.dataset.lerobot_sharded.ShardedLeRobotSubLangSingleActionChunkDatasetDROID
+  all_modality_configs:
+    oxe_droid:
+      video:
+        _target_: groot.vla.data.dataset.ModalityConfig
+        delta_indices:
+        - 0
+        - 1
+        - 2
+        - 3
+        - 4
+        - 5
+        - 6
+        - 7
+        - 8
+        - 9
+        - 10
+        - 11
+        - 12
+        - 13
+        - 14
+        - 15
+        - 16
+        - 17
+        - 18
+        - 19
+        - 20
+        - 21
+        - 22
+        - 23
+        - 24
+        eval_delta_indices:
+        - 0
+        modality_keys:
+        - video.exterior_image_1_left
+        - video.exterior_image_2_left
+        - video.wrist_image_left
+      state:
+        _target_: groot.vla.data.dataset.ModalityConfig
+        delta_indices:
+        - 0
+        modality_keys:
+        - state.joint_position
+        - state.gripper_position
+      action:
+        _target_: groot.vla.data.dataset.ModalityConfig
+        delta_indices:
+        - 0
+        - 1
+        - 2
+        - 3
+        - 4
+        - 5
+        - 6
+        - 7
+        - 8
+        - 9
+        - 10
+        - 11
+        - 12
+        - 13
+        - 14
+        - 15
+        - 16
+        - 17
+        - 18
+        - 19
+        - 20
+        - 21
+        - 22
+        - 23
+        modality_keys:
+        - action.joint_position
+        - action.gripper_position
+      language:
+        _target_: groot.vla.data.dataset.ModalityConfig
+        delta_indices:
+        - 0
+        modality_keys:
+        - annotation.language.language_instruction
+        - annotation.language.language_instruction_2
+        - annotation.language.language_instruction_3
+      lapa_action:
+        _target_: groot.vla.data.dataset.ModalityConfig
+        delta_indices:
+        - 0
+        modality_keys:
+        - lapa_action
+    agibot:
+      video:
+        _target_: groot.vla.data.dataset.ModalityConfig
+        delta_indices:
+        - 0
+        - 1
+        - 2
+        - 3
+        - 4
+        - 5
+        - 6
+        - 7
+        - 8
+        - 9
+        - 10
+        - 11
+        - 12
+        - 13
+        - 14
+        - 15
+        - 16
+        - 17
+        - 18
+        - 19
+        - 20
+        - 21
+        - 22
+        - 23
+        - 24
+        eval_delta_indices:
+        - -3
+        - -2
+        - -1
+        - 0
+        modality_keys:
+        - video.top_head
+        - video.hand_left
+        - video.hand_right
+      state:
+        _target_: groot.vla.data.dataset.ModalityConfig
+        delta_indices:
+        - 0
+        modality_keys:
+        - state.left_arm_joint_position
+        - state.right_arm_joint_position
+        - state.left_effector_position
+        - state.right_effector_position
+        - state.head_position
+        - state.waist_position
+      action:
+        _target_: groot.vla.data.dataset.ModalityConfig
+        delta_indices:
+        - 0
+        - 1
+        - 2
+        - 3
+        - 4
+        - 5
+        - 6
+        - 7
+        - 8
+        - 9
+        - 10
+        - 11
+        - 12
+        - 13
+        - 14
+        - 15
+        - 16
+        - 17
+        - 18
+        - 19
+        - 20
+        - 21
+        - 22
+        - 23
+        modality_keys:
+        - action.left_arm_joint_position
+        - action.right_arm_joint_position
+        - action.left_effector_position
+        - action.right_effector_position
+        - action.head_position
+        - action.waist_position
+        - action.robot_velocity
+      language:
+        _target_: groot.vla.data.dataset.ModalityConfig
+        delta_indices:
+        - 0
+        modality_keys:
+        - annotation.language.action_text
+    yam:
+      video:
+        _target_: groot.vla.data.dataset.ModalityConfig
+        delta_indices:
+        - 0
+        - 1
+        - 2
+        - 3
+        - 4
+        - 5
+        - 6
+        - 7
+        - 8
+        - 9
+        - 10
+        - 11
+        - 12
+        - 13
+        - 14
+        - 15
+        - 16
+        - 17
+        - 18
+        - 19
+        - 20
+        - 21
+        - 22
+        - 23
+        - 24
+        eval_delta_indices:
+        - 0
+        modality_keys:
+        - video.top_camera-images-rgb
+        - video.left_camera-images-rgb
+        - video.right_camera-images-rgb
+      state:
+        _target_: groot.vla.data.dataset.ModalityConfig
+        delta_indices:
+        - 0
+        modality_keys:
+        - state.left_joint_pos
+        - state.left_gripper_pos
+        - state.right_joint_pos
+        - state.right_gripper_pos
+      action:
+        _target_: groot.vla.data.dataset.ModalityConfig
+        delta_indices:
+        - 0
+        - 1
+        - 2
+        - 3
+        - 4
+        - 5
+        - 6
+        - 7
+        - 8
+        - 9
+        - 10
+        - 11
+        - 12
+        - 13
+        - 14
+        - 15
+        - 16
+        - 17
+        - 18
+        - 19
+        - 20
+        - 21
+        - 22
+        - 23
+        modality_keys:
+        - action.left_joint_pos
+        - action.left_gripper_pos
+        - action.right_joint_pos
+        - action.right_gripper_pos
+      language:
+        _target_: groot.vla.data.dataset.ModalityConfig
+        delta_indices:
+        - 0
+        modality_keys:
+        - annotation.task
+    real_teleop_g1:
+      video:
+        _target_: groot.vla.data.dataset.ModalityConfig
+        delta_indices:
+        - 0
+        - 1
+        - 2
+        - 3
+        - 4
+        - 5
+        - 6
+        - 7
+        - 8
+        - 9
+        - 10
+        - 11
+        - 12
+        - 13
+        - 14
+        - 15
+        - 16
+        - 17
+        - 18
+        - 19
+        - 20
+        - 21
+        - 22
+        - 23
+        - 24
+        eval_delta_indices:
+        - 0
+        modality_keys:
+        - video.egocentric
+      state:
+        _target_: groot.vla.data.dataset.ModalityConfig
+        delta_indices:
+        - 0
+        modality_keys:
+        - state.left_hand
+        - state.right_hand
+        - state.left_arm
+        - state.right_arm
+        - state.rpy
+        - state.height
+      action:
+        _target_: groot.vla.data.dataset.ModalityConfig
+        delta_indices:
+        - 0
+        - 1
+        - 2
+        - 3
+        - 4
+        - 5
+        - 6
+        - 7
+        - 8
+        - 9
+        - 10
+        - 11
+        - 12
+        - 13
+        - 14
+        - 15
+        - 16
+        - 17
+        - 18
+        - 19
+        - 20
+        - 21
+        - 22
+        - 23
+        modality_keys:
+        - action.left_hand
+        - action.right_hand
+        - action.left_arm
+        - action.right_arm
+        - action.rpy
+        - action.height
+        - action.torso_vx
+        - action.torso_vy
+        - action.torso_vyaw
+        - action.torso_dyaw
+      language:
+        _target_: groot.vla.data.dataset.ModalityConfig
+        delta_indices:
+        - 0
+        modality_keys:
+        - annotation.language.language_instruction
+  all_transforms:
+    oxe_droid:
+      _target_: groot.vla.data.transform.ComposedModalityTransform
+      transforms:
+      - _target_: groot.vla.data.transform.VideoToTensor
+        apply_to:
+        - video.exterior_image_1_left
+        - video.exterior_image_2_left
+        - video.wrist_image_left
+      - _target_: groot.vla.data.transform.VideoCrop
+        apply_to:
+        - video.exterior_image_1_left
+        - video.exterior_image_2_left
+        - video.wrist_image_left
+        scale: 0.95
+        mode: random
+      - _target_: groot.vla.data.transform.VideoResize
+        apply_to:
+        - video.exterior_image_1_left
+        - video.exterior_image_2_left
+        - video.wrist_image_left
+        height: 176
+        width: 320
+        interpolation: linear
+      - _target_: groot.vla.data.transform.VideoColorJitter
+        apply_to:
+        - video.exterior_image_1_left
+        - video.exterior_image_2_left
+        - video.wrist_image_left
+        brightness: 0.3
+        contrast: 0.4
+        saturation: 0.5
+        hue: 0.08
+      - _target_: groot.vla.data.transform.VideoToNumpy
+        apply_to:
+        - video.exterior_image_1_left
+        - video.exterior_image_2_left
+        - video.wrist_image_left
+      - _target_: groot.vla.data.transform.StateActionToTensor
+        apply_to:
+        - state.joint_position
+        - state.gripper_position
+      - _target_: groot.vla.data.transform.StateActionTransform
+        apply_to:
+        - state.joint_position
+        - state.gripper_position
+        normalization_modes:
+          state.joint_position: q99
+          state.gripper_position: q99
+      - _target_: groot.vla.data.transform.StateActionToTensor
+        apply_to:
+        - action.joint_position
+        - action.gripper_position
+      - _target_: groot.vla.data.transform.StateActionTransform
+        apply_to:
+        - action.joint_position
+        - action.gripper_position
+        normalization_modes:
+          action.joint_position: q99
+          action.gripper_position: q99
+      - _target_: groot.vla.data.transform.ConcatTransform
+        video_concat_order:
+        - video.exterior_image_1_left
+        - video.exterior_image_2_left
+        - video.wrist_image_left
+        state_concat_order:
+        - state.joint_position
+        - state.gripper_position
+        action_concat_order:
+        - action.joint_position
+        - action.gripper_position
+      - _target_: groot.vla.model.dreamzero.transform.dreamzero_cotrain.DreamTransform
+        default_instruction: Perform the default behavior.
+        language_dropout_prob: 0.0
+        always_use_default_instruction: false
+        max_state_dim: 64
+        max_action_dim: 36
+        max_length: 512
+        state_horizon: 1
+        action_horizon: 24
+        embodiment_tag_mapping:
+          real_gr1_arms_only: 0
+          real_gr1_arms_only_annotated: 1
+          real_gr1_arms_waist: 2
+          real_gr1_arms_waist_annotated: 3
+          dexmg_gr1_arms_only_inspire: 4
+          dexmg_gr1_arms_only_fourier: 5
+          dexmg_gr1_arms_waist_fourier: 6
+          robocasa_single_arm: 7
+          onex_eve_gripper: 8
+          robocasa_gr1_arms_only_inspire_hands: 9
+          robocasa_gr1_arms_only_fourier_hands: 10
+          robocasa_gr1_fixed_lower_body_inspire_hands: 11
+          robocasa_gr1_fixed_lower_body_fourier_hands: 12
+          robocasa_panda_omron: 13
+          robocasa_bimanual_panda_parallel_gripper: 15
+          robocasa_bimanual_panda_inspire_hand: 16
+          oxe_droid: 17
+          oxe_fractal: 18
+          oxe_language_table: 19
+          oxe_bridge: 20
+          real_panda_single_arm: 21
+          hot3d_hands_only: 23
+          gr1_unified: 24
+          robocasa_gr1_arms_waist_fourier_hands: 25
+          agibot: 26
+          lapa: 27
+          oxe_mutex: 28
+          oxe_roboset: 29
+          oxe_plex: 30
+          dream: 31
+          yam: 32
+          xdof: 22
+          gr1_unified_segmentation: 14
+          language_table_sim: 7
+          gr1_isaac: 0
+          sim_behavior_r1_pro: 31
+          mecka_hands: 27
+          real_r1_pro_sharpa: 28
+          real_teleop_g1: 33
+        tokenizer_path: /hfm/boqian/liboqian_data/checkpoints/umt5-xxl
+    agibot:
+      _target_: groot.vla.data.transform.ComposedModalityTransform
+      transforms:
+      - _target_: groot.vla.data.transform.VideoToTensor
+        apply_to:
+        - video.top_head
+        - video.hand_left
+        - video.hand_right
+      - _target_: groot.vla.data.transform.VideoCrop
+        apply_to:
+        - video.top_head
+        - video.hand_left
+        - video.hand_right
+        scale: 0.95
+        mode: random
+      - _target_: groot.vla.data.transform.VideoResize
+        apply_to:
+        - video.top_head
+        - video.hand_left
+        - video.hand_right
+        height: 176
+        width: 320
+        interpolation: linear
+      - _target_: groot.vla.data.transform.VideoColorJitter
+        apply_to:
+        - video.top_head
+        - video.hand_left
+        - video.hand_right
+        brightness: 0.3
+        contrast: 0.4
+        saturation: 0.5
+        hue: 0.08
+      - _target_: groot.vla.data.transform.VideoToNumpy
+        apply_to:
+        - video.top_head
+        - video.hand_left
+        - video.hand_right
+      - _target_: groot.vla.data.transform.StateActionToTensor
+        apply_to:
+        - state.left_arm_joint_position
+        - state.right_arm_joint_position
+        - state.left_effector_position
+        - state.right_effector_position
+        - state.head_position
+        - state.waist_position
+      - _target_: groot.vla.data.transform.StateActionTransform
+        apply_to:
+        - state.left_arm_joint_position
+        - state.right_arm_joint_position
+        - state.left_effector_position
+        - state.right_effector_position
+        - state.head_position
+        - state.waist_position
+        normalization_modes:
+          state.left_arm_joint_position: q99
+          state.right_arm_joint_position: q99
+          state.left_effector_position: q99
+          state.right_effector_position: q99
+          state.head_position: q99
+          state.waist_position: q99
+      - _target_: groot.vla.data.transform.StateActionToTensor
+        apply_to:
+        - action.left_arm_joint_position
+        - action.right_arm_joint_position
+        - action.left_effector_position
+        - action.right_effector_position
+        - action.head_position
+        - action.waist_position
+        - action.robot_velocity
+      - _target_: groot.vla.data.transform.StateActionTransform
+        apply_to:
+        - action.left_arm_joint_position
+        - action.right_arm_joint_position
+        - action.left_effector_position
+        - action.right_effector_position
+        - action.head_position
+        - action.waist_position
+        - action.robot_velocity
+        normalization_modes:
+          action.left_arm_joint_position: q99
+          action.right_arm_joint_position: q99
+          action.left_effector_position: q99
+          action.right_effector_position: q99
+          action.head_position: q99
+          action.waist_position: q99
+          action.robot_velocity: q99
+      - _target_: groot.vla.data.transform.ConcatTransform
+        video_concat_order:
+        - video.top_head
+        - video.hand_left
+        - video.hand_right
+        state_concat_order:
+        - state.left_arm_joint_position
+        - state.right_arm_joint_position
+        - state.left_effector_position
+        - state.right_effector_position
+        - state.head_position
+        - state.waist_position
+        action_concat_order:
+        - action.left_arm_joint_position
+        - action.right_arm_joint_position
+        - action.left_effector_position
+        - action.right_effector_position
+        - action.head_position
+        - action.waist_position
+        - action.robot_velocity
+      - _target_: groot.vla.model.dreamzero.transform.dreamzero_cotrain.DreamTransform
+        default_instruction: Perform the default behavior.
+        language_dropout_prob: 0.0
+        always_use_default_instruction: false
+        max_state_dim: 64
+        max_action_dim: 36
+        max_length: 512
+        state_horizon: 1
+        action_horizon: 24
+        embodiment_tag_mapping:
+          real_gr1_arms_only: 0
+          real_gr1_arms_only_annotated: 1
+          real_gr1_arms_waist: 2
+          real_gr1_arms_waist_annotated: 3
+          dexmg_gr1_arms_only_inspire: 4
+          dexmg_gr1_arms_only_fourier: 5
+          dexmg_gr1_arms_waist_fourier: 6
+          robocasa_single_arm: 7
+          onex_eve_gripper: 8
+          robocasa_gr1_arms_only_inspire_hands: 9
+          robocasa_gr1_arms_only_fourier_hands: 10
+          robocasa_gr1_fixed_lower_body_inspire_hands: 11
+          robocasa_gr1_fixed_lower_body_fourier_hands: 12
+          robocasa_panda_omron: 13
+          robocasa_bimanual_panda_parallel_gripper: 15
+          robocasa_bimanual_panda_inspire_hand: 16
+          oxe_droid: 17
+          oxe_fractal: 18
+          oxe_language_table: 19
+          oxe_bridge: 20
+          real_panda_single_arm: 21
+          hot3d_hands_only: 23
+          gr1_unified: 24
+          robocasa_gr1_arms_waist_fourier_hands: 25
+          agibot: 26
+          lapa: 27
+          oxe_mutex: 28
+          oxe_roboset: 29
+          oxe_plex: 30
+          dream: 31
+          yam: 32
+          xdof: 22
+          gr1_unified_segmentation: 14
+          language_table_sim: 7
+          gr1_isaac: 0
+          sim_behavior_r1_pro: 31
+          mecka_hands: 27
+          real_r1_pro_sharpa: 28
+          real_teleop_g1: 33
+        tokenizer_path: /hfm/boqian/liboqian_data/checkpoints/umt5-xxl
+    yam:
+      _target_: groot.vla.data.transform.ComposedModalityTransform
+      transforms:
+      - _target_: groot.vla.data.transform.VideoToTensor
+        apply_to:
+        - video.top_camera-images-rgb
+        - video.left_camera-images-rgb
+        - video.right_camera-images-rgb
+      - _target_: groot.vla.data.transform.VideoCrop
+        apply_to:
+        - video.top_camera-images-rgb
+        - video.left_camera-images-rgb
+        - video.right_camera-images-rgb
+        scale: 0.95
+        mode: random
+      - _target_: groot.vla.data.transform.VideoResize
+        apply_to:
+        - video.top_camera-images-rgb
+        - video.left_camera-images-rgb
+        - video.right_camera-images-rgb
+        height: 176
+        width: 320
+        interpolation: linear
+      - _target_: groot.vla.data.transform.VideoColorJitter
+        apply_to:
+        - video.top_camera-images-rgb
+        - video.left_camera-images-rgb
+        - video.right_camera-images-rgb
+        brightness: 0.3
+        contrast: 0.4
+        saturation: 0.5
+        hue: 0.08
+      - _target_: groot.vla.data.transform.VideoToNumpy
+        apply_to:
+        - video.top_camera-images-rgb
+        - video.left_camera-images-rgb
+        - video.right_camera-images-rgb
+      - _target_: groot.vla.data.transform.StateActionToTensor
+        apply_to:
+        - state.left_joint_pos
+        - state.left_gripper_pos
+        - state.right_joint_pos
+        - state.right_gripper_pos
+      - _target_: groot.vla.data.transform.StateActionTransform
+        apply_to:
+        - state.left_joint_pos
+        - state.left_gripper_pos
+        - state.right_joint_pos
+        - state.right_gripper_pos
+        normalization_modes:
+          state.left_joint_pos: q99
+          state.left_gripper_pos: q99
+          state.right_joint_pos: q99
+          state.right_gripper_pos: q99
+      - _target_: groot.vla.data.transform.StateActionToTensor
+        apply_to:
+        - action.left_joint_pos
+        - action.left_gripper_pos
+        - action.right_joint_pos
+        - action.right_gripper_pos
+      - _target_: groot.vla.data.transform.StateActionTransform
+        apply_to:
+        - action.left_joint_pos
+        - action.left_gripper_pos
+        - action.right_joint_pos
+        - action.right_gripper_pos
+        normalization_modes:
+          action.left_joint_pos: q99
+          action.left_gripper_pos: q99
+          action.right_joint_pos: q99
+          action.right_gripper_pos: q99
+      - _target_: groot.vla.data.transform.ConcatTransform
+        video_concat_order:
+        - video.top_camera-images-rgb
+        - video.left_camera-images-rgb
+        - video.right_camera-images-rgb
+        state_concat_order:
+        - state.left_joint_pos
+        - state.left_gripper_pos
+        - state.right_joint_pos
+        - state.right_gripper_pos
+        action_concat_order:
+        - action.left_joint_pos
+        - action.left_gripper_pos
+        - action.right_joint_pos
+        - action.right_gripper_pos
+      - _target_: groot.vla.model.dreamzero.transform.dreamzero_cotrain.DreamTransform
+        default_instruction: Perform the default behavior.
+        language_dropout_prob: 0.0
+        always_use_default_instruction: false
+        max_state_dim: 64
+        max_action_dim: 36
+        max_length: 512
+        state_horizon: 1
+        action_horizon: 24
+        embodiment_tag_mapping:
+          real_gr1_arms_only: 0
+          real_gr1_arms_only_annotated: 1
+          real_gr1_arms_waist: 2
+          real_gr1_arms_waist_annotated: 3
+          dexmg_gr1_arms_only_inspire: 4
+          dexmg_gr1_arms_only_fourier: 5
+          dexmg_gr1_arms_waist_fourier: 6
+          robocasa_single_arm: 7
+          onex_eve_gripper: 8
+          robocasa_gr1_arms_only_inspire_hands: 9
+          robocasa_gr1_arms_only_fourier_hands: 10
+          robocasa_gr1_fixed_lower_body_inspire_hands: 11
+          robocasa_gr1_fixed_lower_body_fourier_hands: 12
+          robocasa_panda_omron: 13
+          robocasa_bimanual_panda_parallel_gripper: 15
+          robocasa_bimanual_panda_inspire_hand: 16
+          oxe_droid: 17
+          oxe_fractal: 18
+          oxe_language_table: 19
+          oxe_bridge: 20
+          real_panda_single_arm: 21
+          hot3d_hands_only: 23
+          gr1_unified: 24
+          robocasa_gr1_arms_waist_fourier_hands: 25
+          agibot: 26
+          lapa: 27
+          oxe_mutex: 28
+          oxe_roboset: 29
+          oxe_plex: 30
+          dream: 31
+          yam: 32
+          xdof: 22
+          gr1_unified_segmentation: 14
+          language_table_sim: 7
+          gr1_isaac: 0
+          sim_behavior_r1_pro: 31
+          mecka_hands: 27
+          real_r1_pro_sharpa: 28
+          real_teleop_g1: 33
+        tokenizer_path: /hfm/boqian/liboqian_data/checkpoints/umt5-xxl
+    real_teleop_g1:
+      _target_: groot.vla.data.transform.ComposedModalityTransform
+      transforms:
+      - _target_: groot.vla.data.transform.VideoToTensor
+        apply_to:
+        - video.egocentric
+      - _target_: groot.vla.data.transform.VideoCrop
+        apply_to:
+        - video.egocentric
+        scale: 0.95
+        mode: random
+      - _target_: groot.vla.data.transform.VideoResize
+        apply_to:
+        - video.egocentric
+        height: 176
+        width: 320
+        interpolation: linear
+      - _target_: groot.vla.data.transform.VideoColorJitter
+        apply_to:
+        - video.egocentric
+        brightness: 0.3
+        contrast: 0.4
+        saturation: 0.5
+        hue: 0.08
+      - _target_: groot.vla.data.transform.VideoToNumpy
+        apply_to:
+        - video.egocentric
+      - _target_: groot.vla.data.transform.StateActionToTensor
+        apply_to:
+        - state.left_hand
+        - state.right_hand
+        - state.left_arm
+        - state.right_arm
+        - state.rpy
+        - state.height
+      - _target_: groot.vla.data.transform.StateActionTransform
+        apply_to:
+        - state.left_hand
+        - state.right_hand
+        - state.left_arm
+        - state.right_arm
+        - state.rpy
+        - state.height
+        normalization_modes:
+          state.left_hand: q99
+          state.right_hand: q99
+          state.left_arm: q99
+          state.right_arm: q99
+          state.rpy: q99
+          state.height: q99
+      - _target_: groot.vla.data.transform.StateActionToTensor
+        apply_to:
+        - action.left_hand
+        - action.right_hand
+        - action.left_arm
+        - action.right_arm
+        - action.rpy
+        - action.height
+        - action.torso_vx
+        - action.torso_vy
+        - action.torso_vyaw
+        - action.torso_dyaw
+      - _target_: groot.vla.data.transform.StateActionTransform
+        apply_to:
+        - action.left_hand
+        - action.right_hand
+        - action.left_arm
+        - action.right_arm
+        - action.rpy
+        - action.height
+        - action.torso_vx
+        - action.torso_vy
+        - action.torso_vyaw
+        - action.torso_dyaw
+        normalization_modes:
+          action.left_hand: q99
+          action.right_hand: q99
+          action.left_arm: q99
+          action.right_arm: q99
+          action.rpy: q99
+          action.height: q99
+          action.torso_vx: q99
+          action.torso_vy: q99
+          action.torso_vyaw: q99
+          action.torso_dyaw: q99
+      - _target_: groot.vla.data.transform.ConcatTransform
+        video_concat_order:
+        - video.egocentric
+        state_concat_order:
+        - state.left_hand
+        - state.right_hand
+        - state.left_arm
+        - state.right_arm
+        - state.rpy
+        - state.height
+        action_concat_order:
+        - action.left_hand
+        - action.right_hand
+        - action.left_arm
+        - action.right_arm
+        - action.rpy
+        - action.height
+        - action.torso_vx
+        - action.torso_vy
+        - action.torso_vyaw
+        - action.torso_dyaw
+      - _target_: groot.vla.model.dreamzero.transform.dreamzero_cotrain.DreamTransform
+        default_instruction: Perform the default behavior.
+        language_dropout_prob: 0.0
+        always_use_default_instruction: false
+        max_state_dim: 64
+        max_action_dim: 36
+        max_length: 512
+        state_horizon: 1
+        action_horizon: 24
+        embodiment_tag_mapping:
+          real_gr1_arms_only: 0
+          real_gr1_arms_only_annotated: 1
+          real_gr1_arms_waist: 2
+          real_gr1_arms_waist_annotated: 3
+          dexmg_gr1_arms_only_inspire: 4
+          dexmg_gr1_arms_only_fourier: 5
+          dexmg_gr1_arms_waist_fourier: 6
+          robocasa_single_arm: 7
+          onex_eve_gripper: 8
+          robocasa_gr1_arms_only_inspire_hands: 9
+          robocasa_gr1_arms_only_fourier_hands: 10
+          robocasa_gr1_fixed_lower_body_inspire_hands: 11
+          robocasa_gr1_fixed_lower_body_fourier_hands: 12
+          robocasa_panda_omron: 13
+          robocasa_bimanual_panda_parallel_gripper: 15
+          robocasa_bimanual_panda_inspire_hand: 16
+          oxe_droid: 17
+          oxe_fractal: 18
+          oxe_language_table: 19
+          oxe_bridge: 20
+          real_panda_single_arm: 21
+          hot3d_hands_only: 23
+          gr1_unified: 24
+          robocasa_gr1_arms_waist_fourier_hands: 25
+          agibot: 26
+          lapa: 27
+          oxe_mutex: 28
+          oxe_roboset: 29
+          oxe_plex: 30
+          dream: 31
+          yam: 32
+          xdof: 22
+          gr1_unified_segmentation: 14
+          language_table_sim: 7
+          gr1_isaac: 0
+          sim_behavior_r1_pro: 31
+          mecka_hands: 27
+          real_r1_pro_sharpa: 28
+          real_teleop_g1: 33
+        tokenizer_path: /hfm/boqian/liboqian_data/checkpoints/umt5-xxl
+  metadata_versions:
+    oxe_droid: '0221'
+    agibot: '0221'
+    yam: '0221'
+    real_teleop_g1: '0221'
+  fps:
+    yam: 30
+    real_teleop_g1: 30
+  dataset_kwargs:
+    video_backend: decord
+    use_global_metadata: false
+    max_chunk_size: 4
+    relative_action: false
+    relative_action_keys: []
+    relative_action_per_horizon: false
+  mixture_kwargs:
+    training: true
+    balance_dataset_weights: false
+    seed: 42
+    shard_sampling_rate: 0.1
+trainer:
+  _target_: groot.vla.experiment.VLATrainer
+  _partial_: true
+  _recursive_: false
+  callbacks: null
+  model: ???
+  train_dataset: ???
+  compute_dtype: ???
+  benchmark_time: false
+  enable_profiling: false
+  profiling_steps: 5
+  enable_prof_callback: false
+  profile_start_step: 50
+  profile_warmup_steps: 1
+  profile_active_steps: 3
+  profile_record_shapes: false
+  profile_with_stack: false
+  profile_memory: false
+wandb_project: dreamzero
+output_dir: ./checkpoints/dreamzero_real_teleop_g1_full_finetune_absolute
+load_from_yaml: null
+gear_credentials: null
+upload_checkpoints: false
+upload_every: 1000
+upload_last_n_checkpoints: 5
+remove_unused_columns: false
+bf16: true
+tf32: true
+global_batch_size: null
+raise_error_if_global_batch_size_not_set: false
+per_device_train_batch_size: 1
+per_device_eval_batch_size: 64
+gradient_accumulation_steps: 1
+dataloader_num_workers: 1
+dataloader_pin_memory: false
+dataloader_persistent_workers: true
+optim: adamw_torch
+learning_rate: 0.0001
+adam_beta1: 0.95
+adam_beta2: 0.999
+adam_epsilon: 1.0e-08
+weight_decay: 1.0e-05
+lr_scheduler_type: cosine
+warmup_ratio: 0.05
+logging_steps: 10.0
+num_train_epochs: 1000
+max_steps: 20000
+save_strategy: steps
+save_steps: 8000
+eval_strategy: 'no'
+save_total_limit: 10
+report_to: wandb
+seed: 42
+do_eval: false
+gradient_checkpointing: false
+ddp_find_unused_parameters: false
+ddp_bucket_cap_mb: 100
+ray_num_workers: ???
+eval_bf16: true
+torch_compile_mode: null
+pretrained_model_path: /hfm/boqian/liboqian_data/checkpoints/DreamZero-AgiBot
+only_tune_projectors: false
+save_llm: false
+save_lora_only: false
+save_value_model: false
+save_q_model: false
+download_cache: false
+training_args:
+  _target_: transformers.TrainingArguments
+  output_dir: ./checkpoints/dreamzero_real_teleop_g1_full_finetune_absolute
+  run_name: dreamzero_real_teleop_g1_full_finetune_absolute
+  remove_unused_columns: false
+  deepspeed: groot/vla/configs/deepspeed/zero2_offload.json
+  gradient_checkpointing: false
+  bf16: true
+  tf32: true
+  per_device_train_batch_size: 1
+  per_device_eval_batch_size: 64
+  gradient_accumulation_steps: 1
+  dataloader_num_workers: 1
+  dataloader_pin_memory: false
+  dataloader_persistent_workers: true
+  optim: adamw_torch
+  adam_beta1: 0.95
+  adam_beta2: 0.999
+  adam_epsilon: 1.0e-08
+  learning_rate: 1.0e-05
+  weight_decay: 1.0e-05
+  warmup_ratio: 0.05
+  lr_scheduler_type: cosine
+  logging_steps: 10.0
+  num_train_epochs: 1000
+  max_steps: 20000
+  save_strategy: steps
+  save_steps: 8000
+  save_total_limit: 10
+  report_to: wandb
+  seed: 42
+  do_eval: false
+  ddp_find_unused_parameters: false
+  ddp_bucket_cap_mb: 100
+  torch_compile_mode: null
+profile_dir: null
+backbone_hidden_size: 0
+backbone_cfg:
+  _target_: groot.vla.model.dreamzero.backbone.identity.IdentityBackbone
+action_head_cfg:
+  config:
+    backbone_features_projector_cfg: null
+    _target_: groot.vla.model.dreamzero.action_head.wan_flow_matching_action_tf.WANPolicyHeadConfig
+    _recursive_: false
+    tiled: false
+    tile_size_height: 34
+    tile_size_width: 34
+    tile_stride_height: 18
+    tile_stride_width: 16
+    lora_rank: 4
+    lora_alpha: 4
+    num_frames: 33
+    num_frame_per_block: 2
+    lora_target_modules: q,k,v,o,ffn.0,ffn.2
+    init_lora_weights: kaiming
+    train_architecture: full
+    use_gradient_checkpointing: true
+    add_pos_embed: true
+    model_dtype: float32
+    max_state_dim: 64
+    max_action_dim: 36
+    action_loss_embodiment_ids:
+    - 26
+    - 17
+    - 32
+    hidden_size: 64
+    input_embedding_dim: 1536
+    backbone_embedding_dim: 0
+    repa_layer: 8
+    repa_coeff: 1.0
+    load_pretrained_det_decode_layer_path: null
+    freeze_decode_layer: false
+    expand_batch: null
+    use_vlln: true
+    vl_self_attention_cfg:
+      _target_: groot.vla.model.n1_5.modules.cross_attention_dit.SelfAttentionTransformer
+      positional_embeddings: null
+      num_layers: 4
+      num_attention_heads: 24
+      attention_head_dim: 64
+      dropout: 0.2
+      final_dropout: true
+    diffusion_model_cfg:
+      _target_: groot.vla.model.dreamzero.modules.wan_video_dit_action_casual_chunk.CausalWanModel
+      _convert_: object
+      diffusion_model_pretrained_path: /hfm/boqian/liboqian_data/checkpoints/Wan2.1-I2V-14B-480P
+      model_type: i2v
+      frame_seqlen: 220
+      dim: 5120
+      in_dim: 36
+      ffn_dim: 13824
+      out_dim: 16
+      freq_dim: 256
+      eps: 1.0e-06
+      num_heads: 40
+      num_layers: 40
+      max_chunk_size: 4
+      num_frame_per_block: 2
+      num_action_per_block: 24
+      num_state_per_block: 1
+      action_dim: 36
+    text_encoder_cfg:
+      _target_: groot.vla.model.dreamzero.modules.wan_video_text_encoder.WanTextEncoder
+      _convert_: object
+      text_encoder_pretrained_path: /hfm/boqian/liboqian_data/checkpoints/Wan2.1-I2V-14B-480P/models_t5_umt5-xxl-enc-bf16.pth
+    image_encoder_cfg:
+      _target_: groot.vla.model.dreamzero.modules.wan_video_image_encoder.WanImageEncoder
+      _convert_: object
+      image_encoder_pretrained_path: /hfm/boqian/liboqian_data/checkpoints/Wan2.1-I2V-14B-480P/models_clip_open-clip-xlm-roberta-large-vit-huge-14.pth
+    vae_cfg:
+      _target_: groot.vla.model.dreamzero.modules.wan_video_vae.WanVideoVAE
+      _convert_: object
+      vae_pretrained_path: /hfm/boqian/liboqian_data/checkpoints/Wan2.1-I2V-14B-480P/Wan2.1_VAE.pth
+    action_dim: 36
+    action_horizon: 24
+    num_inference_timesteps: 4
+    noise_beta_alpha: 1.5
+    noise_beta_beta: 1.0
+    noise_s: 0.999
+    num_timestep_buckets: 1000
+    decouple_video_action_noise: false
+    video_noise_beta_alpha: 3.0
+    video_noise_beta_beta: 1.0
+    tune_projector: true
+    tune_diffusion_model: true
+    skip_component_loading: true
+  _target_: groot.vla.model.dreamzero.action_head.wan_flow_matching_action_tf.WANPolicyHead
+  _convert_: object
+add_pos_embed: true
+hidden_size: 64
+attn_dropout: 0.2
+repa_layer: 8
+repa_coeff: 1.0
+load_pretrained_det_decode_layer_path: null
+expand_batch: null
+dit_version: /hfm/boqian/liboqian_data/checkpoints/Wan2.1-I2V-14B-480P
+text_encoder_pretrained_path: /hfm/boqian/liboqian_data/checkpoints/Wan2.1-I2V-14B-480P/models_t5_umt5-xxl-enc-bf16.pth
+image_encoder_pretrained_path: /hfm/boqian/liboqian_data/checkpoints/Wan2.1-I2V-14B-480P/models_clip_open-clip-xlm-roberta-large-vit-huge-14.pth
+vae_pretrained_path: /hfm/boqian/liboqian_data/checkpoints/Wan2.1-I2V-14B-480P/Wan2.1_VAE.pth
+train_architecture: full
+num_frame_per_block: 2
+num_action_per_block: 24
+num_state_per_block: 1
+frame_seqlen: 220
+embodiment_tag_to_projector_index:
+  real_gr1_arms_only: 0
+  real_gr1_arms_only_annotated: 1
+  real_gr1_arms_waist: 2
+  real_gr1_arms_waist_annotated: 3
+  dexmg_gr1_arms_only_inspire: 4
+  dexmg_gr1_arms_only_fourier: 5
+  dexmg_gr1_arms_waist_fourier: 6
+  robocasa_single_arm: 7
+  onex_eve_gripper: 8
+  robocasa_gr1_arms_only_inspire_hands: 9
+  robocasa_gr1_arms_only_fourier_hands: 10
+  robocasa_gr1_fixed_lower_body_inspire_hands: 11
+  robocasa_gr1_fixed_lower_body_fourier_hands: 12
+  robocasa_panda_omron: 13
+  robocasa_bimanual_panda_parallel_gripper: 15
+  robocasa_bimanual_panda_inspire_hand: 16
+  oxe_droid: 17
+  oxe_fractal: 18
+  oxe_language_table: 19
+  oxe_bridge: 20
+  real_panda_single_arm: 21
+  hot3d_hands_only: 23
+  gr1_unified: 24
+  robocasa_gr1_arms_waist_fourier_hands: 25
+  agibot: 26
+  lapa: 27
+  oxe_mutex: 28
+  oxe_roboset: 29
+  oxe_plex: 30
+  dream: 31
+  yam: 32
+  xdof: 22
+  gr1_unified_segmentation: 14
+  language_table_sim: 7
+  gr1_isaac: 0
+  sim_behavior_r1_pro: 31
+  mecka_hands: 27
+  real_r1_pro_sharpa: 28
+  real_teleop_g1: 33
+max_length: 512
+num_views: 1
+tokenizer_path: /hfm/boqian/liboqian_data/checkpoints/umt5-xxl
+data_collator:
+  _target_: groot.vla.model.dreamzero.transform.dreamzero_cotrain.DefaultDataCollator
+  tokenizer_path: /hfm/boqian/liboqian_data/checkpoints/umt5-xxl
+  max_length: 512
+  num_views: 1
+  embodiment_tag_mapping:
+    real_gr1_arms_only: 0
+    real_gr1_arms_only_annotated: 1
+    real_gr1_arms_waist: 2
+    real_gr1_arms_waist_annotated: 3
+    dexmg_gr1_arms_only_inspire: 4
+    dexmg_gr1_arms_only_fourier: 5
+    dexmg_gr1_arms_waist_fourier: 6
+    robocasa_single_arm: 7
+    onex_eve_gripper: 8
+    robocasa_gr1_arms_only_inspire_hands: 9
+    robocasa_gr1_arms_only_fourier_hands: 10
+    robocasa_gr1_fixed_lower_body_inspire_hands: 11
+    robocasa_gr1_fixed_lower_body_fourier_hands: 12
+    robocasa_panda_omron: 13
+    robocasa_bimanual_panda_parallel_gripper: 15
+    robocasa_bimanual_panda_inspire_hand: 16
+    oxe_droid: 17
+    oxe_fractal: 18
+    oxe_language_table: 19
+    oxe_bridge: 20
+    real_panda_single_arm: 21
+    hot3d_hands_only: 23
+    gr1_unified: 24
+    robocasa_gr1_arms_waist_fourier_hands: 25
+    agibot: 26
+    lapa: 27
+    oxe_mutex: 28
+    oxe_roboset: 29
+    oxe_plex: 30
+    dream: 31
+    yam: 32
+    xdof: 22
+    gr1_unified_segmentation: 14
+    language_table_sim: 7
+    gr1_isaac: 0
+    sim_behavior_r1_pro: 31
+    mecka_hands: 27
+    real_r1_pro_sharpa: 28
+    real_teleop_g1: 33
+num_visual_tokens_per_frame: 16
+max_state_dim: 64
+max_action_dim: 36
+language_dropout_prob: 0.0
+model_specific_transform:
+  _target_: groot.vla.model.dreamzero.transform.dreamzero_cotrain.DreamTransform
+  default_instruction: Perform the default behavior.
+  language_dropout_prob: 0.0
+  always_use_default_instruction: false
+  max_state_dim: 64
+  max_action_dim: 36
+  max_length: 512
+  state_horizon: 1
+  action_horizon: 24
+  embodiment_tag_mapping:
+    real_gr1_arms_only: 0
+    real_gr1_arms_only_annotated: 1
+    real_gr1_arms_waist: 2
+    real_gr1_arms_waist_annotated: 3
+    dexmg_gr1_arms_only_inspire: 4
+    dexmg_gr1_arms_only_fourier: 5
+    dexmg_gr1_arms_waist_fourier: 6
+    robocasa_single_arm: 7
+    onex_eve_gripper: 8
+    robocasa_gr1_arms_only_inspire_hands: 9
+    robocasa_gr1_arms_only_fourier_hands: 10
+    robocasa_gr1_fixed_lower_body_inspire_hands: 11
+    robocasa_gr1_fixed_lower_body_fourier_hands: 12
+    robocasa_panda_omron: 13
+    robocasa_bimanual_panda_parallel_gripper: 15
+    robocasa_bimanual_panda_inspire_hand: 16
+    oxe_droid: 17
+    oxe_fractal: 18
+    oxe_language_table: 19
+    oxe_bridge: 20
+    real_panda_single_arm: 21
+    hot3d_hands_only: 23
+    gr1_unified: 24
+    robocasa_gr1_arms_waist_fourier_hands: 25
+    agibot: 26
+    lapa: 27
+    oxe_mutex: 28
+    oxe_roboset: 29
+    oxe_plex: 30
+    dream: 31
+    yam: 32
+    xdof: 22
+    gr1_unified_segmentation: 14
+    language_table_sim: 7
+    gr1_isaac: 0
+    sim_behavior_r1_pro: 31
+    mecka_hands: 27
+    real_r1_pro_sharpa: 28
+    real_teleop_g1: 33
+  tokenizer_path: /hfm/boqian/liboqian_data/checkpoints/umt5-xxl
+use_global_metadata: false
+num_frames: 33
+action_horizon: 24
+state_horizon: 1
+image_resolution_width: 320
+image_resolution_height: 176
+image_resolution_width_single_frame: 256
+image_resolution_height_single_frame: 256
+totensor_cfg:
+  _target_: groot.vla.data.transform.VideoToTensor
+  apply_to: ???
+crop_cfg:
+  _target_: groot.vla.data.transform.VideoCrop
+  apply_to: ???
+  scale: 0.95
+  mode: random
+resize_cfg:
+  _target_: groot.vla.data.transform.VideoResize
+  apply_to: ???
+  height: 176
+  width: 320
+  interpolation: linear
+resize_cfg_single_frame:
+  _target_: groot.vla.data.transform.VideoResize
+  apply_to: ???
+  height: 256
+  width: 256
+  interpolation: linear
+color_jitter_cfg:
+  _target_: groot.vla.data.transform.VideoColorJitter
+  apply_to: ???
+  brightness: 0.3
+  contrast: 0.4
+  saturation: 0.5
+  hue: 0.08
+random_grayscale_cfg:
+  _target_: groot.vla.data.transform.VideoRandomGrayscale
+  apply_to: ???
+  p: 0.1
+random_posterize_cfg:
+  _target_: groot.vla.data.transform.VideoRandomPosterize
+  apply_to: ???
+  bits: 4
+  p: 0.1
+normalize_cfg:
+  _target_: groot.vla.data.transform.VideoNormalize
+  apply_to: ???
+  mean:
+  - 0.5
+  - 0.5
+  - 0.5
+  std:
+  - 0.5
+  - 0.5
+  - 0.5
+to_numpy_cfg:
+  _target_: groot.vla.data.transform.VideoToNumpy
+  apply_to: ???
+modality_config_oxe_droid:
+  video:
+    _target_: groot.vla.data.dataset.ModalityConfig
+    delta_indices:
+    - 0
+    - 1
+    - 2
+    - 3
+    - 4
+    - 5
+    - 6
+    - 7
+    - 8
+    - 9
+    - 10
+    - 11
+    - 12
+    - 13
+    - 14
+    - 15
+    - 16
+    - 17
+    - 18
+    - 19
+    - 20
+    - 21
+    - 22
+    - 23
+    - 24
+    eval_delta_indices:
+    - 0
+    modality_keys:
+    - video.exterior_image_1_left
+    - video.exterior_image_2_left
+    - video.wrist_image_left
+  state:
+    _target_: groot.vla.data.dataset.ModalityConfig
+    delta_indices:
+    - 0
+    modality_keys:
+    - state.joint_position
+    - state.gripper_position
+  action:
+    _target_: groot.vla.data.dataset.ModalityConfig
+    delta_indices:
+    - 0
+    - 1
+    - 2
+    - 3
+    - 4
+    - 5
+    - 6
+    - 7
+    - 8
+    - 9
+    - 10
+    - 11
+    - 12
+    - 13
+    - 14
+    - 15
+    - 16
+    - 17
+    - 18
+    - 19
+    - 20
+    - 21
+    - 22
+    - 23
+    modality_keys:
+    - action.joint_position
+    - action.gripper_position
+  language:
+    _target_: groot.vla.data.dataset.ModalityConfig
+    delta_indices:
+    - 0
+    modality_keys:
+    - annotation.language.language_instruction
+    - annotation.language.language_instruction_2
+    - annotation.language.language_instruction_3
+  lapa_action:
+    _target_: groot.vla.data.dataset.ModalityConfig
+    delta_indices:
+    - 0
+    modality_keys:
+    - lapa_action
+transform_oxe_droid:
+  _target_: groot.vla.data.transform.ComposedModalityTransform
+  transforms:
+  - _target_: groot.vla.data.transform.VideoToTensor
+    apply_to:
+    - video.exterior_image_1_left
+    - video.exterior_image_2_left
+    - video.wrist_image_left
+  - _target_: groot.vla.data.transform.VideoCrop
+    apply_to:
+    - video.exterior_image_1_left
+    - video.exterior_image_2_left
+    - video.wrist_image_left
+    scale: 0.95
+    mode: random
+  - _target_: groot.vla.data.transform.VideoResize
+    apply_to:
+    - video.exterior_image_1_left
+    - video.exterior_image_2_left
+    - video.wrist_image_left
+    height: 176
+    width: 320
+    interpolation: linear
+  - _target_: groot.vla.data.transform.VideoColorJitter
+    apply_to:
+    - video.exterior_image_1_left
+    - video.exterior_image_2_left
+    - video.wrist_image_left
+    brightness: 0.3
+    contrast: 0.4
+    saturation: 0.5
+    hue: 0.08
+  - _target_: groot.vla.data.transform.VideoToNumpy
+    apply_to:
+    - video.exterior_image_1_left
+    - video.exterior_image_2_left
+    - video.wrist_image_left
+  - _target_: groot.vla.data.transform.StateActionToTensor
+    apply_to:
+    - state.joint_position
+    - state.gripper_position
+  - _target_: groot.vla.data.transform.StateActionTransform
+    apply_to:
+    - state.joint_position
+    - state.gripper_position
+    normalization_modes:
+      state.joint_position: q99
+      state.gripper_position: q99
+  - _target_: groot.vla.data.transform.StateActionToTensor
+    apply_to:
+    - action.joint_position
+    - action.gripper_position
+  - _target_: groot.vla.data.transform.StateActionTransform
+    apply_to:
+    - action.joint_position
+    - action.gripper_position
+    normalization_modes:
+      action.joint_position: q99
+      action.gripper_position: q99
+  - _target_: groot.vla.data.transform.ConcatTransform
+    video_concat_order:
+    - video.exterior_image_1_left
+    - video.exterior_image_2_left
+    - video.wrist_image_left
+    state_concat_order:
+    - state.joint_position
+    - state.gripper_position
+    action_concat_order:
+    - action.joint_position
+    - action.gripper_position
+  - _target_: groot.vla.model.dreamzero.transform.dreamzero_cotrain.DreamTransform
+    default_instruction: Perform the default behavior.
+    language_dropout_prob: 0.0
+    always_use_default_instruction: false
+    max_state_dim: 64
+    max_action_dim: 36
+    max_length: 512
+    state_horizon: 1
+    action_horizon: 24
+    embodiment_tag_mapping:
+      real_gr1_arms_only: 0
+      real_gr1_arms_only_annotated: 1
+      real_gr1_arms_waist: 2
+      real_gr1_arms_waist_annotated: 3
+      dexmg_gr1_arms_only_inspire: 4
+      dexmg_gr1_arms_only_fourier: 5
+      dexmg_gr1_arms_waist_fourier: 6
+      robocasa_single_arm: 7
+      onex_eve_gripper: 8
+      robocasa_gr1_arms_only_inspire_hands: 9
+      robocasa_gr1_arms_only_fourier_hands: 10
+      robocasa_gr1_fixed_lower_body_inspire_hands: 11
+      robocasa_gr1_fixed_lower_body_fourier_hands: 12
+      robocasa_panda_omron: 13
+      robocasa_bimanual_panda_parallel_gripper: 15
+      robocasa_bimanual_panda_inspire_hand: 16
+      oxe_droid: 17
+      oxe_fractal: 18
+      oxe_language_table: 19
+      oxe_bridge: 20
+      real_panda_single_arm: 21
+      hot3d_hands_only: 23
+      gr1_unified: 24
+      robocasa_gr1_arms_waist_fourier_hands: 25
+      agibot: 26
+      lapa: 27
+      oxe_mutex: 28
+      oxe_roboset: 29
+      oxe_plex: 30
+      dream: 31
+      yam: 32
+      xdof: 22
+      gr1_unified_segmentation: 14
+      language_table_sim: 7
+      gr1_isaac: 0
+      sim_behavior_r1_pro: 31
+      mecka_hands: 27
+      real_r1_pro_sharpa: 28
+      real_teleop_g1: 33
+    tokenizer_path: /hfm/boqian/liboqian_data/checkpoints/umt5-xxl
+modality_config_agibot:
+  video:
+    _target_: groot.vla.data.dataset.ModalityConfig
+    delta_indices:
+    - 0
+    - 1
+    - 2
+    - 3
+    - 4
+    - 5
+    - 6
+    - 7
+    - 8
+    - 9
+    - 10
+    - 11
+    - 12
+    - 13
+    - 14
+    - 15
+    - 16
+    - 17
+    - 18
+    - 19
+    - 20
+    - 21
+    - 22
+    - 23
+    - 24
+    eval_delta_indices:
+    - -3
+    - -2
+    - -1
+    - 0
+    modality_keys:
+    - video.top_head
+    - video.hand_left
+    - video.hand_right
+  state:
+    _target_: groot.vla.data.dataset.ModalityConfig
+    delta_indices:
+    - 0
+    modality_keys:
+    - state.left_arm_joint_position
+    - state.right_arm_joint_position
+    - state.left_effector_position
+    - state.right_effector_position
+    - state.head_position
+    - state.waist_position
+  action:
+    _target_: groot.vla.data.dataset.ModalityConfig
+    delta_indices:
+    - 0
+    - 1
+    - 2
+    - 3
+    - 4
+    - 5
+    - 6
+    - 7
+    - 8
+    - 9
+    - 10
+    - 11
+    - 12
+    - 13
+    - 14
+    - 15
+    - 16
+    - 17
+    - 18
+    - 19
+    - 20
+    - 21
+    - 22
+    - 23
+    modality_keys:
+    - action.left_arm_joint_position
+    - action.right_arm_joint_position
+    - action.left_effector_position
+    - action.right_effector_position
+    - action.head_position
+    - action.waist_position
+    - action.robot_velocity
+  language:
+    _target_: groot.vla.data.dataset.ModalityConfig
+    delta_indices:
+    - 0
+    modality_keys:
+    - annotation.language.action_text
+transform_agibot:
+  _target_: groot.vla.data.transform.ComposedModalityTransform
+  transforms:
+  - _target_: groot.vla.data.transform.VideoToTensor
+    apply_to:
+    - video.top_head
+    - video.hand_left
+    - video.hand_right
+  - _target_: groot.vla.data.transform.VideoCrop
+    apply_to:
+    - video.top_head
+    - video.hand_left
+    - video.hand_right
+    scale: 0.95
+    mode: random
+  - _target_: groot.vla.data.transform.VideoResize
+    apply_to:
+    - video.top_head
+    - video.hand_left
+    - video.hand_right
+    height: 176
+    width: 320
+    interpolation: linear
+  - _target_: groot.vla.data.transform.VideoColorJitter
+    apply_to:
+    - video.top_head
+    - video.hand_left
+    - video.hand_right
+    brightness: 0.3
+    contrast: 0.4
+    saturation: 0.5
+    hue: 0.08
+  - _target_: groot.vla.data.transform.VideoToNumpy
+    apply_to:
+    - video.top_head
+    - video.hand_left
+    - video.hand_right
+  - _target_: groot.vla.data.transform.StateActionToTensor
+    apply_to:
+    - state.left_arm_joint_position
+    - state.right_arm_joint_position
+    - state.left_effector_position
+    - state.right_effector_position
+    - state.head_position
+    - state.waist_position
+  - _target_: groot.vla.data.transform.StateActionTransform
+    apply_to:
+    - state.left_arm_joint_position
+    - state.right_arm_joint_position
+    - state.left_effector_position
+    - state.right_effector_position
+    - state.head_position
+    - state.waist_position
+    normalization_modes:
+      state.left_arm_joint_position: q99
+      state.right_arm_joint_position: q99
+      state.left_effector_position: q99
+      state.right_effector_position: q99
+      state.head_position: q99
+      state.waist_position: q99
+  - _target_: groot.vla.data.transform.StateActionToTensor
+    apply_to:
+    - action.left_arm_joint_position
+    - action.right_arm_joint_position
+    - action.left_effector_position
+    - action.right_effector_position
+    - action.head_position
+    - action.waist_position
+    - action.robot_velocity
+  - _target_: groot.vla.data.transform.StateActionTransform
+    apply_to:
+    - action.left_arm_joint_position
+    - action.right_arm_joint_position
+    - action.left_effector_position
+    - action.right_effector_position
+    - action.head_position
+    - action.waist_position
+    - action.robot_velocity
+    normalization_modes:
+      action.left_arm_joint_position: q99
+      action.right_arm_joint_position: q99
+      action.left_effector_position: q99
+      action.right_effector_position: q99
+      action.head_position: q99
+      action.waist_position: q99
+      action.robot_velocity: q99
+  - _target_: groot.vla.data.transform.ConcatTransform
+    video_concat_order:
+    - video.top_head
+    - video.hand_left
+    - video.hand_right
+    state_concat_order:
+    - state.left_arm_joint_position
+    - state.right_arm_joint_position
+    - state.left_effector_position
+    - state.right_effector_position
+    - state.head_position
+    - state.waist_position
+    action_concat_order:
+    - action.left_arm_joint_position
+    - action.right_arm_joint_position
+    - action.left_effector_position
+    - action.right_effector_position
+    - action.head_position
+    - action.waist_position
+    - action.robot_velocity
+  - _target_: groot.vla.model.dreamzero.transform.dreamzero_cotrain.DreamTransform
+    default_instruction: Perform the default behavior.
+    language_dropout_prob: 0.0
+    always_use_default_instruction: false
+    max_state_dim: 64
+    max_action_dim: 36
+    max_length: 512
+    state_horizon: 1
+    action_horizon: 24
+    embodiment_tag_mapping:
+      real_gr1_arms_only: 0
+      real_gr1_arms_only_annotated: 1
+      real_gr1_arms_waist: 2
+      real_gr1_arms_waist_annotated: 3
+      dexmg_gr1_arms_only_inspire: 4
+      dexmg_gr1_arms_only_fourier: 5
+      dexmg_gr1_arms_waist_fourier: 6
+      robocasa_single_arm: 7
+      onex_eve_gripper: 8
+      robocasa_gr1_arms_only_inspire_hands: 9
+      robocasa_gr1_arms_only_fourier_hands: 10
+      robocasa_gr1_fixed_lower_body_inspire_hands: 11
+      robocasa_gr1_fixed_lower_body_fourier_hands: 12
+      robocasa_panda_omron: 13
+      robocasa_bimanual_panda_parallel_gripper: 15
+      robocasa_bimanual_panda_inspire_hand: 16
+      oxe_droid: 17
+      oxe_fractal: 18
+      oxe_language_table: 19
+      oxe_bridge: 20
+      real_panda_single_arm: 21
+      hot3d_hands_only: 23
+      gr1_unified: 24
+      robocasa_gr1_arms_waist_fourier_hands: 25
+      agibot: 26
+      lapa: 27
+      oxe_mutex: 28
+      oxe_roboset: 29
+      oxe_plex: 30
+      dream: 31
+      yam: 32
+      xdof: 22
+      gr1_unified_segmentation: 14
+      language_table_sim: 7
+      gr1_isaac: 0
+      sim_behavior_r1_pro: 31
+      mecka_hands: 27
+      real_r1_pro_sharpa: 28
+      real_teleop_g1: 33
+    tokenizer_path: /hfm/boqian/liboqian_data/checkpoints/umt5-xxl
+modality_config_yam:
+  video:
+    _target_: groot.vla.data.dataset.ModalityConfig
+    delta_indices:
+    - 0
+    - 1
+    - 2
+    - 3
+    - 4
+    - 5
+    - 6
+    - 7
+    - 8
+    - 9
+    - 10
+    - 11
+    - 12
+    - 13
+    - 14
+    - 15
+    - 16
+    - 17
+    - 18
+    - 19
+    - 20
+    - 21
+    - 22
+    - 23
+    - 24
+    eval_delta_indices:
+    - 0
+    modality_keys:
+    - video.top_camera-images-rgb
+    - video.left_camera-images-rgb
+    - video.right_camera-images-rgb
+  state:
+    _target_: groot.vla.data.dataset.ModalityConfig
+    delta_indices:
+    - 0
+    modality_keys:
+    - state.left_joint_pos
+    - state.left_gripper_pos
+    - state.right_joint_pos
+    - state.right_gripper_pos
+  action:
+    _target_: groot.vla.data.dataset.ModalityConfig
+    delta_indices:
+    - 0
+    - 1
+    - 2
+    - 3
+    - 4
+    - 5
+    - 6
+    - 7
+    - 8
+    - 9
+    - 10
+    - 11
+    - 12
+    - 13
+    - 14
+    - 15
+    - 16
+    - 17
+    - 18
+    - 19
+    - 20
+    - 21
+    - 22
+    - 23
+    modality_keys:
+    - action.left_joint_pos
+    - action.left_gripper_pos
+    - action.right_joint_pos
+    - action.right_gripper_pos
+  language:
+    _target_: groot.vla.data.dataset.ModalityConfig
+    delta_indices:
+    - 0
+    modality_keys:
+    - annotation.task
+transform_yam:
+  _target_: groot.vla.data.transform.ComposedModalityTransform
+  transforms:
+  - _target_: groot.vla.data.transform.VideoToTensor
+    apply_to:
+    - video.top_camera-images-rgb
+    - video.left_camera-images-rgb
+    - video.right_camera-images-rgb
+  - _target_: groot.vla.data.transform.VideoCrop
+    apply_to:
+    - video.top_camera-images-rgb
+    - video.left_camera-images-rgb
+    - video.right_camera-images-rgb
+    scale: 0.95
+    mode: random
+  - _target_: groot.vla.data.transform.VideoResize
+    apply_to:
+    - video.top_camera-images-rgb
+    - video.left_camera-images-rgb
+    - video.right_camera-images-rgb
+    height: 176
+    width: 320
+    interpolation: linear
+  - _target_: groot.vla.data.transform.VideoColorJitter
+    apply_to:
+    - video.top_camera-images-rgb
+    - video.left_camera-images-rgb
+    - video.right_camera-images-rgb
+    brightness: 0.3
+    contrast: 0.4
+    saturation: 0.5
+    hue: 0.08
+  - _target_: groot.vla.data.transform.VideoToNumpy
+    apply_to:
+    - video.top_camera-images-rgb
+    - video.left_camera-images-rgb
+    - video.right_camera-images-rgb
+  - _target_: groot.vla.data.transform.StateActionToTensor
+    apply_to:
+    - state.left_joint_pos
+    - state.left_gripper_pos
+    - state.right_joint_pos
+    - state.right_gripper_pos
+  - _target_: groot.vla.data.transform.StateActionTransform
+    apply_to:
+    - state.left_joint_pos
+    - state.left_gripper_pos
+    - state.right_joint_pos
+    - state.right_gripper_pos
+    normalization_modes:
+      state.left_joint_pos: q99
+      state.left_gripper_pos: q99
+      state.right_joint_pos: q99
+      state.right_gripper_pos: q99
+  - _target_: groot.vla.data.transform.StateActionToTensor
+    apply_to:
+    - action.left_joint_pos
+    - action.left_gripper_pos
+    - action.right_joint_pos
+    - action.right_gripper_pos
+  - _target_: groot.vla.data.transform.StateActionTransform
+    apply_to:
+    - action.left_joint_pos
+    - action.left_gripper_pos
+    - action.right_joint_pos
+    - action.right_gripper_pos
+    normalization_modes:
+      action.left_joint_pos: q99
+      action.left_gripper_pos: q99
+      action.right_joint_pos: q99
+      action.right_gripper_pos: q99
+  - _target_: groot.vla.data.transform.ConcatTransform
+    video_concat_order:
+    - video.top_camera-images-rgb
+    - video.left_camera-images-rgb
+    - video.right_camera-images-rgb
+    state_concat_order:
+    - state.left_joint_pos
+    - state.left_gripper_pos
+    - state.right_joint_pos
+    - state.right_gripper_pos
+    action_concat_order:
+    - action.left_joint_pos
+    - action.left_gripper_pos
+    - action.right_joint_pos
+    - action.right_gripper_pos
+  - _target_: groot.vla.model.dreamzero.transform.dreamzero_cotrain.DreamTransform
+    default_instruction: Perform the default behavior.
+    language_dropout_prob: 0.0
+    always_use_default_instruction: false
+    max_state_dim: 64
+    max_action_dim: 36
+    max_length: 512
+    state_horizon: 1
+    action_horizon: 24
+    embodiment_tag_mapping:
+      real_gr1_arms_only: 0
+      real_gr1_arms_only_annotated: 1
+      real_gr1_arms_waist: 2
+      real_gr1_arms_waist_annotated: 3
+      dexmg_gr1_arms_only_inspire: 4
+      dexmg_gr1_arms_only_fourier: 5
+      dexmg_gr1_arms_waist_fourier: 6
+      robocasa_single_arm: 7
+      onex_eve_gripper: 8
+      robocasa_gr1_arms_only_inspire_hands: 9
+      robocasa_gr1_arms_only_fourier_hands: 10
+      robocasa_gr1_fixed_lower_body_inspire_hands: 11
+      robocasa_gr1_fixed_lower_body_fourier_hands: 12
+      robocasa_panda_omron: 13
+      robocasa_bimanual_panda_parallel_gripper: 15
+      robocasa_bimanual_panda_inspire_hand: 16
+      oxe_droid: 17
+      oxe_fractal: 18
+      oxe_language_table: 19
+      oxe_bridge: 20
+      real_panda_single_arm: 21
+      hot3d_hands_only: 23
+      gr1_unified: 24
+      robocasa_gr1_arms_waist_fourier_hands: 25
+      agibot: 26
+      lapa: 27
+      oxe_mutex: 28
+      oxe_roboset: 29
+      oxe_plex: 30
+      dream: 31
+      yam: 32
+      xdof: 22
+      gr1_unified_segmentation: 14
+      language_table_sim: 7
+      gr1_isaac: 0
+      sim_behavior_r1_pro: 31
+      mecka_hands: 27
+      real_r1_pro_sharpa: 28
+      real_teleop_g1: 33
+    tokenizer_path: /hfm/boqian/liboqian_data/checkpoints/umt5-xxl
+modality_config_real_teleop_g1:
+  video:
+    _target_: groot.vla.data.dataset.ModalityConfig
+    delta_indices:
+    - 0
+    - 1
+    - 2
+    - 3
+    - 4
+    - 5
+    - 6
+    - 7
+    - 8
+    - 9
+    - 10
+    - 11
+    - 12
+    - 13
+    - 14
+    - 15
+    - 16
+    - 17
+    - 18
+    - 19
+    - 20
+    - 21
+    - 22
+    - 23
+    - 24
+    eval_delta_indices:
+    - 0
+    modality_keys:
+    - video.egocentric
+  state:
+    _target_: groot.vla.data.dataset.ModalityConfig
+    delta_indices:
+    - 0
+    modality_keys:
+    - state.left_hand
+    - state.right_hand
+    - state.left_arm
+    - state.right_arm
+    - state.rpy
+    - state.height
+  action:
+    _target_: groot.vla.data.dataset.ModalityConfig
+    delta_indices:
+    - 0
+    - 1
+    - 2
+    - 3
+    - 4
+    - 5
+    - 6
+    - 7
+    - 8
+    - 9
+    - 10
+    - 11
+    - 12
+    - 13
+    - 14
+    - 15
+    - 16
+    - 17
+    - 18
+    - 19
+    - 20
+    - 21
+    - 22
+    - 23
+    modality_keys:
+    - action.left_hand
+    - action.right_hand
+    - action.left_arm
+    - action.right_arm
+    - action.rpy
+    - action.height
+    - action.torso_vx
+    - action.torso_vy
+    - action.torso_vyaw
+    - action.torso_dyaw
+  language:
+    _target_: groot.vla.data.dataset.ModalityConfig
+    delta_indices:
+    - 0
+    modality_keys:
+    - annotation.language.language_instruction
+transform_real_teleop_g1:
+  _target_: groot.vla.data.transform.ComposedModalityTransform
+  transforms:
+  - _target_: groot.vla.data.transform.VideoToTensor
+    apply_to:
+    - video.egocentric
+  - _target_: groot.vla.data.transform.VideoCrop
+    apply_to:
+    - video.egocentric
+    scale: 0.95
+    mode: random
+  - _target_: groot.vla.data.transform.VideoResize
+    apply_to:
+    - video.egocentric
+    height: 176
+    width: 320
+    interpolation: linear
+  - _target_: groot.vla.data.transform.VideoColorJitter
+    apply_to:
+    - video.egocentric
+    brightness: 0.3
+    contrast: 0.4
+    saturation: 0.5
+    hue: 0.08
+  - _target_: groot.vla.data.transform.VideoToNumpy
+    apply_to:
+    - video.egocentric
+  - _target_: groot.vla.data.transform.StateActionToTensor
+    apply_to:
+    - state.left_hand
+    - state.right_hand
+    - state.left_arm
+    - state.right_arm
+    - state.rpy
+    - state.height
+  - _target_: groot.vla.data.transform.StateActionTransform
+    apply_to:
+    - state.left_hand
+    - state.right_hand
+    - state.left_arm
+    - state.right_arm
+    - state.rpy
+    - state.height
+    normalization_modes:
+      state.left_hand: q99
+      state.right_hand: q99
+      state.left_arm: q99
+      state.right_arm: q99
+      state.rpy: q99
+      state.height: q99
+  - _target_: groot.vla.data.transform.StateActionToTensor
+    apply_to:
+    - action.left_hand
+    - action.right_hand
+    - action.left_arm
+    - action.right_arm
+    - action.rpy
+    - action.height
+    - action.torso_vx
+    - action.torso_vy
+    - action.torso_vyaw
+    - action.torso_dyaw
+  - _target_: groot.vla.data.transform.StateActionTransform
+    apply_to:
+    - action.left_hand
+    - action.right_hand
+    - action.left_arm
+    - action.right_arm
+    - action.rpy
+    - action.height
+    - action.torso_vx
+    - action.torso_vy
+    - action.torso_vyaw
+    - action.torso_dyaw
+    normalization_modes:
+      action.left_hand: q99
+      action.right_hand: q99
+      action.left_arm: q99
+      action.right_arm: q99
+      action.rpy: q99
+      action.height: q99
+      action.torso_vx: q99
+      action.torso_vy: q99
+      action.torso_vyaw: q99
+      action.torso_dyaw: q99
+  - _target_: groot.vla.data.transform.ConcatTransform
+    video_concat_order:
+    - video.egocentric
+    state_concat_order:
+    - state.left_hand
+    - state.right_hand
+    - state.left_arm
+    - state.right_arm
+    - state.rpy
+    - state.height
+    action_concat_order:
+    - action.left_hand
+    - action.right_hand
+    - action.left_arm
+    - action.right_arm
+    - action.rpy
+    - action.height
+    - action.torso_vx
+    - action.torso_vy
+    - action.torso_vyaw
+    - action.torso_dyaw
+  - _target_: groot.vla.model.dreamzero.transform.dreamzero_cotrain.DreamTransform
+    default_instruction: Perform the default behavior.
+    language_dropout_prob: 0.0
+    always_use_default_instruction: false
+    max_state_dim: 64
+    max_action_dim: 36
+    max_length: 512
+    state_horizon: 1
+    action_horizon: 24
+    embodiment_tag_mapping:
+      real_gr1_arms_only: 0
+      real_gr1_arms_only_annotated: 1
+      real_gr1_arms_waist: 2
+      real_gr1_arms_waist_annotated: 3
+      dexmg_gr1_arms_only_inspire: 4
+      dexmg_gr1_arms_only_fourier: 5
+      dexmg_gr1_arms_waist_fourier: 6
+      robocasa_single_arm: 7
+      onex_eve_gripper: 8
+      robocasa_gr1_arms_only_inspire_hands: 9
+      robocasa_gr1_arms_only_fourier_hands: 10
+      robocasa_gr1_fixed_lower_body_inspire_hands: 11
+      robocasa_gr1_fixed_lower_body_fourier_hands: 12
+      robocasa_panda_omron: 13
+      robocasa_bimanual_panda_parallel_gripper: 15
+      robocasa_bimanual_panda_inspire_hand: 16
+      oxe_droid: 17
+      oxe_fractal: 18
+      oxe_language_table: 19
+      oxe_bridge: 20
+      real_panda_single_arm: 21
+      hot3d_hands_only: 23
+      gr1_unified: 24
+      robocasa_gr1_arms_waist_fourier_hands: 25
+      agibot: 26
+      lapa: 27
+      oxe_mutex: 28
+      oxe_roboset: 29
+      oxe_plex: 30
+      dream: 31
+      yam: 32
+      xdof: 22
+      gr1_unified_segmentation: 14
+      language_table_sim: 7
+      gr1_isaac: 0
+      sim_behavior_r1_pro: 31
+      mecka_hands: 27
+      real_r1_pro_sharpa: 28
+      real_teleop_g1: 33
+    tokenizer_path: /hfm/boqian/liboqian_data/checkpoints/umt5-xxl
+modality_configs:
+  oxe_droid:
+    video:
+      _target_: groot.vla.data.dataset.ModalityConfig
+      delta_indices:
+      - 0
+      - 1
+      - 2
+      - 3
+      - 4
+      - 5
+      - 6
+      - 7
+      - 8
+      - 9
+      - 10
+      - 11
+      - 12
+      - 13
+      - 14
+      - 15
+      - 16
+      - 17
+      - 18
+      - 19
+      - 20
+      - 21
+      - 22
+      - 23
+      - 24
+      eval_delta_indices:
+      - 0
+      modality_keys:
+      - video.exterior_image_1_left
+      - video.exterior_image_2_left
+      - video.wrist_image_left
+    state:
+      _target_: groot.vla.data.dataset.ModalityConfig
+      delta_indices:
+      - 0
+      modality_keys:
+      - state.joint_position
+      - state.gripper_position
+    action:
+      _target_: groot.vla.data.dataset.ModalityConfig
+      delta_indices:
+      - 0
+      - 1
+      - 2
+      - 3
+      - 4
+      - 5
+      - 6
+      - 7
+      - 8
+      - 9
+      - 10
+      - 11
+      - 12
+      - 13
+      - 14
+      - 15
+      - 16
+      - 17
+      - 18
+      - 19
+      - 20
+      - 21
+      - 22
+      - 23
+      modality_keys:
+      - action.joint_position
+      - action.gripper_position
+    language:
+      _target_: groot.vla.data.dataset.ModalityConfig
+      delta_indices:
+      - 0
+      modality_keys:
+      - annotation.language.language_instruction
+      - annotation.language.language_instruction_2
+      - annotation.language.language_instruction_3
+    lapa_action:
+      _target_: groot.vla.data.dataset.ModalityConfig
+      delta_indices:
+      - 0
+      modality_keys:
+      - lapa_action
+  agibot:
+    video:
+      _target_: groot.vla.data.dataset.ModalityConfig
+      delta_indices:
+      - 0
+      - 1
+      - 2
+      - 3
+      - 4
+      - 5
+      - 6
+      - 7
+      - 8
+      - 9
+      - 10
+      - 11
+      - 12
+      - 13
+      - 14
+      - 15
+      - 16
+      - 17
+      - 18
+      - 19
+      - 20
+      - 21
+      - 22
+      - 23
+      - 24
+      eval_delta_indices:
+      - -3
+      - -2
+      - -1
+      - 0
+      modality_keys:
+      - video.top_head
+      - video.hand_left
+      - video.hand_right
+    state:
+      _target_: groot.vla.data.dataset.ModalityConfig
+      delta_indices:
+      - 0
+      modality_keys:
+      - state.left_arm_joint_position
+      - state.right_arm_joint_position
+      - state.left_effector_position
+      - state.right_effector_position
+      - state.head_position
+      - state.waist_position
+    action:
+      _target_: groot.vla.data.dataset.ModalityConfig
+      delta_indices:
+      - 0
+      - 1
+      - 2
+      - 3
+      - 4
+      - 5
+      - 6
+      - 7
+      - 8
+      - 9
+      - 10
+      - 11
+      - 12
+      - 13
+      - 14
+      - 15
+      - 16
+      - 17
+      - 18
+      - 19
+      - 20
+      - 21
+      - 22
+      - 23
+      modality_keys:
+      - action.left_arm_joint_position
+      - action.right_arm_joint_position
+      - action.left_effector_position
+      - action.right_effector_position
+      - action.head_position
+      - action.waist_position
+      - action.robot_velocity
+    language:
+      _target_: groot.vla.data.dataset.ModalityConfig
+      delta_indices:
+      - 0
+      modality_keys:
+      - annotation.language.action_text
+  yam:
+    video:
+      _target_: groot.vla.data.dataset.ModalityConfig
+      delta_indices:
+      - 0
+      - 1
+      - 2
+      - 3
+      - 4
+      - 5
+      - 6
+      - 7
+      - 8
+      - 9
+      - 10
+      - 11
+      - 12
+      - 13
+      - 14
+      - 15
+      - 16
+      - 17
+      - 18
+      - 19
+      - 20
+      - 21
+      - 22
+      - 23
+      - 24
+      eval_delta_indices:
+      - 0
+      modality_keys:
+      - video.top_camera-images-rgb
+      - video.left_camera-images-rgb
+      - video.right_camera-images-rgb
+    state:
+      _target_: groot.vla.data.dataset.ModalityConfig
+      delta_indices:
+      - 0
+      modality_keys:
+      - state.left_joint_pos
+      - state.left_gripper_pos
+      - state.right_joint_pos
+      - state.right_gripper_pos
+    action:
+      _target_: groot.vla.data.dataset.ModalityConfig
+      delta_indices:
+      - 0
+      - 1
+      - 2
+      - 3
+      - 4
+      - 5
+      - 6
+      - 7
+      - 8
+      - 9
+      - 10
+      - 11
+      - 12
+      - 13
+      - 14
+      - 15
+      - 16
+      - 17
+      - 18
+      - 19
+      - 20
+      - 21
+      - 22
+      - 23
+      modality_keys:
+      - action.left_joint_pos
+      - action.left_gripper_pos
+      - action.right_joint_pos
+      - action.right_gripper_pos
+    language:
+      _target_: groot.vla.data.dataset.ModalityConfig
+      delta_indices:
+      - 0
+      modality_keys:
+      - annotation.task
+  real_teleop_g1:
+    video:
+      _target_: groot.vla.data.dataset.ModalityConfig
+      delta_indices:
+      - 0
+      - 1
+      - 2
+      - 3
+      - 4
+      - 5
+      - 6
+      - 7
+      - 8
+      - 9
+      - 10
+      - 11
+      - 12
+      - 13
+      - 14
+      - 15
+      - 16
+      - 17
+      - 18
+      - 19
+      - 20
+      - 21
+      - 22
+      - 23
+      - 24
+      eval_delta_indices:
+      - 0
+      modality_keys:
+      - video.egocentric
+    state:
+      _target_: groot.vla.data.dataset.ModalityConfig
+      delta_indices:
+      - 0
+      modality_keys:
+      - state.left_hand
+      - state.right_hand
+      - state.left_arm
+      - state.right_arm
+      - state.rpy
+      - state.height
+    action:
+      _target_: groot.vla.data.dataset.ModalityConfig
+      delta_indices:
+      - 0
+      - 1
+      - 2
+      - 3
+      - 4
+      - 5
+      - 6
+      - 7
+      - 8
+      - 9
+      - 10
+      - 11
+      - 12
+      - 13
+      - 14
+      - 15
+      - 16
+      - 17
+      - 18
+      - 19
+      - 20
+      - 21
+      - 22
+      - 23
+      modality_keys:
+      - action.left_hand
+      - action.right_hand
+      - action.left_arm
+      - action.right_arm
+      - action.rpy
+      - action.height
+      - action.torso_vx
+      - action.torso_vy
+      - action.torso_vyaw
+      - action.torso_dyaw
+    language:
+      _target_: groot.vla.data.dataset.ModalityConfig
+      delta_indices:
+      - 0
+      modality_keys:
+      - annotation.language.language_instruction
+transforms:
+  oxe_droid:
+    _target_: groot.vla.data.transform.ComposedModalityTransform
+    transforms:
+    - _target_: groot.vla.data.transform.VideoToTensor
+      apply_to:
+      - video.exterior_image_1_left
+      - video.exterior_image_2_left
+      - video.wrist_image_left
+    - _target_: groot.vla.data.transform.VideoCrop
+      apply_to:
+      - video.exterior_image_1_left
+      - video.exterior_image_2_left
+      - video.wrist_image_left
+      scale: 0.95
+      mode: random
+    - _target_: groot.vla.data.transform.VideoResize
+      apply_to:
+      - video.exterior_image_1_left
+      - video.exterior_image_2_left
+      - video.wrist_image_left
+      height: 176
+      width: 320
+      interpolation: linear
+    - _target_: groot.vla.data.transform.VideoColorJitter
+      apply_to:
+      - video.exterior_image_1_left
+      - video.exterior_image_2_left
+      - video.wrist_image_left
+      brightness: 0.3
+      contrast: 0.4
+      saturation: 0.5
+      hue: 0.08
+    - _target_: groot.vla.data.transform.VideoToNumpy
+      apply_to:
+      - video.exterior_image_1_left
+      - video.exterior_image_2_left
+      - video.wrist_image_left
+    - _target_: groot.vla.data.transform.StateActionToTensor
+      apply_to:
+      - state.joint_position
+      - state.gripper_position
+    - _target_: groot.vla.data.transform.StateActionTransform
+      apply_to:
+      - state.joint_position
+      - state.gripper_position
+      normalization_modes:
+        state.joint_position: q99
+        state.gripper_position: q99
+    - _target_: groot.vla.data.transform.StateActionToTensor
+      apply_to:
+      - action.joint_position
+      - action.gripper_position
+    - _target_: groot.vla.data.transform.StateActionTransform
+      apply_to:
+      - action.joint_position
+      - action.gripper_position
+      normalization_modes:
+        action.joint_position: q99
+        action.gripper_position: q99
+    - _target_: groot.vla.data.transform.ConcatTransform
+      video_concat_order:
+      - video.exterior_image_1_left
+      - video.exterior_image_2_left
+      - video.wrist_image_left
+      state_concat_order:
+      - state.joint_position
+      - state.gripper_position
+      action_concat_order:
+      - action.joint_position
+      - action.gripper_position
+    - _target_: groot.vla.model.dreamzero.transform.dreamzero_cotrain.DreamTransform
+      default_instruction: Perform the default behavior.
+      language_dropout_prob: 0.0
+      always_use_default_instruction: false
+      max_state_dim: 64
+      max_action_dim: 36
+      max_length: 512
+      state_horizon: 1
+      action_horizon: 24
+      embodiment_tag_mapping:
+        real_gr1_arms_only: 0
+        real_gr1_arms_only_annotated: 1
+        real_gr1_arms_waist: 2
+        real_gr1_arms_waist_annotated: 3
+        dexmg_gr1_arms_only_inspire: 4
+        dexmg_gr1_arms_only_fourier: 5
+        dexmg_gr1_arms_waist_fourier: 6
+        robocasa_single_arm: 7
+        onex_eve_gripper: 8
+        robocasa_gr1_arms_only_inspire_hands: 9
+        robocasa_gr1_arms_only_fourier_hands: 10
+        robocasa_gr1_fixed_lower_body_inspire_hands: 11
+        robocasa_gr1_fixed_lower_body_fourier_hands: 12
+        robocasa_panda_omron: 13
+        robocasa_bimanual_panda_parallel_gripper: 15
+        robocasa_bimanual_panda_inspire_hand: 16
+        oxe_droid: 17
+        oxe_fractal: 18
+        oxe_language_table: 19
+        oxe_bridge: 20
+        real_panda_single_arm: 21
+        hot3d_hands_only: 23
+        gr1_unified: 24
+        robocasa_gr1_arms_waist_fourier_hands: 25
+        agibot: 26
+        lapa: 27
+        oxe_mutex: 28
+        oxe_roboset: 29
+        oxe_plex: 30
+        dream: 31
+        yam: 32
+        xdof: 22
+        gr1_unified_segmentation: 14
+        language_table_sim: 7
+        gr1_isaac: 0
+        sim_behavior_r1_pro: 31
+        mecka_hands: 27
+        real_r1_pro_sharpa: 28
+        real_teleop_g1: 33
+      tokenizer_path: /hfm/boqian/liboqian_data/checkpoints/umt5-xxl
+  agibot:
+    _target_: groot.vla.data.transform.ComposedModalityTransform
+    transforms:
+    - _target_: groot.vla.data.transform.VideoToTensor
+      apply_to:
+      - video.top_head
+      - video.hand_left
+      - video.hand_right
+    - _target_: groot.vla.data.transform.VideoCrop
+      apply_to:
+      - video.top_head
+      - video.hand_left
+      - video.hand_right
+      scale: 0.95
+      mode: random
+    - _target_: groot.vla.data.transform.VideoResize
+      apply_to:
+      - video.top_head
+      - video.hand_left
+      - video.hand_right
+      height: 176
+      width: 320
+      interpolation: linear
+    - _target_: groot.vla.data.transform.VideoColorJitter
+      apply_to:
+      - video.top_head
+      - video.hand_left
+      - video.hand_right
+      brightness: 0.3
+      contrast: 0.4
+      saturation: 0.5
+      hue: 0.08
+    - _target_: groot.vla.data.transform.VideoToNumpy
+      apply_to:
+      - video.top_head
+      - video.hand_left
+      - video.hand_right
+    - _target_: groot.vla.data.transform.StateActionToTensor
+      apply_to:
+      - state.left_arm_joint_position
+      - state.right_arm_joint_position
+      - state.left_effector_position
+      - state.right_effector_position
+      - state.head_position
+      - state.waist_position
+    - _target_: groot.vla.data.transform.StateActionTransform
+      apply_to:
+      - state.left_arm_joint_position
+      - state.right_arm_joint_position
+      - state.left_effector_position
+      - state.right_effector_position
+      - state.head_position
+      - state.waist_position
+      normalization_modes:
+        state.left_arm_joint_position: q99
+        state.right_arm_joint_position: q99
+        state.left_effector_position: q99
+        state.right_effector_position: q99
+        state.head_position: q99
+        state.waist_position: q99
+    - _target_: groot.vla.data.transform.StateActionToTensor
+      apply_to:
+      - action.left_arm_joint_position
+      - action.right_arm_joint_position
+      - action.left_effector_position
+      - action.right_effector_position
+      - action.head_position
+      - action.waist_position
+      - action.robot_velocity
+    - _target_: groot.vla.data.transform.StateActionTransform
+      apply_to:
+      - action.left_arm_joint_position
+      - action.right_arm_joint_position
+      - action.left_effector_position
+      - action.right_effector_position
+      - action.head_position
+      - action.waist_position
+      - action.robot_velocity
+      normalization_modes:
+        action.left_arm_joint_position: q99
+        action.right_arm_joint_position: q99
+        action.left_effector_position: q99
+        action.right_effector_position: q99
+        action.head_position: q99
+        action.waist_position: q99
+        action.robot_velocity: q99
+    - _target_: groot.vla.data.transform.ConcatTransform
+      video_concat_order:
+      - video.top_head
+      - video.hand_left
+      - video.hand_right
+      state_concat_order:
+      - state.left_arm_joint_position
+      - state.right_arm_joint_position
+      - state.left_effector_position
+      - state.right_effector_position
+      - state.head_position
+      - state.waist_position
+      action_concat_order:
+      - action.left_arm_joint_position
+      - action.right_arm_joint_position
+      - action.left_effector_position
+      - action.right_effector_position
+      - action.head_position
+      - action.waist_position
+      - action.robot_velocity
+    - _target_: groot.vla.model.dreamzero.transform.dreamzero_cotrain.DreamTransform
+      default_instruction: Perform the default behavior.
+      language_dropout_prob: 0.0
+      always_use_default_instruction: false
+      max_state_dim: 64
+      max_action_dim: 36
+      max_length: 512
+      state_horizon: 1
+      action_horizon: 24
+      embodiment_tag_mapping:
+        real_gr1_arms_only: 0
+        real_gr1_arms_only_annotated: 1
+        real_gr1_arms_waist: 2
+        real_gr1_arms_waist_annotated: 3
+        dexmg_gr1_arms_only_inspire: 4
+        dexmg_gr1_arms_only_fourier: 5
+        dexmg_gr1_arms_waist_fourier: 6
+        robocasa_single_arm: 7
+        onex_eve_gripper: 8
+        robocasa_gr1_arms_only_inspire_hands: 9
+        robocasa_gr1_arms_only_fourier_hands: 10
+        robocasa_gr1_fixed_lower_body_inspire_hands: 11
+        robocasa_gr1_fixed_lower_body_fourier_hands: 12
+        robocasa_panda_omron: 13
+        robocasa_bimanual_panda_parallel_gripper: 15
+        robocasa_bimanual_panda_inspire_hand: 16
+        oxe_droid: 17
+        oxe_fractal: 18
+        oxe_language_table: 19
+        oxe_bridge: 20
+        real_panda_single_arm: 21
+        hot3d_hands_only: 23
+        gr1_unified: 24
+        robocasa_gr1_arms_waist_fourier_hands: 25
+        agibot: 26
+        lapa: 27
+        oxe_mutex: 28
+        oxe_roboset: 29
+        oxe_plex: 30
+        dream: 31
+        yam: 32
+        xdof: 22
+        gr1_unified_segmentation: 14
+        language_table_sim: 7
+        gr1_isaac: 0
+        sim_behavior_r1_pro: 31
+        mecka_hands: 27
+        real_r1_pro_sharpa: 28
+        real_teleop_g1: 33
+      tokenizer_path: /hfm/boqian/liboqian_data/checkpoints/umt5-xxl
+  yam:
+    _target_: groot.vla.data.transform.ComposedModalityTransform
+    transforms:
+    - _target_: groot.vla.data.transform.VideoToTensor
+      apply_to:
+      - video.top_camera-images-rgb
+      - video.left_camera-images-rgb
+      - video.right_camera-images-rgb
+    - _target_: groot.vla.data.transform.VideoCrop
+      apply_to:
+      - video.top_camera-images-rgb
+      - video.left_camera-images-rgb
+      - video.right_camera-images-rgb
+      scale: 0.95
+      mode: random
+    - _target_: groot.vla.data.transform.VideoResize
+      apply_to:
+      - video.top_camera-images-rgb
+      - video.left_camera-images-rgb
+      - video.right_camera-images-rgb
+      height: 176
+      width: 320
+      interpolation: linear
+    - _target_: groot.vla.data.transform.VideoColorJitter
+      apply_to:
+      - video.top_camera-images-rgb
+      - video.left_camera-images-rgb
+      - video.right_camera-images-rgb
+      brightness: 0.3
+      contrast: 0.4
+      saturation: 0.5
+      hue: 0.08
+    - _target_: groot.vla.data.transform.VideoToNumpy
+      apply_to:
+      - video.top_camera-images-rgb
+      - video.left_camera-images-rgb
+      - video.right_camera-images-rgb
+    - _target_: groot.vla.data.transform.StateActionToTensor
+      apply_to:
+      - state.left_joint_pos
+      - state.left_gripper_pos
+      - state.right_joint_pos
+      - state.right_gripper_pos
+    - _target_: groot.vla.data.transform.StateActionTransform
+      apply_to:
+      - state.left_joint_pos
+      - state.left_gripper_pos
+      - state.right_joint_pos
+      - state.right_gripper_pos
+      normalization_modes:
+        state.left_joint_pos: q99
+        state.left_gripper_pos: q99
+        state.right_joint_pos: q99
+        state.right_gripper_pos: q99
+    - _target_: groot.vla.data.transform.StateActionToTensor
+      apply_to:
+      - action.left_joint_pos
+      - action.left_gripper_pos
+      - action.right_joint_pos
+      - action.right_gripper_pos
+    - _target_: groot.vla.data.transform.StateActionTransform
+      apply_to:
+      - action.left_joint_pos
+      - action.left_gripper_pos
+      - action.right_joint_pos
+      - action.right_gripper_pos
+      normalization_modes:
+        action.left_joint_pos: q99
+        action.left_gripper_pos: q99
+        action.right_joint_pos: q99
+        action.right_gripper_pos: q99
+    - _target_: groot.vla.data.transform.ConcatTransform
+      video_concat_order:
+      - video.top_camera-images-rgb
+      - video.left_camera-images-rgb
+      - video.right_camera-images-rgb
+      state_concat_order:
+      - state.left_joint_pos
+      - state.left_gripper_pos
+      - state.right_joint_pos
+      - state.right_gripper_pos
+      action_concat_order:
+      - action.left_joint_pos
+      - action.left_gripper_pos
+      - action.right_joint_pos
+      - action.right_gripper_pos
+    - _target_: groot.vla.model.dreamzero.transform.dreamzero_cotrain.DreamTransform
+      default_instruction: Perform the default behavior.
+      language_dropout_prob: 0.0
+      always_use_default_instruction: false
+      max_state_dim: 64
+      max_action_dim: 36
+      max_length: 512
+      state_horizon: 1
+      action_horizon: 24
+      embodiment_tag_mapping:
+        real_gr1_arms_only: 0
+        real_gr1_arms_only_annotated: 1
+        real_gr1_arms_waist: 2
+        real_gr1_arms_waist_annotated: 3
+        dexmg_gr1_arms_only_inspire: 4
+        dexmg_gr1_arms_only_fourier: 5
+        dexmg_gr1_arms_waist_fourier: 6
+        robocasa_single_arm: 7
+        onex_eve_gripper: 8
+        robocasa_gr1_arms_only_inspire_hands: 9
+        robocasa_gr1_arms_only_fourier_hands: 10
+        robocasa_gr1_fixed_lower_body_inspire_hands: 11
+        robocasa_gr1_fixed_lower_body_fourier_hands: 12
+        robocasa_panda_omron: 13
+        robocasa_bimanual_panda_parallel_gripper: 15
+        robocasa_bimanual_panda_inspire_hand: 16
+        oxe_droid: 17
+        oxe_fractal: 18
+        oxe_language_table: 19
+        oxe_bridge: 20
+        real_panda_single_arm: 21
+        hot3d_hands_only: 23
+        gr1_unified: 24
+        robocasa_gr1_arms_waist_fourier_hands: 25
+        agibot: 26
+        lapa: 27
+        oxe_mutex: 28
+        oxe_roboset: 29
+        oxe_plex: 30
+        dream: 31
+        yam: 32
+        xdof: 22
+        gr1_unified_segmentation: 14
+        language_table_sim: 7
+        gr1_isaac: 0
+        sim_behavior_r1_pro: 31
+        mecka_hands: 27
+        real_r1_pro_sharpa: 28
+        real_teleop_g1: 33
+      tokenizer_path: /hfm/boqian/liboqian_data/checkpoints/umt5-xxl
+  real_teleop_g1:
+    _target_: groot.vla.data.transform.ComposedModalityTransform
+    transforms:
+    - _target_: groot.vla.data.transform.VideoToTensor
+      apply_to:
+      - video.egocentric
+    - _target_: groot.vla.data.transform.VideoCrop
+      apply_to:
+      - video.egocentric
+      scale: 0.95
+      mode: random
+    - _target_: groot.vla.data.transform.VideoResize
+      apply_to:
+      - video.egocentric
+      height: 176
+      width: 320
+      interpolation: linear
+    - _target_: groot.vla.data.transform.VideoColorJitter
+      apply_to:
+      - video.egocentric
+      brightness: 0.3
+      contrast: 0.4
+      saturation: 0.5
+      hue: 0.08
+    - _target_: groot.vla.data.transform.VideoToNumpy
+      apply_to:
+      - video.egocentric
+    - _target_: groot.vla.data.transform.StateActionToTensor
+      apply_to:
+      - state.left_hand
+      - state.right_hand
+      - state.left_arm
+      - state.right_arm
+      - state.rpy
+      - state.height
+    - _target_: groot.vla.data.transform.StateActionTransform
+      apply_to:
+      - state.left_hand
+      - state.right_hand
+      - state.left_arm
+      - state.right_arm
+      - state.rpy
+      - state.height
+      normalization_modes:
+        state.left_hand: q99
+        state.right_hand: q99
+        state.left_arm: q99
+        state.right_arm: q99
+        state.rpy: q99
+        state.height: q99
+    - _target_: groot.vla.data.transform.StateActionToTensor
+      apply_to:
+      - action.left_hand
+      - action.right_hand
+      - action.left_arm
+      - action.right_arm
+      - action.rpy
+      - action.height
+      - action.torso_vx
+      - action.torso_vy
+      - action.torso_vyaw
+      - action.torso_dyaw
+    - _target_: groot.vla.data.transform.StateActionTransform
+      apply_to:
+      - action.left_hand
+      - action.right_hand
+      - action.left_arm
+      - action.right_arm
+      - action.rpy
+      - action.height
+      - action.torso_vx
+      - action.torso_vy
+      - action.torso_vyaw
+      - action.torso_dyaw
+      normalization_modes:
+        action.left_hand: q99
+        action.right_hand: q99
+        action.left_arm: q99
+        action.right_arm: q99
+        action.rpy: q99
+        action.height: q99
+        action.torso_vx: q99
+        action.torso_vy: q99
+        action.torso_vyaw: q99
+        action.torso_dyaw: q99
+    - _target_: groot.vla.data.transform.ConcatTransform
+      video_concat_order:
+      - video.egocentric
+      state_concat_order:
+      - state.left_hand
+      - state.right_hand
+      - state.left_arm
+      - state.right_arm
+      - state.rpy
+      - state.height
+      action_concat_order:
+      - action.left_hand
+      - action.right_hand
+      - action.left_arm
+      - action.right_arm
+      - action.rpy
+      - action.height
+      - action.torso_vx
+      - action.torso_vy
+      - action.torso_vyaw
+      - action.torso_dyaw
+    - _target_: groot.vla.model.dreamzero.transform.dreamzero_cotrain.DreamTransform
+      default_instruction: Perform the default behavior.
+      language_dropout_prob: 0.0
+      always_use_default_instruction: false
+      max_state_dim: 64
+      max_action_dim: 36
+      max_length: 512
+      state_horizon: 1
+      action_horizon: 24
+      embodiment_tag_mapping:
+        real_gr1_arms_only: 0
+        real_gr1_arms_only_annotated: 1
+        real_gr1_arms_waist: 2
+        real_gr1_arms_waist_annotated: 3
+        dexmg_gr1_arms_only_inspire: 4
+        dexmg_gr1_arms_only_fourier: 5
+        dexmg_gr1_arms_waist_fourier: 6
+        robocasa_single_arm: 7
+        onex_eve_gripper: 8
+        robocasa_gr1_arms_only_inspire_hands: 9
+        robocasa_gr1_arms_only_fourier_hands: 10
+        robocasa_gr1_fixed_lower_body_inspire_hands: 11
+        robocasa_gr1_fixed_lower_body_fourier_hands: 12
+        robocasa_panda_omron: 13
+        robocasa_bimanual_panda_parallel_gripper: 15
+        robocasa_bimanual_panda_inspire_hand: 16
+        oxe_droid: 17
+        oxe_fractal: 18
+        oxe_language_table: 19
+        oxe_bridge: 20
+        real_panda_single_arm: 21
+        hot3d_hands_only: 23
+        gr1_unified: 24
+        robocasa_gr1_arms_waist_fourier_hands: 25
+        agibot: 26
+        lapa: 27
+        oxe_mutex: 28
+        oxe_roboset: 29
+        oxe_plex: 30
+        dream: 31
+        yam: 32
+        xdof: 22
+        gr1_unified_segmentation: 14
+        language_table_sim: 7
+        gr1_isaac: 0
+        sim_behavior_r1_pro: 31
+        mecka_hands: 27
+        real_r1_pro_sharpa: 28
+        real_teleop_g1: 33
+      tokenizer_path: /hfm/boqian/liboqian_data/checkpoints/umt5-xxl
+metadata_versions:
+  oxe_droid: '0221'
+  agibot: '0221'
+  yam: '0221'
+  real_teleop_g1: '0221'
+fps:
+  yam: 30
+  real_teleop_g1: 30
+relative_action: false
+relative_action_per_horizon: false
+relative_action_keys: []
+max_chunk_size: 4
+dataset_shard_sampling_rate: 0.1
+mixture_dataset_cls: groot.vla.data.dataset.lerobot_sharded.ShardedLeRobotMixtureDataset.from_mixture_spec
+single_dataset_cls: groot.vla.data.dataset.lerobot_sharded.ShardedLeRobotSubLangSingleActionChunkDatasetDROID
+real_teleop_g1_data_root: /hfm/boqian/liboqian_data/data/real_data/gear/g1/Pick_bottle_and_turn_and_pour_into_cup_absolute
+total_training_steps: 1048576000000

dreamzero_real_teleop_g1_full_finetune_absolute_Pick_bottle_and_turn_and_pour_into_cup/checkpoint-20000/experiment_cfg/metadata.json

@@ -0,0 +1,787 @@
+{
+    "real_teleop_g1": {
+        "statistics": {
+            "state": {
+                "left_hand": {
+                    "max": [
+                        -0.15012885630130768,
+                        0.9005824327468872,
+                        1.240552306175232,
+                        -0.001318894443102181,
+                        -0.029013408347964287,
+                        -0.0026860591024160385,
+                        -0.032535798847675323
+                    ],
+                    "min": [
+                        -1.0365108251571655,
+                        0.3635019063949585,
+                        0.6077943444252014,
+                        -0.8983967900276184,
+                        -1.4610702991485596,
+                        -0.9210805296897888,
+                        -1.2054297924041748
+                    ],
+                    "mean": [
+                        -0.7106521057783574,
+                        0.7397562400548329,
+                        0.7932539759105839,
+                        -0.18307749281402536,
+                        -0.3996594926481087,
+                        -0.22440346846797105,
+                        -0.3648747590553514
+                    ],
+                    "std": [
+                        0.1800224325521231,
+                        0.08892492705684825,
+                        0.17347929747061464,
+                        0.1250548816433564,
+                        0.14322181312600368,
+                        0.12746796614686143,
+                        0.1004095686025728
+                    ],
+                    "q01": [
+                        -1.0010048174858093,
+                        0.5051653385162354,
+                        0.6091835498809814,
+                        -0.5192373794317245,
+                        -0.7626793467998505,
+                        -0.5990579128265381,
+                        -0.6296854817867279
+                    ],
+                    "q99": [
+                        -0.35215023159980774,
+                        0.8828813862800599,
+                        1.1794202327728271,
+                        -0.012645758632570512,
+                        -0.08962718859314919,
+                        -0.01236837636679411,
+                        -0.14306417107582092
+                    ]
+                },
+                "right_hand": {
+                    "max": [
+                        0.013422729447484016,
+                        -0.0029037927743047476,
+                        -0.4307878315448761,
+                        1.1691919565200806,
+                        1.5508151054382324,
+                        1.008196473121643,
+                        1.4388744831085205
+                    ],
+                    "min": [
+                        -0.9901090860366821,
+                        -0.8877645134925842,
+                        -1.2071785926818848,
+                        -0.03135598078370094,
+                        0.03392980992794037,
+                        0.023357456550002098,
+                        0.019666638225317
+                    ],
+                    "mean": [
+                        -0.4129581264587618,
+                        -0.4582520757612821,
+                        -0.664123836057758,
+                        0.4952124991587111,
+                        0.9231042625678995,
+                        0.5022273107153479,
+                        0.669409344487608
+                    ],
+                    "std": [
+                        0.19006874361243062,
+                        0.11976727609652792,
+                        0.10119243113959121,
+                        0.2266161525462106,
+                        0.4655465641238765,
+                        0.21432269096327275,
+                        0.31467715432459836
+                    ],
+                    "q01": [
+                        -0.9081442606449127,
+                        -0.8110105991363525,
+                        -0.9831316405534745,
+                        0.023696813061833383,
+                        0.037188550531864165,
+                        0.024981562048196793,
+                        0.02064503453671932
+                    ],
+                    "q99": [
+                        -0.10628944709897117,
+                        -0.18123969539999962,
+                        -0.5385138392448425,
+                        0.9067972785234452,
+                        1.4886025190353394,
+                        0.871771514415741,
+                        1.3075256443023682
+                    ]
+                },
+                "left_arm": {
+                    "max": [
+                        0.6784030199050903,
+                        0.5408600568771362,
+                        0.261579692363739,
+                        1.2853081226348877,
+                        0.6677011251449585,
+                        0.01986505277454853,
+                        0.5779042840003967
+                    ],
+                    "min": [
+                        -0.6697743535041809,
+                        -0.04027898237109184,
+                        -0.6574546098709106,
+                        -0.02732403390109539,
+                        -0.18146513402462006,
+                        -1.2264176607131958,
+                        -0.8434670567512512
+                    ],
+                    "mean": [
+                        0.06576749867032948,
+                        0.25579782742676094,
+                        -0.06004482964572005,
+                        0.808768983211113,
+                        0.10850686643277818,
+                        -0.7254681426808276,
+                        -0.2878231007376268
+                    ],
+                    "std": [
+                        0.16998997778227107,
+                        0.09886796724754655,
+                        0.1532526690961023,
+                        0.3788245539398352,
+                        0.1304561854786525,
+                        0.3305712410884026,
+                        0.20458272924414123
+                    ],
+                    "q01": [
+                        -0.472993403673172,
+                        0.01923264417797327,
+                        -0.4175487527251244,
+                        0.11523827515542508,
+                        -0.13969017088413238,
+                        -1.190878005027771,
+                        -0.7066819667816162
+                    ],
+                    "q99": [
+                        0.4539884480834007,
+                        0.5193723440170288,
+                        0.2179892195761203,
+                        1.2710349559783936,
+                        0.6194926702976227,
+                        -0.09952061586081994,
+                        0.17473116487264606
+                    ]
+                },
+                "right_arm": {
+                    "max": [
+                        0.9002070426940918,
+                        0.11829628795385361,
+                        0.8515990376472473,
+                        1.2844573259353638,
+                        0.8022719621658325,
+                        0.6399767994880676,
+                        1.2358516454696655
+                    ],
+                    "min": [
+                        -1.2026649713516235,
+                        -0.6969066858291626,
+                        -0.815610408782959,
+                        -0.9874761700630188,
+                        -1.6946533918380737,
+                        -1.4213329553604126,
+                        -1.342279314994812
+                    ],
+                    "mean": [
+                        -0.0662775668920557,
+                        -0.1345353524434497,
+                        0.0634982026036378,
+                        -0.10989526924031702,
+                        -0.05263142642975812,
+                        -0.12096835662218453,
+                        0.120861711803444
+                    ],
+                    "std": [
+                        0.4008494462356859,
+                        0.1700786431413106,
+                        0.21295159275658426,
+                        0.43649950708158436,
+                        0.4497950790039162,
+                        0.36514804111543175,
+                        0.1967684071659634
+                    ],
+                    "q01": [
+                        -0.950262793302536,
+                        -0.5722128254175186,
+                        -0.4087426000833511,
+                        -0.8998615843057632,
+                        -1.200806084871292,
+                        -1.1520812797546387,
+                        -0.3325912955403328
+                    ],
+                    "q99": [
+                        0.7100517773628232,
+                        0.10524750150740145,
+                        0.6194546324014664,
+                        0.9450081789493561,
+                        0.559214276075363,
+                        0.5243198603391647,
+                        0.6973656332492828
+                    ]
+                },
+                "rpy": {
+                    "max": [
+                        0.1827869415283203,
+                        0.12582087516784668,
+                        0.7600905299186707
+                    ],
+                    "min": [
+                        -0.25714462995529175,
+                        -0.27553248405456543,
+                        -0.3924732208251953
+                    ],
+                    "mean": [
+                        -0.07450471914543803,
+                        -0.016358907944713997,
+                        0.08461220301037418
+                    ],
+                    "std": [
+                        0.052686434459038575,
+                        0.05521563296798423,
+                        0.1871730371455983
+                    ],
+                    "q01": [
+                        -0.21652273923158646,
+                        -0.15164137467741967,
+                        -0.3411825066804886
+                    ],
+                    "q99": [
+                        0.023736179061233972,
+                        0.0924126328527927,
+                        0.6843682318925858
+                    ]
+                },
+                "height": {
+                    "max": [
+                        0.75
+                    ],
+                    "min": [
+                        0.6516909599304199
+                    ],
+                    "mean": [
+                        0.717019414027915
+                    ],
+                    "std": [
+                        0.010611909948945071
+                    ],
+                    "q01": [
+                        0.681534378528595
+                    ],
+                    "q99": [
+                        0.7377596861124038
+                    ]
+                }
+            },
+            "action": {
+                "left_hand": {
+                    "max": [
+                        -0.1213064193725586,
+                        0.9210000038146973,
+                        1.2908918857574463,
+                        0.0010000000474974513,
+                        0.0010000000474974513,
+                        0.0010000000474974513,
+                        0.0010000000474974513
+                    ],
+                    "min": [
+                        -1.0481975078582764,
+                        0.33770662546157837,
+                        0.5833444595336914,
+                        -0.9677509069442749,
+                        -1.5101021528244019,
+                        -0.9662452340126038,
+                        -1.252585768699646
+                    ],
+                    "mean": [
+                        -0.7283108163247403,
+                        0.7484512520233001,
+                        0.7898890272276098,
+                        -0.1841689376362202,
+                        -0.40007694289939594,
+                        -0.22100483851739403,
+                        -0.3606009050772844
+                    ],
+                    "std": [
+                        0.18855574743509407,
+                        0.10765775142317323,
+                        0.20562452513338061,
+                        0.13352257457524555,
+                        0.14831147850778914,
+                        0.13177602593521492,
+                        0.10852811415283538
+                    ],
+                    "q01": [
+                        -1.0481975078582764,
+                        0.47301085710525514,
+                        0.5974299025535583,
+                        -0.5396523159742356,
+                        -0.7698984146118164,
+                        -0.569393122792244,
+                        -0.6210935115814209
+                    ],
+                    "q99": [
+                        -0.32839281708002094,
+                        0.9210000038146973,
+                        1.2100253427028655,
+                        0.0010000000474974513,
+                        -0.07790654666721833,
+                        0.0010000000474974513,
+                        -0.12742940008640297
+                    ]
+                },
+                "right_hand": {
+                    "max": [
+                        0.11635252088308334,
+                        0.11013627797365189,
+                        -0.4304533302783966,
+                        1.4093444347381592,
+                        1.7463293075561523,
+                        1.2701730728149414,
+                        1.7463293075561523
+                    ],
+                    "min": [
+                        -1.0481975078582764,
+                        -0.9210000038146973,
+                        -1.4877983331680298,
+                        -0.0010000000474974513,
+                        -0.0010000000474974513,
+                        -0.0010000000474974513,
+                        -0.0010000000474974513
+                    ],
+                    "mean": [
+                        -0.3458960654680007,
+                        -0.6374010453148075,
+                        -0.7414860319631214,
+                        0.6248424045903762,
+                        1.0210826742269492,
+                        0.624888935664034,
+                        0.7309311322729288
+                    ],
+                    "std": [
+                        0.2477478868975401,
+                        0.12149676878810435,
+                        0.14591807616135066,
+                        0.31889999783341705,
+                        0.550414547421302,
+                        0.29655947330065524,
+                        0.37683330327414893
+                    ],
+                    "q01": [
+                        -0.943056748509407,
+                        -0.8472843170166016,
+                        -1.144005218744278,
+                        -0.0010000000474974513,
+                        -0.0010000000474974513,
+                        -0.0010000000474974513,
+                        -0.0010000000474974513
+                    ],
+                    "q99": [
+                        -0.019174456279724885,
+                        -0.30395907163619995,
+                        -0.5490964096784592,
+                        1.2382159757614122,
+                        1.7386069679260252,
+                        1.021970037221905,
+                        1.4857414960861206
+                    ]
+                },
+                "left_arm": {
+                    "max": [
+                        0.7063568830490112,
+                        0.5787747502326965,
+                        0.2658364176750183,
+                        1.36533784866333,
+                        0.7970759272575378,
+                        0.050355639308691025,
+                        0.6250148415565491
+                    ],
+                    "min": [
+                        -0.6973711848258972,
+                        -0.04661950096487999,
+                        -0.7617664933204651,
+                        -0.0605352558195591,
+                        -0.21390148997306824,
+                        -1.2438902854919434,
+                        -0.8552998304367065
+                    ],
+                    "mean": [
+                        0.05742595636422137,
+                        0.2634212185929878,
+                        -0.06515526749555424,
+                        0.8164667552012993,
+                        0.10927570881930437,
+                        -0.7389155860415086,
+                        -0.295913822816809
+                    ],
+                    "std": [
+                        0.16957499887908717,
+                        0.0950404677458485,
+                        0.15050156924776992,
+                        0.38010919103376684,
+                        0.13206946446875728,
+                        0.32367770103457033,
+                        0.20343907571654762
+                    ],
+                    "q01": [
+                        -0.5008356976509094,
+                        0.03911703012883663,
+                        -0.4255762630701065,
+                        0.11195705555379391,
+                        -0.15061241388320923,
+                        -1.1989330887794494,
+                        -0.7123385232686996
+                    ],
+                    "q99": [
+                        0.4477514466643332,
+                        0.524748660326004,
+                        0.216304791867733,
+                        1.2753526496887206,
+                        0.6235151898860929,
+                        -0.11624279774725446,
+                        0.17421353697776795
+                    ]
+                },
+                "right_arm": {
+                    "max": [
+                        1.0034496784210205,
+                        0.6385414600372314,
+                        0.8838359117507935,
+                        1.3011894226074219,
+                        0.9103013873100281,
+                        0.6697143316268921,
+                        1.300451397895813
+                    ],
+                    "min": [
+                        -1.2567458152770996,
+                        -0.7329762578010559,
+                        -0.8994792103767395,
+                        -1.029470682144165,
+                        -1.7312105894088745,
+                        -1.614429235458374,
+                        -1.554446816444397
+                    ],
+                    "mean": [
+                        -0.06658668015508451,
+                        -0.1276185496595779,
+                        0.05797919617399393,
+                        -0.11602684028492932,
+                        -0.03613449704007667,
+                        -0.1405436867128357,
+                        0.1209290054543264
+                    ],
+                    "std": [
+                        0.4080472941147137,
+                        0.19441447416204616,
+                        0.22406206897544698,
+                        0.46053246124217484,
+                        0.4648783663000821,
+                        0.3964305914547449,
+                        0.21165879464009837
+                    ],
+                    "q01": [
+                        -0.9834537935256957,
+                        -0.5832101339101792,
+                        -0.46410051107406614,
+                        -0.9261943989992142,
+                        -1.2178711235523223,
+                        -1.2455947709083557,
+                        -0.374444442987442
+                    ],
+                    "q99": [
+                        0.7256432592868804,
+                        0.30993463188409803,
+                        0.6237642824649811,
+                        1.089427736997604,
+                        0.6462495481967916,
+                        0.5251477748155594,
+                        0.7147807055711739
+                    ]
+                },
+                "rpy": {
+                    "max": [
+                        0.1827869415283203,
+                        0.12582087516784668,
+                        0.7600905299186707
+                    ],
+                    "min": [
+                        -0.25714462995529175,
+                        -0.27553248405456543,
+                        -0.3924732208251953
+                    ],
+                    "mean": [
+                        -0.07457712200498402,
+                        -0.016338004632023968,
+                        0.08467273931155996
+                    ],
+                    "std": [
+                        0.05265008883277812,
+                        0.05524160676308374,
+                        0.18724355327591102
+                    ],
+                    "q01": [
+                        -0.21652273923158646,
+                        -0.15164137467741967,
+                        -0.34133789867162706
+                    ],
+                    "q99": [
+                        0.023736179061233972,
+                        0.09255022160708899,
+                        0.6843682318925858
+                    ]
+                },
+                "height": {
+                    "max": [
+                        0.7456391453742981
+                    ],
+                    "min": [
+                        0.6516909599304199
+                    ],
+                    "mean": [
+                        0.7169688045156508
+                    ],
+                    "std": [
+                        0.01058211087414312
+                    ],
+                    "q01": [
+                        0.6814439028501511
+                    ],
+                    "q99": [
+                        0.7374992072582245
+                    ]
+                },
+                "torso_vx": {
+                    "max": [
+                        0.3499999940395355
+                    ],
+                    "min": [
+                        0.0
+                    ],
+                    "mean": [
+                        0.017269024247052768
+                    ],
+                    "std": [
+                        0.07580197349073346
+                    ],
+                    "q01": [
+                        0.0
+                    ],
+                    "q99": [
+                        0.3499999940395355
+                    ]
+                },
+                "torso_vy": {
+                    "max": [
+                        0.5
+                    ],
+                    "min": [
+                        0.0
+                    ],
+                    "mean": [
+                        0.0002320065992988245
+                    ],
+                    "std": [
+                        0.01076798368252486
+                    ],
+                    "q01": [
+                        0.0
+                    ],
+                    "q99": [
+                        0.0
+                    ]
+                },
+                "torso_vyaw": {
+                    "max": [
+                        0.5
+                    ],
+                    "min": [
+                        -0.5
+                    ],
+                    "mean": [
+                        -0.03733115137363985
+                    ],
+                    "std": [
+                        0.1199053455233051
+                    ],
+                    "q01": [
+                        -0.5
+                    ],
+                    "q99": [
+                        0.0
+                    ]
+                },
+                "torso_dyaw": {
+                    "max": [
+                        0.3072107136249542
+                    ],
+                    "min": [
+                        -0.1875186562538147
+                    ],
+                    "mean": [
+                        0.016342195046893805
+                    ],
+                    "std": [
+                        0.048962870720188865
+                    ],
+                    "q01": [
+                        -0.005849878406152129
+                    ],
+                    "q99": [
+                        0.22258896410465237
+                    ]
+                }
+            }
+        },
+        "modalities": {
+            "video": {
+                "egocentric": {
+                    "resolution": [
+                        640,
+                        480
+                    ],
+                    "channels": 3,
+                    "fps": 30.0
+                }
+            },
+            "state": {
+                "left_hand": {
+                    "absolute": true,
+                    "rotation_type": null,
+                    "shape": [
+                        7
+                    ],
+                    "continuous": true
+                },
+                "right_hand": {
+                    "absolute": true,
+                    "rotation_type": null,
+                    "shape": [
+                        7
+                    ],
+                    "continuous": true
+                },
+                "left_arm": {
+                    "absolute": true,
+                    "rotation_type": null,
+                    "shape": [
+                        7
+                    ],
+                    "continuous": true
+                },
+                "right_arm": {
+                    "absolute": true,
+                    "rotation_type": null,
+                    "shape": [
+                        7
+                    ],
+                    "continuous": true
+                },
+                "rpy": {
+                    "absolute": true,
+                    "rotation_type": null,
+                    "shape": [
+                        3
+                    ],
+                    "continuous": true
+                },
+                "height": {
+                    "absolute": true,
+                    "rotation_type": null,
+                    "shape": [
+                        1
+                    ],
+                    "continuous": true
+                }
+            },
+            "action": {
+                "left_hand": {
+                    "absolute": true,
+                    "rotation_type": null,
+                    "shape": [
+                        7
+                    ],
+                    "continuous": true
+                },
+                "right_hand": {
+                    "absolute": true,
+                    "rotation_type": null,
+                    "shape": [
+                        7
+                    ],
+                    "continuous": true
+                },
+                "left_arm": {
+                    "absolute": true,
+                    "rotation_type": null,
+                    "shape": [
+                        7
+                    ],
+                    "continuous": true
+                },
+                "right_arm": {
+                    "absolute": true,
+                    "rotation_type": null,
+                    "shape": [
+                        7
+                    ],
+                    "continuous": true
+                },
+                "rpy": {
+                    "absolute": true,
+                    "rotation_type": null,
+                    "shape": [
+                        3
+                    ],
+                    "continuous": true
+                },
+                "height": {
+                    "absolute": true,
+                    "rotation_type": null,
+                    "shape": [
+                        1
+                    ],
+                    "continuous": true
+                },
+                "torso_vx": {
+                    "absolute": false,
+                    "rotation_type": null,
+                    "shape": [
+                        1
+                    ],
+                    "continuous": true
+                },
+                "torso_vy": {
+                    "absolute": false,
+                    "rotation_type": null,
+                    "shape": [
+                        1
+                    ],
+                    "continuous": true
+                },
+                "torso_vyaw": {
+                    "absolute": false,
+                    "rotation_type": null,
+                    "shape": [
+                        1
+                    ],
+                    "continuous": true
+                },
+                "torso_dyaw": {
+                    "absolute": false,
+                    "rotation_type": null,
+                    "shape": [
+                        1
+                    ],
+                    "continuous": true
+                }
+            }
+        },
+        "embodiment_tag": "real_teleop_g1"
+    }
+}

dreamzero_real_teleop_g1_full_finetune_absolute_Pick_bottle_and_turn_and_pour_into_cup/checkpoint-20000/global_step20000/bf16_zero_pp_rank_0_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3caff8ec74a8b6a9567475ebe40d3861fe8892f2b8c400d9ca996004310f3761
+size 24726494997

dreamzero_real_teleop_g1_full_finetune_absolute_Pick_bottle_and_turn_and_pour_into_cup/checkpoint-20000/global_step20000/bf16_zero_pp_rank_1_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:06d2a1b9ce3e5d4fb81678db9f7b4bf1347e4257d560b6675f2796546b2499eb
+size 24726494869

dreamzero_real_teleop_g1_full_finetune_absolute_Pick_bottle_and_turn_and_pour_into_cup/checkpoint-20000/global_step20000/bf16_zero_pp_rank_2_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7f1c8ed0278612e4bb5024be81202eaa2bdc612bb0e5ef8a1cbf8f631aec6f89
+size 24726494933

dreamzero_real_teleop_g1_full_finetune_absolute_Pick_bottle_and_turn_and_pour_into_cup/checkpoint-20000/global_step20000/bf16_zero_pp_rank_3_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:584d0b5d5dfa69327cd95583c1e442fba0c5509b07fe630ed9190c2a0d629983
+size 24726495189

dreamzero_real_teleop_g1_full_finetune_absolute_Pick_bottle_and_turn_and_pour_into_cup/checkpoint-20000/global_step20000/bf16_zero_pp_rank_4_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c9f62e5478f4b8d0139fe536d4e2ea88c0a8fcb31aaaca1937273ba6add59b11
+size 24726494869

dreamzero_real_teleop_g1_full_finetune_absolute_Pick_bottle_and_turn_and_pour_into_cup/checkpoint-20000/global_step20000/bf16_zero_pp_rank_5_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5cea7b073d45494887b902099a666ef0d51739f35d8d92def0f1864429cb7fec
+size 24726494805

dreamzero_real_teleop_g1_full_finetune_absolute_Pick_bottle_and_turn_and_pour_into_cup/checkpoint-20000/global_step20000/bf16_zero_pp_rank_6_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:53a042f0b18e2a32e863d50f6d772b6761f496680288cae301a1edfe2c227d98
+size 24726494805

dreamzero_real_teleop_g1_full_finetune_absolute_Pick_bottle_and_turn_and_pour_into_cup/checkpoint-20000/global_step20000/bf16_zero_pp_rank_7_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9f383a20036309f2b056fd11380d45b741b9e6637804de3739943ab4baaa9a95
+size 24726495317

dreamzero_real_teleop_g1_full_finetune_absolute_Pick_bottle_and_turn_and_pour_into_cup/checkpoint-20000/global_step20000/mp_rank_00_model_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:241bbf9b897f8e86fe844221bf2fd2fd3009ee2089b5467a9a859e379de0f754
+size 58729254967

dreamzero_real_teleop_g1_full_finetune_absolute_Pick_bottle_and_turn_and_pour_into_cup/checkpoint-20000/latest

@@ -0,0 +1 @@
+global_step20000

dreamzero_real_teleop_g1_full_finetune_absolute_Pick_bottle_and_turn_and_pour_into_cup/checkpoint-20000/model-00001-of-00010.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e0e0ff763ed8bb6b855782e10f1577d676bc7abce2df0e46cea5d340a05963a5
+size 4935816720

dreamzero_real_teleop_g1_full_finetune_absolute_Pick_bottle_and_turn_and_pour_into_cup/checkpoint-20000/model-00002-of-00010.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3dd6a6ed2735b920badc8b58864ac567f921b5e1ab537f9c755e35e6c65ebd8c
+size 4983107552

dreamzero_real_teleop_g1_full_finetune_absolute_Pick_bottle_and_turn_and_pour_into_cup/checkpoint-20000/model-00003-of-00010.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:90a95d54d28bbc3ddbc55719ae709e904548b740f25600c875adb639f6870888
+size 4937616256

dreamzero_real_teleop_g1_full_finetune_absolute_Pick_bottle_and_turn_and_pour_into_cup/checkpoint-20000/model-00004-of-00010.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7ca52a1d81944c771ba484b89adef9c8df476c7fca3e4ffac7f2619d0b307746
+size 4987670360

dreamzero_real_teleop_g1_full_finetune_absolute_Pick_bottle_and_turn_and_pour_into_cup/checkpoint-20000/model-00005-of-00010.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e5eb8b845be51f2f71959a4b75ffcae684bf87c1832f9affa2c3631be54040b8
+size 4987714608

dreamzero_real_teleop_g1_full_finetune_absolute_Pick_bottle_and_turn_and_pour_into_cup/checkpoint-20000/model-00006-of-00010.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5c90d9c99202128fd727ea9f10746b4716184989e5a0edf80f7bc2d3ffbae071
+size 4950963456

dreamzero_real_teleop_g1_full_finetune_absolute_Pick_bottle_and_turn_and_pour_into_cup/checkpoint-20000/model-00007-of-00010.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:36a44e22d355d0c600e60523182930d30b5c5139f33758c9fc8f66f153a706fa
+size 4951004896

dreamzero_real_teleop_g1_full_finetune_absolute_Pick_bottle_and_turn_and_pour_into_cup/checkpoint-20000/model-00008-of-00010.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:41528a5bcf542bc3e7f76c49d9bd6a14b696313d7183e085e9d331ddd8ce00e1
+size 4950963472

dreamzero_real_teleop_g1_full_finetune_absolute_Pick_bottle_and_turn_and_pour_into_cup/checkpoint-20000/model-00009-of-00010.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5824aff7cb28287d558676dd698be4aa781bb676ac87f2fb12a2b51793a79044
+size 4950984200

dreamzero_real_teleop_g1_full_finetune_absolute_Pick_bottle_and_turn_and_pour_into_cup/checkpoint-20000/model-00010-of-00010.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3346e9c1cbcd384e3694dfff4f52b22e1f0b841885faefe941bce96c8ebe40b3
+size 1212823208

dreamzero_real_teleop_g1_full_finetune_absolute_Pick_bottle_and_turn_and_pour_into_cup/checkpoint-20000/rng_state_0.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7c438dafa01e947a44498e649040021017fe8ac3136f94a6557fd9a73aa6fc87
+size 16389

dreamzero_real_teleop_g1_full_finetune_absolute_Pick_bottle_and_turn_and_pour_into_cup/checkpoint-20000/rng_state_1.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:67d51182ad9d940b5250d0f4063d4f029f086b6e9cf752e8ce5fef7190bbe2a6
+size 16389

dreamzero_real_teleop_g1_full_finetune_absolute_Pick_bottle_and_turn_and_pour_into_cup/checkpoint-20000/rng_state_2.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3c2469460a6376abd134137c65e6c5c24654fc30f4aa68654584ad0118a8a799
+size 16389

dreamzero_real_teleop_g1_full_finetune_absolute_Pick_bottle_and_turn_and_pour_into_cup/checkpoint-20000/rng_state_3.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:66302fb556ed76741228268c5831ac3c0a4d4c66bbc86fdef4bcf6ff391db590
+size 16389

dreamzero_real_teleop_g1_full_finetune_absolute_Pick_bottle_and_turn_and_pour_into_cup/checkpoint-20000/rng_state_4.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f366fae6f95875cd115f941788689420c7ea5dde16996514d0da7f7cf0614e0b
+size 16389

dreamzero_real_teleop_g1_full_finetune_absolute_Pick_bottle_and_turn_and_pour_into_cup/checkpoint-20000/rng_state_5.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8723611dff729f148415a3396f5beb7a562b2fd9a73d99dd46a6329f42cd678c
+size 16389

dreamzero_real_teleop_g1_full_finetune_absolute_Pick_bottle_and_turn_and_pour_into_cup/checkpoint-20000/rng_state_6.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a3b337e3ce96bf47ec270f4a05861e7868b21555fb7e8e181e0d106b0a8674e8
+size 16389

dreamzero_real_teleop_g1_full_finetune_absolute_Pick_bottle_and_turn_and_pour_into_cup/checkpoint-20000/rng_state_7.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f19001c88c670951f34333ba015a06134a3298870fc8c97ce8c5ec9b787261a1
+size 16389

dreamzero_real_teleop_g1_full_finetune_absolute_Pick_bottle_and_turn_and_pour_into_cup/checkpoint-20000/scheduler.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:28508b0a940eb7f712597070ad628ccf8634a0efd412740f51a02d7843f5d668
+size 1465

dreamzero_real_teleop_g1_full_finetune_absolute_Pick_bottle_and_turn_and_pour_into_cup/checkpoint-20000/wandb_config.json

@@ -0,0 +1 @@
+{"project": "dreamzero", "run_id": ""}

dreamzero_real_teleop_g1_full_finetune_absolute_Pick_bottle_and_turn_and_pour_into_cup/checkpoint-20000/zero_to_fp32.py

@@ -0,0 +1,760 @@
+#!/usr/bin/env python
+
+# Copyright (c) Microsoft Corporation.
+# SPDX-License-Identifier: Apache-2.0
+
+# DeepSpeed Team
+
+# This script extracts fp32 consolidated weights from a zero 1, 2 and 3 DeepSpeed checkpoints. It gets
+# copied into the top level checkpoint dir, so the user can easily do the conversion at any point in
+# the future. Once extracted, the weights don't require DeepSpeed and can be used in any
+# application.
+#
+# example:
+#   python zero_to_fp32.py . output_dir/
+#   or
+#   python zero_to_fp32.py . output_dir/ --safe_serialization
+
+import argparse
+import torch
+import glob
+import math
+import os
+import re
+import gc
+import json
+import numpy as np
+from tqdm import tqdm
+from collections import OrderedDict
+from dataclasses import dataclass
+
+# while this script doesn't use deepspeed to recover data, since the checkpoints are pickled with
+# DeepSpeed data structures it has to be available in the current python environment.
+from deepspeed.utils import logger
+from deepspeed.checkpoint.constants import (DS_VERSION, OPTIMIZER_STATE_DICT, SINGLE_PARTITION_OF_FP32_GROUPS,
+                                            FP32_FLAT_GROUPS, ZERO_STAGE, PARTITION_COUNT, PARAM_SHAPES, BUFFER_NAMES,
+                                            FROZEN_PARAM_SHAPES, FROZEN_PARAM_FRAGMENTS)
+
+
+@dataclass
+class zero_model_state:
+    buffers: dict()
+    param_shapes: dict()
+    shared_params: list
+    ds_version: int
+    frozen_param_shapes: dict()
+    frozen_param_fragments: dict()
+
+
+debug = 0
+
+# load to cpu
+device = torch.device('cpu')
+
+
+def atoi(text):
+    return int(text) if text.isdigit() else text
+
+
+def natural_keys(text):
+    '''
+    alist.sort(key=natural_keys) sorts in human order
+    http://nedbatchelder.com/blog/200712/human_sorting.html
+    (See Toothy's implementation in the comments)
+    '''
+    return [atoi(c) for c in re.split(r'(\d+)', text)]
+
+
+def get_model_state_file(checkpoint_dir, zero_stage):
+    if not os.path.isdir(checkpoint_dir):
+        raise FileNotFoundError(f"Directory '{checkpoint_dir}' doesn't exist")
+
+    # there should be only one file
+    if zero_stage <= 2:
+        file = os.path.join(checkpoint_dir, "mp_rank_00_model_states.pt")
+    elif zero_stage == 3:
+        file = os.path.join(checkpoint_dir, "zero_pp_rank_0_mp_rank_00_model_states.pt")
+
+    if not os.path.exists(file):
+        raise FileNotFoundError(f"can't find model states file at '{file}'")
+
+    return file
+
+
+def get_checkpoint_files(checkpoint_dir, glob_pattern):
+    # XXX: need to test that this simple glob rule works for multi-node setup too
+    ckpt_files = sorted(glob.glob(os.path.join(checkpoint_dir, glob_pattern)), key=natural_keys)
+
+    if len(ckpt_files) == 0:
+        raise FileNotFoundError(f"can't find {glob_pattern} files in directory '{checkpoint_dir}'")
+
+    return ckpt_files
+
+
+def get_optim_files(checkpoint_dir):
+    return get_checkpoint_files(checkpoint_dir, "*_optim_states.pt")
+
+
+def get_model_state_files(checkpoint_dir):
+    return get_checkpoint_files(checkpoint_dir, "*_model_states.pt")
+
+
+def parse_model_states(files):
+    zero_model_states = []
+    for file in files:
+        state_dict = torch.load(file, map_location=device, weights_only=False)
+
+        if BUFFER_NAMES not in state_dict:
+            raise ValueError(f"{file} is not a model state checkpoint")
+        buffer_names = state_dict[BUFFER_NAMES]
+        if debug:
+            print("Found buffers:", buffer_names)
+
+        # recover just the buffers while restoring them to fp32 if they were saved in fp16
+        buffers = {k: v.float() for k, v in state_dict["module"].items() if k in buffer_names}
+        param_shapes = state_dict[PARAM_SHAPES]
+
+        # collect parameters that are included in param_shapes
+        param_names = []
+        for s in param_shapes:
+            for name in s.keys():
+                param_names.append(name)
+
+        # update with frozen parameters
+        frozen_param_shapes = state_dict.get(FROZEN_PARAM_SHAPES, None)
+        if frozen_param_shapes is not None:
+            if debug:
+                print(f"Found frozen_param_shapes: {frozen_param_shapes}")
+            param_names += list(frozen_param_shapes.keys())
+
+        # handle shared params
+        shared_params = [[k, v] for k, v in state_dict["shared_params"].items()]
+
+        ds_version = state_dict.get(DS_VERSION, None)
+
+        frozen_param_fragments = state_dict.get(FROZEN_PARAM_FRAGMENTS, None)
+
+        z_model_state = zero_model_state(buffers=buffers,
+                                         param_shapes=param_shapes,
+                                         shared_params=shared_params,
+                                         ds_version=ds_version,
+                                         frozen_param_shapes=frozen_param_shapes,
+                                         frozen_param_fragments=frozen_param_fragments)
+        zero_model_states.append(z_model_state)
+
+    return zero_model_states
+
+
+def parse_optim_states(files, ds_checkpoint_dir):
+    total_files = len(files)
+    state_dicts = []
+    for f in tqdm(files, desc='Loading checkpoint shards'):
+        state_dict = torch.load(f, map_location=device, mmap=True, weights_only=False)
+        # immediately discard the potentially huge 2 optimizer states as we only care for fp32 master weights
+        # and also handle the case where it was already removed by another helper script
+        state_dict["optimizer_state_dict"].pop("optimizer_state_dict", None)
+        state_dicts.append(state_dict)
+
+    if ZERO_STAGE not in state_dicts[0][OPTIMIZER_STATE_DICT]:
+        raise ValueError(f"{files[0]} is not a zero checkpoint")
+    zero_stage = state_dicts[0][OPTIMIZER_STATE_DICT][ZERO_STAGE]
+    world_size = state_dicts[0][OPTIMIZER_STATE_DICT][PARTITION_COUNT]
+
+    # For ZeRO-2 each param group can have different partition_count as data parallelism for expert
+    # parameters can be different from data parallelism for non-expert parameters. So we can just
+    # use the max of the partition_count to get the dp world_size.
+
+    if type(world_size) is list:
+        world_size = max(world_size)
+
+    if world_size != total_files:
+        raise ValueError(
+            f"Expected {world_size} of '*_optim_states.pt' under '{ds_checkpoint_dir}' but found {total_files} files. "
+            "Possibly due to an overwrite of an old checkpoint, or a checkpoint didn't get saved by one or more processes."
+        )
+
+    # the groups are named differently in each stage
+    if zero_stage <= 2:
+        fp32_groups_key = SINGLE_PARTITION_OF_FP32_GROUPS
+    elif zero_stage == 3:
+        fp32_groups_key = FP32_FLAT_GROUPS
+    else:
+        raise ValueError(f"unknown zero stage {zero_stage}")
+
+    fp32_flat_groups = [state_dicts[i][OPTIMIZER_STATE_DICT][fp32_groups_key] for i in range(len(state_dicts))]
+    return zero_stage, world_size, fp32_flat_groups
+
+
+def _get_fp32_state_dict_from_zero_checkpoint(ds_checkpoint_dir, exclude_frozen_parameters):
+    """
+    Returns fp32 state_dict reconstructed from ds checkpoint
+
+    Args:
+        - ``ds_checkpoint_dir``: path to the deepspeed checkpoint folder (where the optimizer files are)
+
+    """
+    print(f"Processing zero checkpoint '{ds_checkpoint_dir}'")
+
+    optim_files = get_optim_files(ds_checkpoint_dir)
+    zero_stage, world_size, fp32_flat_groups = parse_optim_states(optim_files, ds_checkpoint_dir)
+    print(f"Detected checkpoint of type zero stage {zero_stage}, world_size: {world_size}")
+
+    model_files = get_model_state_files(ds_checkpoint_dir)
+
+    zero_model_states = parse_model_states(model_files)
+    print(f'Parsing checkpoint created by deepspeed=={zero_model_states[0].ds_version}')
+
+    if zero_stage <= 2:
+        return _get_fp32_state_dict_from_zero2_checkpoint(world_size, fp32_flat_groups, zero_model_states,
+                                                          exclude_frozen_parameters)
+    elif zero_stage == 3:
+        return _get_fp32_state_dict_from_zero3_checkpoint(world_size, fp32_flat_groups, zero_model_states,
+                                                          exclude_frozen_parameters)
+
+
+def _zero2_merge_frozen_params(state_dict, zero_model_states):
+    if zero_model_states[0].frozen_param_shapes is None or len(zero_model_states[0].frozen_param_shapes) == 0:
+        return
+
+    frozen_param_shapes = zero_model_states[0].frozen_param_shapes
+    frozen_param_fragments = zero_model_states[0].frozen_param_fragments
+
+    if debug:
+        num_elem = sum(s.numel() for s in frozen_param_shapes.values())
+        print(f'rank 0: {FROZEN_PARAM_SHAPES}.numel = {num_elem}')
+
+        wanted_params = len(frozen_param_shapes)
+        wanted_numel = sum(s.numel() for s in frozen_param_shapes.values())
+        avail_numel = sum([p.numel() for p in frozen_param_fragments.values()])
+        print(f'Frozen params: Have {avail_numel} numels to process.')
+        print(f'Frozen params: Need {wanted_numel} numels in {wanted_params} params')
+
+    total_params = 0
+    total_numel = 0
+    for name, shape in frozen_param_shapes.items():
+        total_params += 1
+        unpartitioned_numel = shape.numel()
+        total_numel += unpartitioned_numel
+
+        state_dict[name] = frozen_param_fragments[name]
+
+        if debug:
+            print(f"{name} full shape: {shape} unpartitioned numel {unpartitioned_numel} ")
+
+    print(f"Reconstructed Frozen fp32 state dict with {total_params} params {total_numel} elements")
+
+
+def _has_callable(obj, fn):
+    attr = getattr(obj, fn, None)
+    return callable(attr)
+
+
+def _zero2_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states):
+    param_shapes = zero_model_states[0].param_shapes
+
+    # Reconstruction protocol:
+    #
+    # XXX: document this
+
+    if debug:
+        for i in range(world_size):
+            for j in range(len(fp32_flat_groups[0])):
+                print(f"{FP32_FLAT_GROUPS}[{i}][{j}].shape={fp32_flat_groups[i][j].shape}")
+
+    # XXX: memory usage doubles here (zero2)
+    num_param_groups = len(fp32_flat_groups[0])
+    merged_single_partition_of_fp32_groups = []
+    for i in range(num_param_groups):
+        merged_partitions = [sd[i] for sd in fp32_flat_groups]
+        full_single_fp32_vector = torch.cat(merged_partitions, 0)
+        merged_single_partition_of_fp32_groups.append(full_single_fp32_vector)
+    avail_numel = sum(
+        [full_single_fp32_vector.numel() for full_single_fp32_vector in merged_single_partition_of_fp32_groups])
+
+    if debug:
+        wanted_params = sum([len(shapes) for shapes in param_shapes])
+        wanted_numel = sum([sum(shape.numel() for shape in shapes.values()) for shapes in param_shapes])
+        # not asserting if there is a mismatch due to possible padding
+        print(f"Have {avail_numel} numels to process.")
+        print(f"Need {wanted_numel} numels in {wanted_params} params.")
+
+    # params
+    # XXX: for huge models that can't fit into the host's RAM we will have to recode this to support
+    # out-of-core computing solution
+    total_numel = 0
+    total_params = 0
+    for shapes, full_single_fp32_vector in zip(param_shapes, merged_single_partition_of_fp32_groups):
+        offset = 0
+        avail_numel = full_single_fp32_vector.numel()
+        for name, shape in shapes.items():
+
+            unpartitioned_numel = shape.numel() if _has_callable(shape, 'numel') else math.prod(shape)
+            total_numel += unpartitioned_numel
+            total_params += 1
+
+            if debug:
+                print(f"{name} full shape: {shape} unpartitioned numel {unpartitioned_numel} ")
+            state_dict[name] = full_single_fp32_vector.narrow(0, offset, unpartitioned_numel).view(shape)
+            offset += unpartitioned_numel
+
+        # Z2 started to align to 2*world_size to improve nccl performance. Therefore both offset and
+        # avail_numel can differ by anywhere between 0..2*world_size. Due to two unrelated complex
+        # paddings performed in the code it's almost impossible to predict the exact numbers w/o the
+        # live optimizer object, so we are checking that the numbers are within the right range
+        align_to = 2 * world_size
+
+        def zero2_align(x):
+            return align_to * math.ceil(x / align_to)
+
+        if debug:
+            print(f"original offset={offset}, avail_numel={avail_numel}")
+
+        offset = zero2_align(offset)
+        avail_numel = zero2_align(avail_numel)
+
+        if debug:
+            print(f"aligned  offset={offset}, avail_numel={avail_numel}")
+
+        # Sanity check
+        if offset != avail_numel:
+            raise ValueError(f"consumed {offset} numels out of {avail_numel} - something is wrong")
+
+    print(f"Reconstructed fp32 state dict with {total_params} params {total_numel} elements")
+
+
+def _get_fp32_state_dict_from_zero2_checkpoint(world_size, fp32_flat_groups, zero_model_states,
+                                               exclude_frozen_parameters):
+    state_dict = OrderedDict()
+
+    # buffers
+    buffers = zero_model_states[0].buffers
+    state_dict.update(buffers)
+    if debug:
+        print(f"added {len(buffers)} buffers")
+
+    if not exclude_frozen_parameters:
+        _zero2_merge_frozen_params(state_dict, zero_model_states)
+
+    _zero2_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states)
+
+    # recover shared parameters
+    for pair in zero_model_states[0].shared_params:
+        if pair[1] in state_dict:
+            state_dict[pair[0]] = state_dict[pair[1]]
+
+    return state_dict
+
+
+def zero3_partitioned_param_info(unpartitioned_numel, world_size):
+    remainder = unpartitioned_numel % world_size
+    padding_numel = (world_size - remainder) if remainder else 0
+    partitioned_numel = math.ceil(unpartitioned_numel / world_size)
+    return partitioned_numel, padding_numel
+
+
+def _zero3_merge_frozen_params(state_dict, world_size, zero_model_states):
+    if zero_model_states[0].frozen_param_shapes is None or len(zero_model_states[0].frozen_param_shapes) == 0:
+        return
+
+    if debug:
+        for i in range(world_size):
+            num_elem = sum(s.numel() for s in zero_model_states[i].frozen_param_fragments.values())
+            print(f'rank {i}: {FROZEN_PARAM_SHAPES}.numel = {num_elem}')
+
+        frozen_param_shapes = zero_model_states[0].frozen_param_shapes
+        wanted_params = len(frozen_param_shapes)
+        wanted_numel = sum(s.numel() for s in frozen_param_shapes.values())
+        avail_numel = sum([p.numel() for p in zero_model_states[0].frozen_param_fragments.values()]) * world_size
+        print(f'Frozen params: Have {avail_numel} numels to process.')
+        print(f'Frozen params: Need {wanted_numel} numels in {wanted_params} params')
+
+    total_params = 0
+    total_numel = 0
+    for name, shape in zero_model_states[0].frozen_param_shapes.items():
+        total_params += 1
+        unpartitioned_numel = shape.numel()
+        total_numel += unpartitioned_numel
+
+        param_frags = tuple(model_state.frozen_param_fragments[name] for model_state in zero_model_states)
+        state_dict[name] = torch.cat(param_frags, 0).narrow(0, 0, unpartitioned_numel).view(shape)
+
+        partitioned_numel, partitioned_padding_numel = zero3_partitioned_param_info(unpartitioned_numel, world_size)
+
+        if debug:
+            print(
+                f"Frozen params: {total_params} {name} full shape: {shape} partition0 numel={partitioned_numel} partitioned_padding_numel={partitioned_padding_numel}"
+            )
+
+    print(f"Reconstructed Frozen fp32 state dict with {total_params} params {total_numel} elements")
+
+
+class GatheredTensor:
+    """
+    A pseudo tensor that collects partitioned weights.
+    It is more memory efficient when there are multiple groups.
+    """
+
+    def __init__(self, flat_groups, flat_groups_offset, offset, partitioned_numel, shape):
+        self.flat_groups = flat_groups
+        self.flat_groups_offset = flat_groups_offset
+        self.offset = offset
+        self.partitioned_numel = partitioned_numel
+        self.shape = shape
+        self.dtype = self.flat_groups[0][0].dtype
+
+    def contiguous(self):
+        """
+        Merge partitioned weights from flat_groups into a single tensor.
+        """
+        end_idx = self.offset + self.partitioned_numel
+        world_size = len(self.flat_groups)
+        pad_flat_param_chunks = []
+
+        for rank_i in range(world_size):
+            # for each rank, we need to collect weights from related group/groups
+            flat_groups_at_rank_i = self.flat_groups[rank_i]
+            start_group_id = None
+            end_group_id = None
+            for group_id in range(len(self.flat_groups_offset)):
+                if self.flat_groups_offset[group_id] <= self.offset < self.flat_groups_offset[group_id + 1]:
+                    start_group_id = group_id
+                if self.flat_groups_offset[group_id] < end_idx <= self.flat_groups_offset[group_id + 1]:
+                    end_group_id = group_id
+                    break
+            # collect weights from related group/groups
+            for group_id in range(start_group_id, end_group_id + 1):
+                flat_tensor = flat_groups_at_rank_i[group_id]
+                start_offset = self.offset - self.flat_groups_offset[group_id]
+                end_offset = min(end_idx, self.flat_groups_offset[group_id + 1]) - self.flat_groups_offset[group_id]
+                pad_flat_param_chunks.append(flat_tensor[start_offset:end_offset])
+
+        # collect weights from all ranks
+        pad_flat_param = torch.cat(pad_flat_param_chunks, dim=0)
+        param = pad_flat_param[:self.shape.numel()].view(self.shape).contiguous()
+        return param
+
+
+def _zero3_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states):
+    param_shapes = zero_model_states[0].param_shapes
+    avail_numel = sum([flat_group.numel() for flat_group in fp32_flat_groups[0]]) * world_size
+
+    # Reconstruction protocol: For zero3 we need to zip the partitions together at boundary of each
+    # param, re-consolidating each param, while dealing with padding if any
+
+    # merge list of dicts, preserving order
+    param_shapes = {k: v for d in param_shapes for k, v in d.items()}
+
+    if debug:
+        for i in range(world_size):
+            print(f"{FP32_FLAT_GROUPS}[{i}].shape={fp32_flat_groups[i].shape}")
+
+        wanted_params = len(param_shapes)
+        wanted_numel = sum(shape.numel() for shape in param_shapes.values())
+        # not asserting if there is a mismatch due to possible padding
+        avail_numel = fp32_flat_groups[0].numel() * world_size
+        print(f"Trainable params: Have {avail_numel} numels to process.")
+        print(f"Trainable params: Need {wanted_numel} numels in {wanted_params} params.")
+
+    # params
+    # XXX: for huge models that can't fit into the host's RAM we will have to recode this to support
+    # out-of-core computing solution
+    offset = 0
+    total_numel = 0
+    total_params = 0
+    flat_groups_offset = [0] + list(np.cumsum([flat_tensor.numel() for flat_tensor in fp32_flat_groups[0]]))
+    for name, shape in tqdm(param_shapes.items(), desc='Gathering sharded weights'):
+        unpartitioned_numel = shape.numel()
+        total_numel += unpartitioned_numel
+        total_params += 1
+        partitioned_numel, partitioned_padding_numel = zero3_partitioned_param_info(unpartitioned_numel, world_size)
+
+        if debug:
+            print(
+                f"Trainable params: {total_params} {name} full shape: {shape} partition0 numel={partitioned_numel} partitioned_padding_numel={partitioned_padding_numel}"
+            )
+
+        # memory efficient tensor
+        tensor = GatheredTensor(fp32_flat_groups, flat_groups_offset, offset, partitioned_numel, shape)
+        state_dict[name] = tensor
+        offset += partitioned_numel
+
+    offset *= world_size
+
+    # Sanity check
+    if offset != avail_numel:
+        raise ValueError(f"consumed {offset} numels out of {avail_numel} - something is wrong")
+
+    print(f"Reconstructed Trainable fp32 state dict with {total_params} params {total_numel} elements")
+
+
+def _get_fp32_state_dict_from_zero3_checkpoint(world_size, fp32_flat_groups, zero_model_states,
+                                               exclude_frozen_parameters):
+    state_dict = OrderedDict()
+
+    # buffers
+    buffers = zero_model_states[0].buffers
+    state_dict.update(buffers)
+    if debug:
+        print(f"added {len(buffers)} buffers")
+
+    if not exclude_frozen_parameters:
+        _zero3_merge_frozen_params(state_dict, world_size, zero_model_states)
+
+    _zero3_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states)
+
+    # recover shared parameters
+    for pair in zero_model_states[0].shared_params:
+        if pair[1] in state_dict:
+            state_dict[pair[0]] = state_dict[pair[1]]
+
+    return state_dict
+
+
+def to_torch_tensor(state_dict, return_empty_tensor=False):
+    """
+    Convert state_dict of GatheredTensor to torch tensor
+    """
+    torch_state_dict = {}
+    converted_tensors = {}
+    for name, tensor in state_dict.items():
+        tensor_id = id(tensor)
+        if tensor_id in converted_tensors:  # shared tensors
+            shared_tensor = torch_state_dict[converted_tensors[tensor_id]]
+            torch_state_dict[name] = shared_tensor
+        else:
+            converted_tensors[tensor_id] = name
+            if return_empty_tensor:
+                torch_state_dict[name] = torch.empty(tensor.shape, dtype=tensor.dtype)
+            else:
+                torch_state_dict[name] = tensor.contiguous()
+    return torch_state_dict
+
+
+def get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir,
+                                             tag=None,
+                                             exclude_frozen_parameters=False,
+                                             lazy_mode=False):
+    """
+    Convert ZeRO 2 or 3 checkpoint into a single fp32 consolidated state_dict that can be loaded with
+    ``load_state_dict()`` and used for training without DeepSpeed or shared with others, for example
+    via a model hub.
+
+    Args:
+        - ``checkpoint_dir``: path to the desired checkpoint folder
+        - ``tag``: checkpoint tag used as a unique identifier for checkpoint. If not provided will attempt to load tag in 'latest' file. e.g., ``global_step14``
+        - ``exclude_frozen_parameters``: exclude frozen parameters
+        - ``lazy_mode``: get state_dict in lazy mode. It returns a dict of pesduo tensor instead of torch tensor, which is more memory efficient.
+          Convert the pesduo tensor to torch tensor by ``.contiguous()``
+
+    Returns:
+        - pytorch ``state_dict``
+
+    A typical usage might be ::
+
+        from deepspeed.utils.zero_to_fp32 import get_fp32_state_dict_from_zero_checkpoint
+        # do the training and checkpoint saving
+        state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir) # already on cpu
+        model = model.cpu() # move to cpu
+        model.load_state_dict(state_dict)
+        # submit to model hub or save the model to share with others
+
+    In this example the ``model`` will no longer be usable in the deepspeed context of the same
+    application. i.e. you will need to re-initialize the deepspeed engine, since
+    ``model.load_state_dict(state_dict)`` will remove all the deepspeed magic from it.
+
+    If you want it all done for you, use ``load_state_dict_from_zero_checkpoint`` instead.
+
+    Note: the above usage may not work if your application doesn't have sufficient free CPU memory.
+    You may need to use the offline approach using the ``zero_to_fp32.py`` script that is saved with
+    the checkpoint. Or you can load state_dict in lazy mode ::
+
+        from deepspeed.utils.zero_to_fp32 import get_fp32_state_dict_from_zero_checkpoint
+        state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, lazy_mode=True) # not on cpu
+        for name, lazy_tensor in state_dict.item():
+            tensor = lazy_tensor.contiguous()  # to cpu
+            print(name, tensor)
+            # del tensor to release memory if it no longer in use
+    """
+    if tag is None:
+        latest_path = os.path.join(checkpoint_dir, 'latest')
+        if os.path.isfile(latest_path):
+            with open(latest_path, 'r') as fd:
+                tag = fd.read().strip()
+        else:
+            raise ValueError(f"Unable to find 'latest' file at {latest_path}")
+
+    ds_checkpoint_dir = os.path.join(checkpoint_dir, tag)
+
+    if not os.path.isdir(ds_checkpoint_dir):
+        raise FileNotFoundError(f"Directory '{ds_checkpoint_dir}' doesn't exist")
+
+    state_dict = _get_fp32_state_dict_from_zero_checkpoint(ds_checkpoint_dir, exclude_frozen_parameters)
+    if lazy_mode:
+        return state_dict
+    else:
+        return to_torch_tensor(state_dict)
+
+
+def convert_zero_checkpoint_to_fp32_state_dict(checkpoint_dir,
+                                               output_dir,
+                                               max_shard_size="5GB",
+                                               safe_serialization=False,
+                                               tag=None,
+                                               exclude_frozen_parameters=False):
+    """
+    Convert ZeRO 2 or 3 checkpoint into a single fp32 consolidated ``state_dict`` file that can be
+    loaded with ``torch.load(file)`` + ``load_state_dict()`` and used for training without DeepSpeed.
+
+    Args:
+        - ``checkpoint_dir``: path to the desired checkpoint folder. (one that contains the tag-folder, like ``global_step14``)
+        - ``output_dir``: directory to the pytorch fp32 state_dict output files
+        - ``max_shard_size``: the maximum size for a checkpoint before being sharded, default value is 5GB
+        - ``safe_serialization``:  whether to save the model using `safetensors` or the traditional PyTorch way (that uses `pickle`).
+        - ``tag``: checkpoint tag used as a unique identifier for checkpoint. If not provided will attempt to load tag in the file named ``latest`` in the checkpoint folder, e.g., ``global_step14``
+        - ``exclude_frozen_parameters``: exclude frozen parameters
+    """
+
+    # Dependency pre-check
+    if safe_serialization:
+        try:
+            from safetensors.torch import save_file
+        except ImportError:
+            print('If you want to use `safe_serialization`, please `pip install safetensors`')
+            raise
+    if max_shard_size is not None:
+        try:
+            from huggingface_hub import split_torch_state_dict_into_shards
+        except ImportError:
+            print('If you want to use `max_shard_size`, please `pip install huggingface_hub`')
+            raise
+
+    # Convert zero checkpoint to state_dict
+    state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir,
+                                                          tag,
+                                                          exclude_frozen_parameters,
+                                                          lazy_mode=True)
+
+    # Shard the model if it is too big.
+    weights_name = "model.safetensors" if safe_serialization else "pytorch_model.bin"
+    if max_shard_size is not None:
+        filename_pattern = weights_name.replace(".bin", "{suffix}.bin").replace(".safetensors", "{suffix}.safetensors")
+        # an memory-efficient approach for sharding
+        empty_state_dict = to_torch_tensor(state_dict, return_empty_tensor=True)
+        state_dict_split = split_torch_state_dict_into_shards(empty_state_dict,
+                                                              filename_pattern=filename_pattern,
+                                                              max_shard_size=max_shard_size)
+    else:
+        from collections import namedtuple
+        StateDictSplit = namedtuple("StateDictSplit", ["is_sharded", "filename_to_tensors"])
+        state_dict_split = StateDictSplit(is_sharded=False,
+                                          filename_to_tensors={weights_name: list(state_dict.keys())})
+
+    # Save the model by shard
+    os.makedirs(output_dir, exist_ok=True)
+    filename_to_tensors = state_dict_split.filename_to_tensors.items()
+    for shard_file, tensors in tqdm(filename_to_tensors, desc="Saving checkpoint shards"):
+        shard_state_dict = {tensor_name: state_dict[tensor_name] for tensor_name in tensors}
+        shard_state_dict = to_torch_tensor(shard_state_dict)
+        output_path = os.path.join(output_dir, shard_file)
+        if safe_serialization:
+            save_file(shard_state_dict, output_path, metadata={"format": "pt"})
+        else:
+            torch.save(shard_state_dict, output_path)
+        # release the memory of current shard
+        for tensor_name in list(shard_state_dict.keys()):
+            del state_dict[tensor_name]
+            del shard_state_dict[tensor_name]
+        del shard_state_dict
+        gc.collect()
+
+    # Save index if sharded
+    if state_dict_split.is_sharded:
+        index = {
+            "metadata": state_dict_split.metadata,
+            "weight_map": state_dict_split.tensor_to_filename,
+        }
+        save_index_file = "model.safetensors.index.json" if safe_serialization else "pytorch_model.bin.index.json"
+        save_index_file = os.path.join(output_dir, save_index_file)
+        with open(save_index_file, "w", encoding="utf-8") as f:
+            content = json.dumps(index, indent=2, sort_keys=True) + "\n"
+            f.write(content)
+
+
+def load_state_dict_from_zero_checkpoint(model, checkpoint_dir, tag=None):
+    """
+    1. Put the provided model to cpu
+    2. Convert ZeRO 2 or 3 checkpoint into a single fp32 consolidated ``state_dict``
+    3. Load it into the provided model
+
+    Args:
+        - ``model``: the model object to update
+        - ``checkpoint_dir``: path to the desired checkpoint folder. (one that contains the tag-folder, like ``global_step14``)
+        - ``tag``: checkpoint tag used as a unique identifier for checkpoint. If not provided will attempt to load tag in the file named ``latest`` in the checkpoint folder, e.g., ``global_step14``
+
+    Returns:
+        - ``model`: modified model
+
+    Make sure you have plenty of CPU memory available before you call this function. If you don't
+    have enough use the ``zero_to_fp32.py`` utility to do the conversion. You will find it
+    conveniently placed for you in the checkpoint folder.
+
+    A typical usage might be ::
+
+        from deepspeed.utils.zero_to_fp32 import load_state_dict_from_zero_checkpoint
+        model = load_state_dict_from_zero_checkpoint(trainer.model, checkpoint_dir)
+        # submit to model hub or save the model to share with others
+
+    Note, that once this was run, the ``model`` will no longer be usable in the deepspeed context
+    of the same application. i.e. you will need to re-initialize the deepspeed engine, since
+    ``model.load_state_dict(state_dict)`` will remove all the deepspeed magic from it.
+
+    """
+    logger.info("Extracting fp32 weights")
+    state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, tag)
+
+    logger.info("Overwriting model with fp32 weights")
+    model = model.cpu()
+    model.load_state_dict(state_dict, strict=False)
+
+    return model
+
+
+if __name__ == "__main__":
+    parser = argparse.ArgumentParser()
+    parser.add_argument("checkpoint_dir",
+                        type=str,
+                        help="path to the desired checkpoint folder, e.g., path/checkpoint-12")
+    parser.add_argument("output_dir",
+                        type=str,
+                        help="directory to the pytorch fp32 state_dict output files"
+                        "(e.g. path/checkpoint-12-output/)")
+    parser.add_argument(
+        "--max_shard_size",
+        type=str,
+        default="5GB",
+        help="The maximum size for a checkpoint before being sharded. Checkpoints shard will then be each of size"
+        "lower than this size. If expressed as a string, needs to be digits followed by a unit (like `5MB`"
+        "We default it to 5GB in order for models to be able to run easily on free-tier google colab instances"
+        "without CPU OOM issues.")
+    parser.add_argument(
+        "--safe_serialization",
+        default=False,
+        action='store_true',
+        help="Whether to save the model using `safetensors` or the traditional PyTorch way (that uses `pickle`).")
+    parser.add_argument("-t",
+                        "--tag",
+                        type=str,
+                        default=None,
+                        help="checkpoint tag used as a unique identifier for checkpoint. e.g., global_step1")
+    parser.add_argument("--exclude_frozen_parameters", action='store_true', help="exclude frozen parameters")
+    parser.add_argument("-d", "--debug", action='store_true', help="enable debug")
+    args = parser.parse_args()
+
+    debug = args.debug
+
+    convert_zero_checkpoint_to_fp32_state_dict(args.checkpoint_dir,
+                                               args.output_dir,
+                                               max_shard_size=args.max_shard_size,
+                                               safe_serialization=args.safe_serialization,
+                                               tag=args.tag,
+                                               exclude_frozen_parameters=args.exclude_frozen_parameters)

dreamzero_real_teleop_g1_full_finetune_absolute_Pick_bottle_and_turn_and_pour_into_cup/loss_log.jsonl

@@ -6290,3 +6290,610 @@
 {"step": 17970, "loss": 0.0586, "learning_rate": 2.7929753142667883e-07}
 {"step": 17970, "dynamics_loss_avg": 0.04188279584050179}
 {"step": 17970, "action_loss_avg": 0.020719322189688683}
+{"step": 17980, "loss": 0.0657, "learning_rate": 2.765795370105401e-07}
+{"step": 17980, "dynamics_loss_avg": 0.04451025417074561}
+{"step": 17980, "action_loss_avg": 0.025259029027074574}
+{"step": 17990, "loss": 0.0769, "learning_rate": 2.738744562455831e-07}
+{"step": 17990, "dynamics_loss_avg": 0.05019981041550636}
+{"step": 17990, "action_loss_avg": 0.019757727766409515}
+{"step": 18000, "loss": 0.0624, "learning_rate": 2.711822965273975e-07}
+{"step": 18000, "dynamics_loss_avg": 0.03530683349817991}
+{"step": 18000, "action_loss_avg": 0.01971647096797824}
+{"step": 18010, "loss": 0.0723, "learning_rate": 2.685030652162424e-07}
+{"step": 18010, "dynamics_loss_avg": 0.04856907473877072}
+{"step": 18010, "action_loss_avg": 0.023391773412004113}
+{"step": 18020, "loss": 0.0713, "learning_rate": 2.658367696370351e-07}
+{"step": 18020, "dynamics_loss_avg": 0.052192554157227275}
+{"step": 18020, "action_loss_avg": 0.01916952694300562}
+{"step": 18030, "loss": 0.0711, "learning_rate": 2.631834170793268e-07}
+{"step": 18030, "dynamics_loss_avg": 0.05118879219517112}
+{"step": 18030, "action_loss_avg": 0.018339621974155305}
+{"step": 18040, "loss": 0.0786, "learning_rate": 2.605430147972804e-07}
+{"step": 18040, "dynamics_loss_avg": 0.056528258882462976}
+{"step": 18040, "action_loss_avg": 0.03153302231803536}
+{"step": 18050, "loss": 0.093, "learning_rate": 2.579155700096575e-07}
+{"step": 18050, "dynamics_loss_avg": 0.06036507375538349}
+{"step": 18050, "action_loss_avg": 0.030267918528988958}
+{"step": 18060, "loss": 0.0572, "learning_rate": 2.5530108989978875e-07}
+{"step": 18060, "dynamics_loss_avg": 0.033854188490659}
+{"step": 18060, "action_loss_avg": 0.011101155891083181}
+{"step": 18070, "loss": 0.0622, "learning_rate": 2.526995816155642e-07}
+{"step": 18070, "dynamics_loss_avg": 0.03401682879775762}
+{"step": 18070, "action_loss_avg": 0.016593949124217033}
+{"step": 18080, "loss": 0.0651, "learning_rate": 2.501110522694089e-07}
+{"step": 18080, "dynamics_loss_avg": 0.04331094124354422}
+{"step": 18080, "action_loss_avg": 0.019172144029289483}
+{"step": 18090, "loss": 0.0722, "learning_rate": 2.475355089382625e-07}
+{"step": 18090, "dynamics_loss_avg": 0.04457311574369669}
+{"step": 18090, "action_loss_avg": 0.022115184273570777}
+{"step": 18100, "loss": 0.0765, "learning_rate": 2.4497295866356295e-07}
+{"step": 18100, "dynamics_loss_avg": 0.04849362997338176}
+{"step": 18100, "action_loss_avg": 0.021924686199054123}
+{"step": 18110, "loss": 0.0662, "learning_rate": 2.424234084512228e-07}
+{"step": 18110, "dynamics_loss_avg": 0.04284500405192375}
+{"step": 18110, "action_loss_avg": 0.013206246495246887}
+{"step": 18120, "loss": 0.0708, "learning_rate": 2.3988686527161686e-07}
+{"step": 18120, "dynamics_loss_avg": 0.04819774907082319}
+{"step": 18120, "action_loss_avg": 0.02406782256439328}
+{"step": 18130, "loss": 0.0681, "learning_rate": 2.373633360595573e-07}
+{"step": 18130, "dynamics_loss_avg": 0.050570879317820075}
+{"step": 18130, "action_loss_avg": 0.01779633527621627}
+{"step": 18140, "loss": 0.0715, "learning_rate": 2.3485282771427586e-07}
+{"step": 18140, "dynamics_loss_avg": 0.03040293287485838}
+{"step": 18140, "action_loss_avg": 0.012802653654944152}
+{"step": 18150, "loss": 0.0647, "learning_rate": 2.3235534709940665e-07}
+{"step": 18150, "dynamics_loss_avg": 0.038292928226292135}
+{"step": 18150, "action_loss_avg": 0.01287715877406299}
+{"step": 18160, "loss": 0.0739, "learning_rate": 2.2987090104296616e-07}
+{"step": 18160, "dynamics_loss_avg": 0.054934194311499596}
+{"step": 18160, "action_loss_avg": 0.026963144540786743}
+{"step": 18170, "loss": 0.0622, "learning_rate": 2.2739949633733549e-07}
+{"step": 18170, "dynamics_loss_avg": 0.0363878505770117}
+{"step": 18170, "action_loss_avg": 0.011499037570320069}
+{"step": 18180, "loss": 0.0633, "learning_rate": 2.2494113973924091e-07}
+{"step": 18180, "dynamics_loss_avg": 0.04405585685744882}
+{"step": 18180, "action_loss_avg": 0.01619838015176356}
+{"step": 18190, "loss": 0.0837, "learning_rate": 2.2249583796973507e-07}
+{"step": 18190, "dynamics_loss_avg": 0.05213240385055542}
+{"step": 18190, "action_loss_avg": 0.022103648679330946}
+{"step": 18200, "loss": 0.0613, "learning_rate": 2.200635977141796e-07}
+{"step": 18200, "dynamics_loss_avg": 0.053078430145978926}
+{"step": 18200, "action_loss_avg": 0.01745017133653164}
+{"step": 18210, "loss": 0.0761, "learning_rate": 2.1764442562222598e-07}
+{"step": 18210, "dynamics_loss_avg": 0.049139073304831984}
+{"step": 18210, "action_loss_avg": 0.024428400211036205}
+{"step": 18220, "loss": 0.0733, "learning_rate": 2.152383283077991e-07}
+{"step": 18220, "dynamics_loss_avg": 0.050191032327711584}
+{"step": 18220, "action_loss_avg": 0.02038680133409798}
+{"step": 18230, "loss": 0.0627, "learning_rate": 2.1284531234907812e-07}
+{"step": 18230, "dynamics_loss_avg": 0.04079515226185322}
+{"step": 18230, "action_loss_avg": 0.018291610572487117}
+{"step": 18240, "loss": 0.0555, "learning_rate": 2.1046538428847464e-07}
+{"step": 18240, "dynamics_loss_avg": 0.03139160391874611}
+{"step": 18240, "action_loss_avg": 0.019224531413055957}
+{"step": 18250, "loss": 0.0781, "learning_rate": 2.0809855063262275e-07}
+{"step": 18250, "dynamics_loss_avg": 0.05537284519523382}
+{"step": 18250, "action_loss_avg": 0.01678796047344804}
+{"step": 18260, "loss": 0.0738, "learning_rate": 2.0574481785235577e-07}
+{"step": 18260, "dynamics_loss_avg": 0.04461161983199417}
+{"step": 18260, "action_loss_avg": 0.01744754402898252}
+{"step": 18270, "loss": 0.0541, "learning_rate": 2.034041923826885e-07}
+{"step": 18270, "dynamics_loss_avg": 0.03249894678592682}
+{"step": 18270, "action_loss_avg": 0.013080501649528741}
+{"step": 18280, "loss": 0.0731, "learning_rate": 2.0107668062280207e-07}
+{"step": 18280, "dynamics_loss_avg": 0.04757325639948249}
+{"step": 18280, "action_loss_avg": 0.017159340134821833}
+{"step": 18290, "loss": 0.0792, "learning_rate": 1.9876228893602355e-07}
+{"step": 18290, "dynamics_loss_avg": 0.06475771861150861}
+{"step": 18290, "action_loss_avg": 0.02554315598681569}
+{"step": 18300, "loss": 0.0815, "learning_rate": 1.9646102364981269e-07}
+{"step": 18300, "dynamics_loss_avg": 0.05235052909702063}
+{"step": 18300, "action_loss_avg": 0.02022950756363571}
+{"step": 18310, "loss": 0.0734, "learning_rate": 1.9417289105574054e-07}
+{"step": 18310, "dynamics_loss_avg": 0.05568016031756997}
+{"step": 18310, "action_loss_avg": 0.018663786817342043}
+{"step": 18320, "loss": 0.0784, "learning_rate": 1.918978974094743e-07}
+{"step": 18320, "dynamics_loss_avg": 0.051876201014965775}
+{"step": 18320, "action_loss_avg": 0.02238959036767483}
+{"step": 18330, "loss": 0.0682, "learning_rate": 1.896360489307597e-07}
+{"step": 18330, "dynamics_loss_avg": 0.04768909700214863}
+{"step": 18330, "action_loss_avg": 0.011938971444033086}
+{"step": 18340, "loss": 0.0702, "learning_rate": 1.8738735180340362e-07}
+{"step": 18340, "dynamics_loss_avg": 0.04586377870291471}
+{"step": 18340, "action_loss_avg": 0.017745603155344725}
+{"step": 18350, "loss": 0.0744, "learning_rate": 1.8515181217525823e-07}
+{"step": 18350, "dynamics_loss_avg": 0.04679070338606835}
+{"step": 18350, "action_loss_avg": 0.019874962512403727}
+{"step": 18360, "loss": 0.0724, "learning_rate": 1.8292943615820458e-07}
+{"step": 18360, "dynamics_loss_avg": 0.0325282221660018}
+{"step": 18360, "action_loss_avg": 0.012186819082126021}
+{"step": 18370, "loss": 0.0692, "learning_rate": 1.8072022982813297e-07}
+{"step": 18370, "dynamics_loss_avg": 0.03504083342850208}
+{"step": 18370, "action_loss_avg": 0.017874133703298867}
+{"step": 18380, "loss": 0.0647, "learning_rate": 1.7852419922492925e-07}
+{"step": 18380, "dynamics_loss_avg": 0.04656160455197096}
+{"step": 18380, "action_loss_avg": 0.02956375349313021}
+{"step": 18390, "loss": 0.0618, "learning_rate": 1.7634135035245692e-07}
+{"step": 18390, "dynamics_loss_avg": 0.03509335024282336}
+{"step": 18390, "action_loss_avg": 0.01892667606007308}
+{"step": 18400, "loss": 0.0679, "learning_rate": 1.7417168917854167e-07}
+{"step": 18400, "dynamics_loss_avg": 0.04116056151688099}
+{"step": 18400, "action_loss_avg": 0.01439917974639684}
+{"step": 18410, "loss": 0.067, "learning_rate": 1.720152216349552e-07}
+{"step": 18410, "dynamics_loss_avg": 0.04588055443018675}
+{"step": 18410, "action_loss_avg": 0.015296191233210266}
+{"step": 18420, "loss": 0.0616, "learning_rate": 1.6987195361739595e-07}
+{"step": 18420, "dynamics_loss_avg": 0.04652413250878453}
+{"step": 18420, "action_loss_avg": 0.02111965431831777}
+{"step": 18430, "loss": 0.0711, "learning_rate": 1.6774189098547833e-07}
+{"step": 18430, "dynamics_loss_avg": 0.05404613558202982}
+{"step": 18430, "action_loss_avg": 0.022632275894284248}
+{"step": 18440, "loss": 0.0579, "learning_rate": 1.656250395627107e-07}
+{"step": 18440, "dynamics_loss_avg": 0.04214397612959146}
+{"step": 18440, "action_loss_avg": 0.011458933795802295}
+{"step": 18450, "loss": 0.0849, "learning_rate": 1.6352140513648418e-07}
+{"step": 18450, "dynamics_loss_avg": 0.06370990518480539}
+{"step": 18450, "action_loss_avg": 0.019021764234639704}
+{"step": 18460, "loss": 0.0576, "learning_rate": 1.6143099345805714e-07}
+{"step": 18460, "dynamics_loss_avg": 0.03974438169971108}
+{"step": 18460, "action_loss_avg": 0.011968593345955014}
+{"step": 18470, "loss": 0.0704, "learning_rate": 1.5935381024253294e-07}
+{"step": 18470, "dynamics_loss_avg": 0.04946697074919939}
+{"step": 18470, "action_loss_avg": 0.02251418838277459}
+{"step": 18480, "loss": 0.0658, "learning_rate": 1.572898611688517e-07}
+{"step": 18480, "dynamics_loss_avg": 0.049329818598926066}
+{"step": 18480, "action_loss_avg": 0.0209303927840665}
+{"step": 18490, "loss": 0.085, "learning_rate": 1.5523915187977135e-07}
+{"step": 18490, "dynamics_loss_avg": 0.06325942352414131}
+{"step": 18490, "action_loss_avg": 0.021714000264182687}
+{"step": 18500, "loss": 0.0679, "learning_rate": 1.532016879818532e-07}
+{"step": 18500, "dynamics_loss_avg": 0.04309999439865351}
+{"step": 18500, "action_loss_avg": 0.01855244832113385}
+{"step": 18510, "loss": 0.0574, "learning_rate": 1.51177475045447e-07}
+{"step": 18510, "dynamics_loss_avg": 0.03668638826347888}
+{"step": 18510, "action_loss_avg": 0.014108453458175064}
+{"step": 18520, "loss": 0.0653, "learning_rate": 1.4916651860467035e-07}
+{"step": 18520, "dynamics_loss_avg": 0.0457814609631896}
+{"step": 18520, "action_loss_avg": 0.01547098788432777}
+{"step": 18530, "loss": 0.0888, "learning_rate": 1.471688241574043e-07}
+{"step": 18530, "dynamics_loss_avg": 0.06023927703499794}
+{"step": 18530, "action_loss_avg": 0.01835664624813944}
+{"step": 18540, "loss": 0.0617, "learning_rate": 1.4518439716526721e-07}
+{"step": 18540, "dynamics_loss_avg": 0.04002130590379238}
+{"step": 18540, "action_loss_avg": 0.011030881525948643}
+{"step": 18550, "loss": 0.0706, "learning_rate": 1.432132430536076e-07}
+{"step": 18550, "dynamics_loss_avg": 0.05326564330607653}
+{"step": 18550, "action_loss_avg": 0.028722409531474112}
+{"step": 18560, "loss": 0.0679, "learning_rate": 1.412553672114869e-07}
+{"step": 18560, "dynamics_loss_avg": 0.03678571954369545}
+{"step": 18560, "action_loss_avg": 0.013442619773559272}
+{"step": 18570, "loss": 0.0762, "learning_rate": 1.3931077499166056e-07}
+{"step": 18570, "dynamics_loss_avg": 0.051235838793218136}
+{"step": 18570, "action_loss_avg": 0.014766597165726125}
+{"step": 18580, "loss": 0.0623, "learning_rate": 1.3737947171057087e-07}
+{"step": 18580, "dynamics_loss_avg": 0.04611561596393585}
+{"step": 18580, "action_loss_avg": 0.024290019972249867}
+{"step": 18590, "loss": 0.0617, "learning_rate": 1.3546146264832583e-07}
+{"step": 18590, "dynamics_loss_avg": 0.043639526097103955}
+{"step": 18590, "action_loss_avg": 0.020817835722118615}
+{"step": 18600, "loss": 0.0542, "learning_rate": 1.3355675304869086e-07}
+{"step": 18600, "dynamics_loss_avg": 0.041601867880672216}
+{"step": 18600, "action_loss_avg": 0.018878667848184705}
+{"step": 18610, "loss": 0.0741, "learning_rate": 1.3166534811906827e-07}
+{"step": 18610, "dynamics_loss_avg": 0.03837147168815136}
+{"step": 18610, "action_loss_avg": 0.012793690583202988}
+{"step": 18620, "loss": 0.0573, "learning_rate": 1.2978725303048666e-07}
+{"step": 18620, "dynamics_loss_avg": 0.040544414147734645}
+{"step": 18620, "action_loss_avg": 0.014463109290227294}
+{"step": 18630, "loss": 0.0763, "learning_rate": 1.2792247291758763e-07}
+{"step": 18630, "dynamics_loss_avg": 0.047443586960434915}
+{"step": 18630, "action_loss_avg": 0.01775962356477976}
+{"step": 18640, "loss": 0.0729, "learning_rate": 1.2607101287860635e-07}
+{"step": 18640, "dynamics_loss_avg": 0.05059406105428934}
+{"step": 18640, "action_loss_avg": 0.017567940149456264}
+{"step": 18650, "loss": 0.0812, "learning_rate": 1.2423287797536654e-07}
+{"step": 18650, "dynamics_loss_avg": 0.0561508284881711}
+{"step": 18650, "action_loss_avg": 0.02486546281725168}
+{"step": 18660, "loss": 0.0649, "learning_rate": 1.2240807323325775e-07}
+{"step": 18660, "dynamics_loss_avg": 0.03691480532288551}
+{"step": 18660, "action_loss_avg": 0.013475889107212424}
+{"step": 18670, "loss": 0.0563, "learning_rate": 1.2059660364122538e-07}
+{"step": 18670, "dynamics_loss_avg": 0.042761373054236176}
+{"step": 18670, "action_loss_avg": 0.015165302250534297}
+{"step": 18680, "loss": 0.0611, "learning_rate": 1.187984741517595e-07}
+{"step": 18680, "dynamics_loss_avg": 0.038918837811797856}
+{"step": 18680, "action_loss_avg": 0.014585279813036322}
+{"step": 18690, "loss": 0.0792, "learning_rate": 1.1701368968087711e-07}
+{"step": 18690, "dynamics_loss_avg": 0.04762068502604962}
+{"step": 18690, "action_loss_avg": 0.02017206819728017}
+{"step": 18700, "loss": 0.0607, "learning_rate": 1.1524225510811116e-07}
+{"step": 18700, "dynamics_loss_avg": 0.04003286678344011}
+{"step": 18700, "action_loss_avg": 0.017731298226863146}
+{"step": 18710, "loss": 0.0674, "learning_rate": 1.1348417527649536e-07}
+{"step": 18710, "dynamics_loss_avg": 0.04693224634975195}
+{"step": 18710, "action_loss_avg": 0.027163309697061776}
+{"step": 18720, "loss": 0.0697, "learning_rate": 1.1173945499255268e-07}
+{"step": 18720, "dynamics_loss_avg": 0.04715399304404855}
+{"step": 18720, "action_loss_avg": 0.020895790844224393}
+{"step": 18730, "loss": 0.0843, "learning_rate": 1.1000809902628307e-07}
+{"step": 18730, "dynamics_loss_avg": 0.056629661656916144}
+{"step": 18730, "action_loss_avg": 0.02812079694122076}
+{"step": 18740, "loss": 0.0694, "learning_rate": 1.0829011211114681e-07}
+{"step": 18740, "dynamics_loss_avg": 0.04512308090925217}
+{"step": 18740, "action_loss_avg": 0.012506886711344123}
+{"step": 18750, "loss": 0.0625, "learning_rate": 1.0658549894405457e-07}
+{"step": 18750, "dynamics_loss_avg": 0.049353060126304624}
+{"step": 18750, "action_loss_avg": 0.01917194863781333}
+{"step": 18760, "loss": 0.0716, "learning_rate": 1.0489426418535342e-07}
+{"step": 18760, "dynamics_loss_avg": 0.045650312304496767}
+{"step": 18760, "action_loss_avg": 0.013405046029947697}
+{"step": 18770, "loss": 0.0768, "learning_rate": 1.0321641245881475e-07}
+{"step": 18770, "dynamics_loss_avg": 0.046050991769880055}
+{"step": 18770, "action_loss_avg": 0.02187738287029788}
+{"step": 18780, "loss": 0.0672, "learning_rate": 1.0155194835162141e-07}
+{"step": 18780, "dynamics_loss_avg": 0.04976412951946259}
+{"step": 18780, "action_loss_avg": 0.023234894685447215}
+{"step": 18790, "loss": 0.0636, "learning_rate": 9.990087641435443e-08}
+{"step": 18790, "dynamics_loss_avg": 0.05185487000271678}
+{"step": 18790, "action_loss_avg": 0.016145678516477347}
+{"step": 18800, "loss": 0.0675, "learning_rate": 9.826320116098131e-08}
+{"step": 18800, "dynamics_loss_avg": 0.04294224474579096}
+{"step": 18800, "action_loss_avg": 0.01435599448159337}
+{"step": 18810, "loss": 0.0695, "learning_rate": 9.663892706884448e-08}
+{"step": 18810, "dynamics_loss_avg": 0.06062636505812406}
+{"step": 18810, "action_loss_avg": 0.022002062015235423}
+{"step": 18820, "loss": 0.0741, "learning_rate": 9.502805857864617e-08}
+{"step": 18820, "dynamics_loss_avg": 0.045694983564317226}
+{"step": 18820, "action_loss_avg": 0.014170713955536485}
+{"step": 18830, "loss": 0.0655, "learning_rate": 9.343060009443961e-08}
+{"step": 18830, "dynamics_loss_avg": 0.056142124626785514}
+{"step": 18830, "action_loss_avg": 0.015817865310236812}
+{"step": 18840, "loss": 0.0752, "learning_rate": 9.184655598361624e-08}
+{"step": 18840, "dynamics_loss_avg": 0.05955648459494114}
+{"step": 18840, "action_loss_avg": 0.019305437244474886}
+{"step": 18850, "loss": 0.0802, "learning_rate": 9.027593057689077e-08}
+{"step": 18850, "dynamics_loss_avg": 0.05589356829877943}
+{"step": 18850, "action_loss_avg": 0.02508344443049282}
+{"step": 18860, "loss": 0.0742, "learning_rate": 8.871872816829442e-08}
+{"step": 18860, "dynamics_loss_avg": 0.04614152507856488}
+{"step": 18860, "action_loss_avg": 0.018156480311881752}
+{"step": 18870, "loss": 0.0594, "learning_rate": 8.717495301515777e-08}
+{"step": 18870, "dynamics_loss_avg": 0.030260950746014714}
+{"step": 18870, "action_loss_avg": 0.017760740919038655}
+{"step": 18880, "loss": 0.0856, "learning_rate": 8.564460933810414e-08}
+{"step": 18880, "dynamics_loss_avg": 0.0606656925752759}
+{"step": 18880, "action_loss_avg": 0.022149211214855313}
+{"step": 18890, "loss": 0.0675, "learning_rate": 8.412770132103454e-08}
+{"step": 18890, "dynamics_loss_avg": 0.0462961982935667}
+{"step": 18890, "action_loss_avg": 0.026284475857391954}
+{"step": 18900, "loss": 0.0674, "learning_rate": 8.262423311111711e-08}
+{"step": 18900, "dynamics_loss_avg": 0.03955181259661913}
+{"step": 18900, "action_loss_avg": 0.017146135633811354}
+{"step": 18910, "loss": 0.0538, "learning_rate": 8.113420881877665e-08}
+{"step": 18910, "dynamics_loss_avg": 0.027236733864992858}
+{"step": 18910, "action_loss_avg": 0.007272660755552352}
+{"step": 18920, "loss": 0.0579, "learning_rate": 7.965763251768289e-08}
+{"step": 18920, "dynamics_loss_avg": 0.03394540082663298}
+{"step": 18920, "action_loss_avg": 0.016379949520342052}
+{"step": 18930, "loss": 0.0757, "learning_rate": 7.819450824473995e-08}
+{"step": 18930, "dynamics_loss_avg": 0.06098019005730748}
+{"step": 18930, "action_loss_avg": 0.019424212723970415}
+{"step": 18940, "loss": 0.0631, "learning_rate": 7.674484000007199e-08}
+{"step": 18940, "dynamics_loss_avg": 0.05012268982827663}
+{"step": 18940, "action_loss_avg": 0.012639198126271367}
+{"step": 18950, "loss": 0.0632, "learning_rate": 7.530863174701752e-08}
+{"step": 18950, "dynamics_loss_avg": 0.046331493742764}
+{"step": 18950, "action_loss_avg": 0.016248486703261733}
+{"step": 18960, "loss": 0.0648, "learning_rate": 7.388588741211511e-08}
+{"step": 18960, "dynamics_loss_avg": 0.04534003343433142}
+{"step": 18960, "action_loss_avg": 0.018011249601840973}
+{"step": 18970, "loss": 0.0579, "learning_rate": 7.247661088509328e-08}
+{"step": 18970, "dynamics_loss_avg": 0.035985720716416834}
+{"step": 18970, "action_loss_avg": 0.009873280115425587}
+{"step": 18980, "loss": 0.0658, "learning_rate": 7.108080601886003e-08}
+{"step": 18980, "dynamics_loss_avg": 0.04430606840178371}
+{"step": 18980, "action_loss_avg": 0.015798577875830232}
+{"step": 18990, "loss": 0.079, "learning_rate": 6.969847662949337e-08}
+{"step": 18990, "dynamics_loss_avg": 0.047527849674224854}
+{"step": 18990, "action_loss_avg": 0.022077369946055116}
+{"step": 19000, "loss": 0.0669, "learning_rate": 6.832962649622799e-08}
+{"step": 19000, "dynamics_loss_avg": 0.04626642623916268}
+{"step": 19000, "action_loss_avg": 0.01617791245225817}
+{"step": 19010, "loss": 0.0754, "learning_rate": 6.697425936144864e-08}
+{"step": 19010, "dynamics_loss_avg": 0.04557799994945526}
+{"step": 19010, "action_loss_avg": 0.021255811862647533}
+{"step": 19020, "loss": 0.0547, "learning_rate": 6.563237893067731e-08}
+{"step": 19020, "dynamics_loss_avg": 0.03289818889461458}
+{"step": 19020, "action_loss_avg": 0.023261255864053965}
+{"step": 19030, "loss": 0.0735, "learning_rate": 6.430398887256329e-08}
+{"step": 19030, "dynamics_loss_avg": 0.05449823662638664}
+{"step": 19030, "action_loss_avg": 0.023226954974234104}
+{"step": 19040, "loss": 0.0756, "learning_rate": 6.298909281887478e-08}
+{"step": 19040, "dynamics_loss_avg": 0.04155576010234654}
+{"step": 19040, "action_loss_avg": 0.02150076034013182}
+{"step": 19050, "loss": 0.0655, "learning_rate": 6.168769436448674e-08}
+{"step": 19050, "dynamics_loss_avg": 0.035638409107923506}
+{"step": 19050, "action_loss_avg": 0.018701795022934675}
+{"step": 19060, "loss": 0.0672, "learning_rate": 6.03997970673742e-08}
+{"step": 19060, "dynamics_loss_avg": 0.05180228874087334}
+{"step": 19060, "action_loss_avg": 0.03301493991166353}
+{"step": 19070, "loss": 0.0723, "learning_rate": 5.9125404448597825e-08}
+{"step": 19070, "dynamics_loss_avg": 0.05123816449195147}
+{"step": 19070, "action_loss_avg": 0.03283484075218439}
+{"step": 19080, "loss": 0.0619, "learning_rate": 5.786451999229836e-08}
+{"step": 19080, "dynamics_loss_avg": 0.042518378654494884}
+{"step": 19080, "action_loss_avg": 0.028800006583333017}
+{"step": 19090, "loss": 0.0646, "learning_rate": 5.6617147145687226e-08}
+{"step": 19090, "dynamics_loss_avg": 0.052104505337774755}
+{"step": 19090, "action_loss_avg": 0.03023729813285172}
+{"step": 19100, "loss": 0.065, "learning_rate": 5.538328931903259e-08}
+{"step": 19100, "dynamics_loss_avg": 0.04455666309222579}
+{"step": 19100, "action_loss_avg": 0.016629776498302818}
+{"step": 19110, "loss": 0.0656, "learning_rate": 5.416294988565551e-08}
+{"step": 19110, "dynamics_loss_avg": 0.03844494493678212}
+{"step": 19110, "action_loss_avg": 0.020590703887864947}
+{"step": 19120, "loss": 0.0788, "learning_rate": 5.29561321819172e-08}
+{"step": 19120, "dynamics_loss_avg": 0.048755235597491266}
+{"step": 19120, "action_loss_avg": 0.04096140079200268}
+{"step": 19130, "loss": 0.0564, "learning_rate": 5.176283950721062e-08}
+{"step": 19130, "dynamics_loss_avg": 0.03957191361114383}
+{"step": 19130, "action_loss_avg": 0.021234600781463088}
+{"step": 19140, "loss": 0.0569, "learning_rate": 5.058307512395333e-08}
+{"step": 19140, "dynamics_loss_avg": 0.03663741974160075}
+{"step": 19140, "action_loss_avg": 0.014955365541391074}
+{"step": 19150, "loss": 0.0559, "learning_rate": 4.941684225757526e-08}
+{"step": 19150, "dynamics_loss_avg": 0.038839674554765224}
+{"step": 19150, "action_loss_avg": 0.014020096138119697}
+{"step": 19160, "loss": 0.0697, "learning_rate": 4.8264144096513145e-08}
+{"step": 19160, "dynamics_loss_avg": 0.0577778372913599}
+{"step": 19160, "action_loss_avg": 0.023691162606701253}
+{"step": 19170, "loss": 0.0836, "learning_rate": 4.712498379219943e-08}
+{"step": 19170, "dynamics_loss_avg": 0.05882037840783596}
+{"step": 19170, "action_loss_avg": 0.03158386466093362}
+{"step": 19180, "loss": 0.0696, "learning_rate": 4.599936445905506e-08}
+{"step": 19180, "dynamics_loss_avg": 0.04895575139671564}
+{"step": 19180, "action_loss_avg": 0.018969870824366807}
+{"step": 19190, "loss": 0.0693, "learning_rate": 4.4887289174480594e-08}
+{"step": 19190, "dynamics_loss_avg": 0.04373278133571148}
+{"step": 19190, "action_loss_avg": 0.025109104276634753}
+{"step": 19200, "loss": 0.0599, "learning_rate": 4.3788760978846215e-08}
+{"step": 19200, "dynamics_loss_avg": 0.04576329002156854}
+{"step": 19200, "action_loss_avg": 0.016592811932787298}
+{"step": 19210, "loss": 0.0666, "learning_rate": 4.2703782875487265e-08}
+{"step": 19210, "dynamics_loss_avg": 0.055230416171252725}
+{"step": 19210, "action_loss_avg": 0.02290729838423431}
+{"step": 19220, "loss": 0.0637, "learning_rate": 4.1632357830692083e-08}
+{"step": 19220, "dynamics_loss_avg": 0.04439005693420768}
+{"step": 19220, "action_loss_avg": 0.024878822336904704}
+{"step": 19230, "loss": 0.0662, "learning_rate": 4.0574488773695855e-08}
+{"step": 19230, "dynamics_loss_avg": 0.05375898201018572}
+{"step": 19230, "action_loss_avg": 0.025994909834116696}
+{"step": 19240, "loss": 0.067, "learning_rate": 3.95301785966723e-08}
+{"step": 19240, "dynamics_loss_avg": 0.04041470270603895}
+{"step": 19240, "action_loss_avg": 0.017607183149084448}
+{"step": 19250, "loss": 0.0556, "learning_rate": 3.849943015472479e-08}
+{"step": 19250, "dynamics_loss_avg": 0.03672714214771986}
+{"step": 19250, "action_loss_avg": 0.019279234297573566}
+{"step": 19260, "loss": 0.0643, "learning_rate": 3.748224626588137e-08}
+{"step": 19260, "dynamics_loss_avg": 0.04293777644634247}
+{"step": 19260, "action_loss_avg": 0.016091548628173767}
+{"step": 19270, "loss": 0.0813, "learning_rate": 3.647862971108307e-08}
+{"step": 19270, "dynamics_loss_avg": 0.06033924128860235}
+{"step": 19270, "action_loss_avg": 0.02317564245313406}
+{"step": 19280, "loss": 0.0676, "learning_rate": 3.5488583234179474e-08}
+{"step": 19280, "dynamics_loss_avg": 0.04067262168973684}
+{"step": 19280, "action_loss_avg": 0.01721664038486779}
+{"step": 19290, "loss": 0.0666, "learning_rate": 3.4512109541920413e-08}
+{"step": 19290, "dynamics_loss_avg": 0.04518165625631809}
+{"step": 19290, "action_loss_avg": 0.025825847452506424}
+{"step": 19300, "loss": 0.0664, "learning_rate": 3.354921130394706e-08}
+{"step": 19300, "dynamics_loss_avg": 0.048086746409535405}
+{"step": 19300, "action_loss_avg": 0.02090142834931612}
+{"step": 19310, "loss": 0.0547, "learning_rate": 3.259989115278639e-08}
+{"step": 19310, "dynamics_loss_avg": 0.03277552807703614}
+{"step": 19310, "action_loss_avg": 0.0179377687163651}
+{"step": 19320, "loss": 0.0756, "learning_rate": 3.1664151683843404e-08}
+{"step": 19320, "dynamics_loss_avg": 0.05201920215040445}
+{"step": 19320, "action_loss_avg": 0.025618729740381242}
+{"step": 19330, "loss": 0.0757, "learning_rate": 3.0741995455394466e-08}
+{"step": 19330, "dynamics_loss_avg": 0.05810132035985589}
+{"step": 19330, "action_loss_avg": 0.02606740880291909}
+{"step": 19340, "loss": 0.0698, "learning_rate": 2.9833424988579554e-08}
+{"step": 19340, "dynamics_loss_avg": 0.04584471378475428}
+{"step": 19340, "action_loss_avg": 0.029441210301592947}
+{"step": 19350, "loss": 0.0621, "learning_rate": 2.8938442767394992e-08}
+{"step": 19350, "dynamics_loss_avg": 0.035817237198352815}
+{"step": 19350, "action_loss_avg": 0.013609715178608895}
+{"step": 19360, "loss": 0.064, "learning_rate": 2.8057051238688515e-08}
+{"step": 19360, "dynamics_loss_avg": 0.047259863931685686}
+{"step": 19360, "action_loss_avg": 0.01589830848388374}
+{"step": 19370, "loss": 0.0646, "learning_rate": 2.71892528121509e-08}
+{"step": 19370, "dynamics_loss_avg": 0.04148031026124954}
+{"step": 19370, "action_loss_avg": 0.02225602187681943}
+{"step": 19380, "loss": 0.0678, "learning_rate": 2.633504986030988e-08}
+{"step": 19380, "dynamics_loss_avg": 0.04275577375665307}
+{"step": 19380, "action_loss_avg": 0.017516118101775648}
+{"step": 19390, "loss": 0.0542, "learning_rate": 2.549444471852347e-08}
+{"step": 19390, "dynamics_loss_avg": 0.032702566031366585}
+{"step": 19390, "action_loss_avg": 0.013057024264708162}
+{"step": 19400, "loss": 0.0557, "learning_rate": 2.4667439684974424e-08}
+{"step": 19400, "dynamics_loss_avg": 0.045774967316538094}
+{"step": 19400, "action_loss_avg": 0.014352619275450706}
+{"step": 19410, "loss": 0.0616, "learning_rate": 2.3854037020662467e-08}
+{"step": 19410, "dynamics_loss_avg": 0.03391762087121606}
+{"step": 19410, "action_loss_avg": 0.02073847660794854}
+{"step": 19420, "loss": 0.0577, "learning_rate": 2.3054238949399287e-08}
+{"step": 19420, "dynamics_loss_avg": 0.04494827501475811}
+{"step": 19420, "action_loss_avg": 0.014736767392605543}
+{"step": 19430, "loss": 0.0617, "learning_rate": 2.2268047657802994e-08}
+{"step": 19430, "dynamics_loss_avg": 0.04049129285849631}
+{"step": 19430, "action_loss_avg": 0.016750149230938406}
+{"step": 19440, "loss": 0.0643, "learning_rate": 2.149546529529034e-08}
+{"step": 19440, "dynamics_loss_avg": 0.05459021460264921}
+{"step": 19440, "action_loss_avg": 0.027651762869209052}
+{"step": 19450, "loss": 0.0685, "learning_rate": 2.0736493974071736e-08}
+{"step": 19450, "dynamics_loss_avg": 0.0578347940929234}
+{"step": 19450, "action_loss_avg": 0.022148777451366185}
+{"step": 19460, "loss": 0.0734, "learning_rate": 1.9991135769145687e-08}
+{"step": 19460, "dynamics_loss_avg": 0.06578245982527733}
+{"step": 19460, "action_loss_avg": 0.025048522651195525}
+{"step": 19470, "loss": 0.0658, "learning_rate": 1.9259392718293247e-08}
+{"step": 19470, "dynamics_loss_avg": 0.04603032823652029}
+{"step": 19470, "action_loss_avg": 0.018549220450222493}
+{"step": 19480, "loss": 0.0842, "learning_rate": 1.8541266822072467e-08}
+{"step": 19480, "dynamics_loss_avg": 0.05411523971706629}
+{"step": 19480, "action_loss_avg": 0.024996600742451847}
+{"step": 19490, "loss": 0.0735, "learning_rate": 1.7836760043811187e-08}
+{"step": 19490, "dynamics_loss_avg": 0.04837729539722204}
+{"step": 19490, "action_loss_avg": 0.029992511868476866}
+{"step": 19500, "loss": 0.0775, "learning_rate": 1.7145874309604793e-08}
+{"step": 19500, "dynamics_loss_avg": 0.05609335135668516}
+{"step": 19500, "action_loss_avg": 0.03079576869495213}
+{"step": 19510, "loss": 0.0737, "learning_rate": 1.6468611508308474e-08}
+{"step": 19510, "dynamics_loss_avg": 0.038016231171786784}
+{"step": 19510, "action_loss_avg": 0.022260204795747994}
+{"step": 19520, "loss": 0.0648, "learning_rate": 1.5804973491532206e-08}
+{"step": 19520, "dynamics_loss_avg": 0.04410122968256473}
+{"step": 19520, "action_loss_avg": 0.020528348349034786}
+{"step": 19530, "loss": 0.0623, "learning_rate": 1.5154962073637425e-08}
+{"step": 19530, "dynamics_loss_avg": 0.03693989543244243}
+{"step": 19530, "action_loss_avg": 0.018374732113443315}
+{"step": 19540, "loss": 0.0586, "learning_rate": 1.4518579031730373e-08}
+{"step": 19540, "dynamics_loss_avg": 0.03409141227602959}
+{"step": 19540, "action_loss_avg": 0.020327942539006473}
+{"step": 19550, "loss": 0.0753, "learning_rate": 1.389582610565876e-08}
+{"step": 19550, "dynamics_loss_avg": 0.051672781072556975}
+{"step": 19550, "action_loss_avg": 0.029321196768432856}
+{"step": 19560, "loss": 0.0669, "learning_rate": 1.3286704998003997e-08}
+{"step": 19560, "dynamics_loss_avg": 0.049665603414177896}
+{"step": 19560, "action_loss_avg": 0.023823793698102235}
+{"step": 19570, "loss": 0.0664, "learning_rate": 1.2691217374080634e-08}
+{"step": 19570, "dynamics_loss_avg": 0.04450676534324884}
+{"step": 19570, "action_loss_avg": 0.020256917690858245}
+{"step": 19580, "loss": 0.0762, "learning_rate": 1.2109364861929707e-08}
+{"step": 19580, "dynamics_loss_avg": 0.05879650972783566}
+{"step": 19580, "action_loss_avg": 0.020828585047274827}
+{"step": 19590, "loss": 0.0646, "learning_rate": 1.1541149052312628e-08}
+{"step": 19590, "dynamics_loss_avg": 0.045380566362291576}
+{"step": 19590, "action_loss_avg": 0.019981274381279945}
+{"step": 19600, "loss": 0.0681, "learning_rate": 1.0986571498710075e-08}
+{"step": 19600, "dynamics_loss_avg": 0.05374464839696884}
+{"step": 19600, "action_loss_avg": 0.01966229798272252}
+{"step": 19610, "loss": 0.0664, "learning_rate": 1.044563371731644e-08}
+{"step": 19610, "dynamics_loss_avg": 0.04756856011226773}
+{"step": 19610, "action_loss_avg": 0.03430402267258614}
+{"step": 19620, "loss": 0.075, "learning_rate": 9.918337187034276e-09}
+{"step": 19620, "dynamics_loss_avg": 0.054150094650685786}
+{"step": 19620, "action_loss_avg": 0.02385778082534671}
+{"step": 19630, "loss": 0.0562, "learning_rate": 9.404683349472642e-09}
+{"step": 19630, "dynamics_loss_avg": 0.041935459896922114}
+{"step": 19630, "action_loss_avg": 0.02104444750584662}
+{"step": 19640, "loss": 0.0703, "learning_rate": 8.904673608940983e-09}
+{"step": 19640, "dynamics_loss_avg": 0.04481700295582414}
+{"step": 19640, "action_loss_avg": 0.025399928074330093}
+{"step": 19650, "loss": 0.0611, "learning_rate": 8.418309332447472e-09}
+{"step": 19650, "dynamics_loss_avg": 0.04475516434758901}
+{"step": 19650, "action_loss_avg": 0.02396374901290983}
+{"step": 19660, "loss": 0.0603, "learning_rate": 7.945591849692901e-09}
+{"step": 19660, "dynamics_loss_avg": 0.042971633933484556}
+{"step": 19660, "action_loss_avg": 0.022925299359485507}
+{"step": 19670, "loss": 0.0587, "learning_rate": 7.486522453069578e-09}
+{"step": 19670, "dynamics_loss_avg": 0.04090790711343288}
+{"step": 19670, "action_loss_avg": 0.018234587763436137}
+{"step": 19680, "loss": 0.0674, "learning_rate": 7.041102397655208e-09}
+{"step": 19680, "dynamics_loss_avg": 0.04668019823729992}
+{"step": 19680, "action_loss_avg": 0.01666556135751307}
+{"step": 19690, "loss": 0.0704, "learning_rate": 6.609332901210686e-09}
+{"step": 19690, "dynamics_loss_avg": 0.04854318592697382}
+{"step": 19690, "action_loss_avg": 0.01791277281008661}
+{"step": 19700, "loss": 0.0628, "learning_rate": 6.191215144178419e-09}
+{"step": 19700, "dynamics_loss_avg": 0.04074359182268381}
+{"step": 19700, "action_loss_avg": 0.022232988104224204}
+{"step": 19710, "loss": 0.0769, "learning_rate": 5.7867502696756785e-09}
+{"step": 19710, "dynamics_loss_avg": 0.05371684115380049}
+{"step": 19710, "action_loss_avg": 0.028047747211530805}
+{"step": 19720, "loss": 0.0738, "learning_rate": 5.395939383494031e-09}
+{"step": 19720, "dynamics_loss_avg": 0.05972814448177814}
+{"step": 19720, "action_loss_avg": 0.020521515607833864}
+{"step": 19730, "loss": 0.0684, "learning_rate": 5.018783554095464e-09}
+{"step": 19730, "dynamics_loss_avg": 0.04780337158590555}
+{"step": 19730, "action_loss_avg": 0.01911618495360017}
+{"step": 19740, "loss": 0.0633, "learning_rate": 4.655283812610156e-09}
+{"step": 19740, "dynamics_loss_avg": 0.04706993810832501}
+{"step": 19740, "action_loss_avg": 0.02198672341182828}
+{"step": 19750, "loss": 0.0565, "learning_rate": 4.305441152831491e-09}
+{"step": 19750, "dynamics_loss_avg": 0.03720401818864048}
+{"step": 19750, "action_loss_avg": 0.01753886609803885}
+{"step": 19760, "loss": 0.0662, "learning_rate": 3.969256531217158e-09}
+{"step": 19760, "dynamics_loss_avg": 0.045891742780804635}
+{"step": 19760, "action_loss_avg": 0.017851687874644993}
+{"step": 19770, "loss": 0.0505, "learning_rate": 3.6467308668824975e-09}
+{"step": 19770, "dynamics_loss_avg": 0.029992605652660133}
+{"step": 19770, "action_loss_avg": 0.01609015348367393}
+{"step": 19780, "loss": 0.0535, "learning_rate": 3.3378650416004965e-09}
+{"step": 19780, "dynamics_loss_avg": 0.030971918720752}
+{"step": 19780, "action_loss_avg": 0.012941750744357706}
+{"step": 19790, "loss": 0.0712, "learning_rate": 3.042659899797906e-09}
+{"step": 19790, "dynamics_loss_avg": 0.05431404039263725}
+{"step": 19790, "action_loss_avg": 0.03052880633622408}
+{"step": 19800, "loss": 0.0776, "learning_rate": 2.7611162485552403e-09}
+{"step": 19800, "dynamics_loss_avg": 0.045865561440587045}
+{"step": 19800, "action_loss_avg": 0.026190615445375442}
+{"step": 19810, "loss": 0.0754, "learning_rate": 2.4932348576017782e-09}
+{"step": 19810, "dynamics_loss_avg": 0.05829719081521034}
+{"step": 19810, "action_loss_avg": 0.025177312176674606}
+{"step": 19820, "loss": 0.0661, "learning_rate": 2.2390164593144583e-09}
+{"step": 19820, "dynamics_loss_avg": 0.04026942318305373}
+{"step": 19820, "action_loss_avg": 0.018375073047354817}
+{"step": 19830, "loss": 0.0754, "learning_rate": 1.9984617487173174e-09}
+{"step": 19830, "dynamics_loss_avg": 0.05735848154872656}
+{"step": 19830, "action_loss_avg": 0.029920194949954747}
+{"step": 19840, "loss": 0.0632, "learning_rate": 1.7715713834776105e-09}
+{"step": 19840, "dynamics_loss_avg": 0.034946164861321447}
+{"step": 19840, "action_loss_avg": 0.020855458034202457}
+{"step": 19850, "loss": 0.063, "learning_rate": 1.5583459839046965e-09}
+{"step": 19850, "dynamics_loss_avg": 0.04919734569266439}
+{"step": 19850, "action_loss_avg": 0.02176156654022634}
+{"step": 19860, "loss": 0.0642, "learning_rate": 1.3587861329489305e-09}
+{"step": 19860, "dynamics_loss_avg": 0.030154460575431585}
+{"step": 19860, "action_loss_avg": 0.01854857854777947}
+{"step": 19870, "loss": 0.064, "learning_rate": 1.1728923761994415e-09}
+{"step": 19870, "dynamics_loss_avg": 0.045471681468188765}
+{"step": 19870, "action_loss_avg": 0.02216670736670494}
+{"step": 19880, "loss": 0.0718, "learning_rate": 1.0006652218819135e-09}
+{"step": 19880, "dynamics_loss_avg": 0.05042210947722196}
+{"step": 19880, "action_loss_avg": 0.019688624422997236}
+{"step": 19890, "loss": 0.0519, "learning_rate": 8.421051408596948e-10}
+{"step": 19890, "dynamics_loss_avg": 0.023838566755875944}
+{"step": 19890, "action_loss_avg": 0.010728829819709063}
+{"step": 19900, "loss": 0.0653, "learning_rate": 6.972125666299123e-10}
+{"step": 19900, "dynamics_loss_avg": 0.042154907062649724}
+{"step": 19900, "action_loss_avg": 0.022989201312884688}
+{"step": 19910, "loss": 0.068, "learning_rate": 5.65987895322917e-10}
+{"step": 19910, "dynamics_loss_avg": 0.053467546589672565}
+{"step": 19910, "action_loss_avg": 0.026885805232450367}
+{"step": 19920, "loss": 0.0676, "learning_rate": 4.48431485701728e-10}
+{"step": 19920, "dynamics_loss_avg": 0.04784501497633755}
+{"step": 19920, "action_loss_avg": 0.021005964279174803}
+{"step": 19930, "loss": 0.0628, "learning_rate": 3.4454365916203326e-10}
+{"step": 19930, "dynamics_loss_avg": 0.04858131390064955}
+{"step": 19930, "action_loss_avg": 0.02204345481004566}
+{"step": 19940, "loss": 0.0743, "learning_rate": 2.5432469972830333e-10}
+{"step": 19940, "dynamics_loss_avg": 0.0493664912879467}
+{"step": 19940, "action_loss_avg": 0.016600931715220212}
+{"step": 19950, "loss": 0.0604, "learning_rate": 1.77774854056012e-10}
+{"step": 19950, "dynamics_loss_avg": 0.04475926365703344}
+{"step": 19950, "action_loss_avg": 0.02512040608562529}
+{"step": 19960, "loss": 0.0679, "learning_rate": 1.1489433142941597e-10}
+{"step": 19960, "dynamics_loss_avg": 0.04144583009183407}
+{"step": 19960, "action_loss_avg": 0.015122923068702221}
+{"step": 19970, "loss": 0.0763, "learning_rate": 6.568330376210964e-11}
+{"step": 19970, "dynamics_loss_avg": 0.0645992343313992}
+{"step": 19970, "action_loss_avg": 0.02953599998727441}
+{"step": 19980, "loss": 0.0735, "learning_rate": 3.014190559424979e-11}
+{"step": 19980, "dynamics_loss_avg": 0.051318105403333904}
+{"step": 19980, "action_loss_avg": 0.024603950278833508}
+{"step": 19990, "loss": 0.0758, "learning_rate": 8.270234094776009e-12}
+{"step": 19990, "dynamics_loss_avg": 0.047976726107299326}
+{"step": 19990, "action_loss_avg": 0.02277223765850067}
+{"step": 20000, "loss": 0.0696, "learning_rate": 6.834906085551041e-14}

dreamzero_real_teleop_g1_full_finetune_absolute_Pick_bottle_and_turn_and_pour_into_cup/wandb/debug-internal.log

@@ -5,3 +5,37 @@
 {"time":"2026-03-30T04:15:10.413395221Z","level":"INFO","msg":"writer: started","stream_id":"ao6lob27"}
 {"time":"2026-03-30T04:15:10.413394431Z","level":"INFO","msg":"sender: started","stream_id":"ao6lob27"}
 {"time":"2026-03-30T22:30:11.007336848Z","level":"INFO","msg":"api: retrying HTTP error","status":502,"url":"https://api.wandb.ai/files/boqianli/dreamzero/ao6lob27/file_stream","body":"\n<html><head>\n<meta http-equiv=\"content-type\" content=\"text/html;charset=utf-8\">\n<title>502 Server Error</title>\n</head>\n<body text=#000000 bgcolor=#ffffff>\n<h1>Error: Server Error</h1>\n<h2>The server encountered a temporary error and could not complete your request.<p>Please try again in 30 seconds.</h2>\n<h2></h2>\n</body></html>\n"}
+{"time":"2026-04-01T10:18:44.020890295Z","level":"INFO","msg":"api: retrying error","error":"Post \"https://api.wandb.ai/files/boqianli/dreamzero/ao6lob27/file_stream\": read tcp 192.168.0.3:41808->35.186.228.49:443: read: connection timed out"}
+{"time":"2026-04-01T10:18:56.087878557Z","level":"INFO","msg":"api: retrying error","error":"Post \"https://api.wandb.ai/files/boqianli/dreamzero/ao6lob27/file_stream\": dial tcp: lookup api.wandb.ai on 127.0.0.53:53: read udp 127.0.0.1:38712->127.0.0.53:53: i/o timeout"}
+{"time":"2026-04-01T10:19:10.908212851Z","level":"INFO","msg":"api: retrying error","error":"Post \"https://api.wandb.ai/files/boqianli/dreamzero/ao6lob27/file_stream\": dial tcp: lookup api.wandb.ai on 127.0.0.53:53: read udp 127.0.0.1:39469->127.0.0.53:53: i/o timeout"}
+{"time":"2026-04-01T10:19:30.190045118Z","level":"INFO","msg":"api: retrying error","error":"Post \"https://api.wandb.ai/files/boqianli/dreamzero/ao6lob27/file_stream\": dial tcp: lookup api.wandb.ai on 127.0.0.53:53: read udp 127.0.0.1:47884->127.0.0.53:53: i/o timeout"}
+{"time":"2026-04-01T10:19:58.216957187Z","level":"INFO","msg":"api: retrying error","error":"Post \"https://api.wandb.ai/files/boqianli/dreamzero/ao6lob27/file_stream\": dial tcp: lookup api.wandb.ai on 127.0.0.53:53: read udp 127.0.0.1:34560->127.0.0.53:53: i/o timeout"}
+{"time":"2026-04-01T10:20:40.445735516Z","level":"INFO","msg":"api: retrying error","error":"Post \"https://api.wandb.ai/files/boqianli/dreamzero/ao6lob27/file_stream\": dial tcp: lookup api.wandb.ai on 127.0.0.53:53: read udp 127.0.0.1:46537->127.0.0.53:53: i/o timeout"}
+{"time":"2026-04-01T10:21:50.450660421Z","level":"INFO","msg":"api: retrying error","error":"Post \"https://api.wandb.ai/files/boqianli/dreamzero/ao6lob27/file_stream\": dial tcp: lookup api.wandb.ai on 127.0.0.53:53: read udp 127.0.0.1:39132->127.0.0.53:53: i/o timeout"}
+{"time":"2026-04-01T10:23:00.454344979Z","level":"INFO","msg":"api: retrying error","error":"Post \"https://api.wandb.ai/files/boqianli/dreamzero/ao6lob27/file_stream\": dial tcp: lookup api.wandb.ai on 127.0.0.53:53: read udp 127.0.0.1:60808->127.0.0.53:53: i/o timeout"}
+{"time":"2026-04-01T10:24:10.457186285Z","level":"INFO","msg":"api: retrying error","error":"Post \"https://api.wandb.ai/files/boqianli/dreamzero/ao6lob27/file_stream\": dial tcp: lookup api.wandb.ai on 127.0.0.53:53: read udp 127.0.0.1:42899->127.0.0.53:53: i/o timeout"}
+{"time":"2026-04-01T10:25:20.462333271Z","level":"INFO","msg":"api: retrying error","error":"Post \"https://api.wandb.ai/files/boqianli/dreamzero/ao6lob27/file_stream\": dial tcp: lookup api.wandb.ai on 127.0.0.53:53: read udp 127.0.0.1:51640->127.0.0.53:53: i/o timeout"}
+{"time":"2026-04-01T10:26:30.466228733Z","level":"INFO","msg":"api: retrying error","error":"Post \"https://api.wandb.ai/files/boqianli/dreamzero/ao6lob27/file_stream\": dial tcp: lookup api.wandb.ai on 127.0.0.53:53: read udp 127.0.0.1:54385->127.0.0.53:53: i/o timeout"}
+{"time":"2026-04-01T10:27:40.469992011Z","level":"INFO","msg":"api: retrying error","error":"Post \"https://api.wandb.ai/files/boqianli/dreamzero/ao6lob27/file_stream\": dial tcp: lookup api.wandb.ai on 127.0.0.53:53: read udp 127.0.0.1:36994->127.0.0.53:53: i/o timeout"}
+{"time":"2026-04-01T10:28:50.474794045Z","level":"INFO","msg":"api: retrying error","error":"Post \"https://api.wandb.ai/files/boqianli/dreamzero/ao6lob27/file_stream\": dial tcp: lookup api.wandb.ai on 127.0.0.53:53: read udp 127.0.0.1:59340->127.0.0.53:53: i/o timeout"}
+{"time":"2026-04-01T10:30:00.479410672Z","level":"INFO","msg":"api: retrying error","error":"Post \"https://api.wandb.ai/files/boqianli/dreamzero/ao6lob27/file_stream\": dial tcp: lookup api.wandb.ai on 127.0.0.53:53: read udp 127.0.0.1:53914->127.0.0.53:53: i/o timeout"}
+{"time":"2026-04-01T10:31:10.483284091Z","level":"INFO","msg":"api: retrying error","error":"Post \"https://api.wandb.ai/files/boqianli/dreamzero/ao6lob27/file_stream\": dial tcp: lookup api.wandb.ai on 127.0.0.53:53: read udp 127.0.0.1:52683->127.0.0.53:53: i/o timeout"}
+{"time":"2026-04-01T10:32:20.487788622Z","level":"INFO","msg":"api: retrying error","error":"Post \"https://api.wandb.ai/files/boqianli/dreamzero/ao6lob27/file_stream\": dial tcp: lookup api.wandb.ai on 127.0.0.53:53: read udp 127.0.0.1:52313->127.0.0.53:53: i/o timeout"}
+{"time":"2026-04-01T10:33:30.492784858Z","level":"INFO","msg":"api: retrying error","error":"Post \"https://api.wandb.ai/files/boqianli/dreamzero/ao6lob27/file_stream\": dial tcp: lookup api.wandb.ai on 127.0.0.53:53: read udp 127.0.0.1:38331->127.0.0.53:53: i/o timeout"}
+{"time":"2026-04-01T10:34:40.497056663Z","level":"INFO","msg":"api: retrying error","error":"Post \"https://api.wandb.ai/files/boqianli/dreamzero/ao6lob27/file_stream\": dial tcp: lookup api.wandb.ai on 127.0.0.53:53: read udp 127.0.0.1:59696->127.0.0.53:53: i/o timeout"}
+{"time":"2026-04-01T10:35:50.500859105Z","level":"INFO","msg":"api: retrying error","error":"Post \"https://api.wandb.ai/files/boqianli/dreamzero/ao6lob27/file_stream\": dial tcp: lookup api.wandb.ai on 127.0.0.53:53: read udp 127.0.0.1:40973->127.0.0.53:53: i/o timeout"}
+{"time":"2026-04-01T10:37:00.5054549Z","level":"INFO","msg":"api: retrying error","error":"Post \"https://api.wandb.ai/files/boqianli/dreamzero/ao6lob27/file_stream\": dial tcp: lookup api.wandb.ai on 127.0.0.53:53: read udp 127.0.0.1:45222->127.0.0.53:53: i/o timeout"}
+{"time":"2026-04-01T10:38:10.508535232Z","level":"INFO","msg":"api: retrying error","error":"Post \"https://api.wandb.ai/files/boqianli/dreamzero/ao6lob27/file_stream\": dial tcp: lookup api.wandb.ai on 127.0.0.53:53: read udp 127.0.0.1:55362->127.0.0.53:53: i/o timeout"}
+{"time":"2026-04-01T10:39:20.512300509Z","level":"INFO","msg":"api: retrying error","error":"Post \"https://api.wandb.ai/files/boqianli/dreamzero/ao6lob27/file_stream\": dial tcp: lookup api.wandb.ai on 127.0.0.53:53: read udp 127.0.0.1:50089->127.0.0.53:53: i/o timeout"}
+{"time":"2026-04-01T10:40:30.517197984Z","level":"INFO","msg":"api: retrying error","error":"Post \"https://api.wandb.ai/files/boqianli/dreamzero/ao6lob27/file_stream\": dial tcp: lookup api.wandb.ai on 127.0.0.53:53: read udp 127.0.0.1:34388->127.0.0.53:53: i/o timeout"}
+{"time":"2026-04-01T10:41:40.521767935Z","level":"INFO","msg":"api: retrying error","error":"Post \"https://api.wandb.ai/files/boqianli/dreamzero/ao6lob27/file_stream\": dial tcp: lookup api.wandb.ai on 127.0.0.53:53: read udp 127.0.0.1:53489->127.0.0.53:53: i/o timeout"}
+{"time":"2026-04-01T10:42:50.525040035Z","level":"INFO","msg":"api: retrying error","error":"Post \"https://api.wandb.ai/files/boqianli/dreamzero/ao6lob27/file_stream\": dial tcp: lookup api.wandb.ai on 127.0.0.53:53: read udp 127.0.0.1:50726->127.0.0.53:53: i/o timeout"}
+{"time":"2026-04-01T10:44:00.527917425Z","level":"INFO","msg":"api: retrying error","error":"Post \"https://api.wandb.ai/files/boqianli/dreamzero/ao6lob27/file_stream\": dial tcp: lookup api.wandb.ai on 127.0.0.53:53: read udp 127.0.0.1:53434->127.0.0.53:53: i/o timeout"}
+{"time":"2026-04-01T10:45:10.532222422Z","level":"INFO","msg":"api: retrying error","error":"Post \"https://api.wandb.ai/files/boqianli/dreamzero/ao6lob27/file_stream\": dial tcp: lookup api.wandb.ai on 127.0.0.53:53: read udp 127.0.0.1:37341->127.0.0.53:53: i/o timeout"}
+{"time":"2026-04-01T10:46:20.535679123Z","level":"INFO","msg":"api: retrying error","error":"Post \"https://api.wandb.ai/files/boqianli/dreamzero/ao6lob27/file_stream\": dial tcp: lookup api.wandb.ai on 127.0.0.53:53: read udp 127.0.0.1:37921->127.0.0.53:53: i/o timeout"}
+{"time":"2026-04-01T10:47:30.539549836Z","level":"INFO","msg":"api: retrying error","error":"Post \"https://api.wandb.ai/files/boqianli/dreamzero/ao6lob27/file_stream\": dial tcp: lookup api.wandb.ai on 127.0.0.53:53: read udp 127.0.0.1:38214->127.0.0.53:53: i/o timeout"}
+{"time":"2026-04-01T14:24:28.755750998Z","level":"INFO","msg":"stream: closing","id":"ao6lob27"}
+{"time":"2026-04-01T14:24:29.342449946Z","level":"INFO","msg":"fileTransfer: Close: file transfer manager closed"}
+{"time":"2026-04-01T14:24:29.828344988Z","level":"INFO","msg":"handler: closed","stream_id":"ao6lob27"}
+{"time":"2026-04-01T14:24:29.82853025Z","level":"INFO","msg":"sender: closed","stream_id":"ao6lob27"}
+{"time":"2026-04-01T14:24:29.8285597Z","level":"INFO","msg":"stream: closed","id":"ao6lob27"}

dreamzero_real_teleop_g1_full_finetune_absolute_Pick_bottle_and_turn_and_pour_into_cup/wandb/debug.log

@@ -20,3 +20,5 @@ config: {'_wandb': {}}
 2026-03-30 04:15:10,932 INFO    MainThread:977089 [wandb_run.py:_config_callback():1403] config_cb None None {'return_dict': True, 'output_hidden_states': False, 'output_attentions': False, 'torchscript': False, 'torch_dtype': None, 'use_bfloat16': False, 'tf_legacy_loss': False, 'pruned_heads': {}, 'tie_word_embeddings': True, 'chunk_size_feed_forward': 0, 'is_encoder_decoder': False, 'is_decoder': False, 'cross_attention_hidden_size': None, 'add_cross_attention': False, 'tie_encoder_decoder': False, 'max_length': 20, 'min_length': 0, 'do_sample': False, 'early_stopping': False, 'num_beams': 1, 'num_beam_groups': 1, 'diversity_penalty': 0.0, 'temperature': 1.0, 'top_k': 50, 'top_p': 1.0, 'typical_p': 1.0, 'repetition_penalty': 1.0, 'length_penalty': 1.0, 'no_repeat_ngram_size': 0, 'encoder_no_repeat_ngram_size': 0, 'bad_words_ids': None, 'num_return_sequences': 1, 'output_scores': False, 'return_dict_in_generate': False, 'forced_bos_token_id': None, 'forced_eos_token_id': None, 'remove_invalid_values': False, 'exponential_decay_length_penalty': None, 'suppress_tokens': None, 'begin_suppress_tokens': None, 'architectures': None, 'finetuning_task': None, 'id2label': {0: 'LABEL_0', 1: 'LABEL_1'}, 'label2id': {'LABEL_0': 0, 'LABEL_1': 1}, 'tokenizer_class': None, 'prefix': None, 'bos_token_id': None, 'pad_token_id': None, 'eos_token_id': None, 'sep_token_id': None, 'decoder_start_token_id': None, 'task_specific_params': None, 'problem_type': None, '_name_or_path': './checkpoints/dreamzero_real_teleop_g1_full_finetune_absolute', '_attn_implementation_autoset': True, 'transformers_version': '4.51.3', 'model_dtype': 'float32', 'hidden_size': 0, 'action_horizon': 24, 'action_dim': 36, 'backbone_cfg': {'_target_': 'groot.vla.model.dreamzero.backbone.identity.IdentityBackbone'}, 'action_head_cfg': {'config': {'backbone_features_projector_cfg': None, '_target_': 'groot.vla.model.dreamzero.action_head.wan_flow_matching_action_tf.WANPolicyHeadConfig', '_recursive_': False, 'tiled': False, 'tile_size_height': 34, 'tile_size_width': 34, 'tile_stride_height': 18, 'tile_stride_width': 16, 'lora_rank': 4, 'lora_alpha': 4, 'num_frames': 33, 'num_frame_per_block': 2, 'lora_target_modules': 'q,k,v,o,ffn.0,ffn.2', 'init_lora_weights': 'kaiming', 'train_architecture': 'full', 'use_gradient_checkpointing': True, 'add_pos_embed': True, 'model_dtype': 'float32', 'max_state_dim': 64, 'max_action_dim': 36, 'action_loss_embodiment_ids': [26, 17, 32], 'hidden_size': 64, 'input_embedding_dim': 1536, 'backbone_embedding_dim': 0, 'repa_layer': 8, 'repa_coeff': 1.0, 'load_pretrained_det_decode_layer_path': None, 'freeze_decode_layer': False, 'expand_batch': None, 'use_vlln': True, 'vl_self_attention_cfg': {'_target_': 'groot.vla.model.n1_5.modules.cross_attention_dit.SelfAttentionTransformer', 'positional_embeddings': None, 'num_layers': 4, 'num_attention_heads': 24, 'attention_head_dim': 64, 'dropout': 0.2, 'final_dropout': True}, 'diffusion_model_cfg': {'_target_': 'groot.vla.model.dreamzero.modules.wan_video_dit_action_casual_chunk.CausalWanModel', '_convert_': 'object', 'diffusion_model_pretrained_path': '/hfm/boqian/liboqian_data/checkpoints/Wan2.1-I2V-14B-480P', 'model_type': 'i2v', 'frame_seqlen': 220, 'dim': 5120, 'in_dim': 36, 'ffn_dim': 13824, 'out_dim': 16, 'freq_dim': 256, 'eps': 1e-06, 'num_heads': 40, 'num_layers': 40, 'max_chunk_size': 4, 'num_frame_per_block': 2, 'num_action_per_block': 24, 'num_state_per_block': 1, 'action_dim': 36}, 'text_encoder_cfg': {'_target_': 'groot.vla.model.dreamzero.modules.wan_video_text_encoder.WanTextEncoder', '_convert_': 'object', 'text_encoder_pretrained_path': '/hfm/boqian/liboqian_data/checkpoints/Wan2.1-I2V-14B-480P/models_t5_umt5-xxl-enc-bf16.pth'}, 'image_encoder_cfg': {'_target_': 'groot.vla.model.dreamzero.modules.wan_video_image_encoder.WanImageEncoder', '_convert_': 'object', 'image_encoder_pretrained_path': '/hfm/boqian/liboqian_data/checkpoints/Wan2.1-I2V-14B-480P/models_clip_open-clip-xlm-roberta-large-vit-huge-14.pth'}, 'vae_cfg': {'_target_': 'groot.vla.model.dreamzero.modules.wan_video_vae.WanVideoVAE', '_convert_': 'object', 'vae_pretrained_path': '/hfm/boqian/liboqian_data/checkpoints/Wan2.1-I2V-14B-480P/Wan2.1_VAE.pth'}, 'action_dim': 36, 'action_horizon': 24, 'num_inference_timesteps': 4, 'noise_beta_alpha': 1.5, 'noise_beta_beta': 1.0, 'noise_s': 0.999, 'num_timestep_buckets': 1000, 'decouple_video_action_noise': False, 'video_noise_beta_alpha': 3.0, 'video_noise_beta_beta': 1.0, 'tune_projector': True, 'tune_diffusion_model': True, 'skip_component_loading': True}, '_target_': 'groot.vla.model.dreamzero.action_head.wan_flow_matching_action_tf.WANPolicyHead', '_convert_': 'object'}, 'resume_path': './checkpoints/dreamzero_real_teleop_g1_full_finetune_absolute', 'model_type': 'vla', 'output_dir': './checkpoints/dreamzero_real_teleop_g1_full_finetune_absolute', 'overwrite_output_dir': False, 'do_train': False, 'do_eval': False, 'do_predict': False, 'eval_strategy': 'no', 'prediction_loss_only': False, 'per_device_train_batch_size': 1, 'per_device_eval_batch_size': 64, 'per_gpu_train_batch_size': None, 'per_gpu_eval_batch_size': None, 'gradient_accumulation_steps': 1, 'eval_accumulation_steps': None, 'eval_delay': 0, 'torch_empty_cache_steps': None, 'learning_rate': 1e-05, 'weight_decay': 1e-05, 'adam_beta1': 0.95, 'adam_beta2': 0.999, 'adam_epsilon': 1e-08, 'max_grad_norm': 1.0, 'num_train_epochs': 1000, 'max_steps': 20000, 'lr_scheduler_type': 'cosine', 'lr_scheduler_kwargs': {}, 'warmup_ratio': 0.05, 'warmup_steps': 0, 'log_level': 'passive', 'log_level_replica': 'warning', 'log_on_each_node': True, 'logging_dir': './checkpoints/dreamzero_real_teleop_g1_full_finetune_absolute/runs/Mar30_04-04-13_nebula100', 'logging_strategy': 'steps', 'logging_first_step': False, 'logging_steps': 10, 'logging_nan_inf_filter': True, 'save_strategy': 'steps', 'save_steps': 8000, 'save_total_limit': 10, 'save_safetensors': True, 'save_on_each_node': False, 'save_only_model': False, 'restore_callback_states_from_checkpoint': False, 'no_cuda': False, 'use_cpu': False, 'use_mps_device': False, 'seed': 42, 'data_seed': None, 'jit_mode_eval': False, 'use_ipex': False, 'bf16': True, 'fp16': False, 'fp16_opt_level': 'O1', 'half_precision_backend': 'auto', 'bf16_full_eval': False, 'fp16_full_eval': False, 'tf32': True, 'local_rank': 0, 'ddp_backend': None, 'tpu_num_cores': None, 'tpu_metrics_debug': False, 'debug': [], 'dataloader_drop_last': False, 'eval_steps': None, 'dataloader_num_workers': 1, 'dataloader_prefetch_factor': None, 'past_index': -1, 'run_name': 'dreamzero_real_teleop_g1_full_finetune_absolute', 'disable_tqdm': False, 'remove_unused_columns': False, 'label_names': None, 'load_best_model_at_end': False, 'metric_for_best_model': None, 'greater_is_better': None, 'ignore_data_skip': True, 'fsdp': [], 'fsdp_min_num_params': 0, 'fsdp_config': {'min_num_params': 0, 'xla': False, 'xla_fsdp_v2': False, 'xla_fsdp_grad_ckpt': False}, 'tp_size': 0, 'fsdp_transformer_layer_cls_to_wrap': None, 'accelerator_config': {'split_batches': False, 'dispatch_batches': None, 'even_batches': True, 'use_seedable_sampler': True, 'non_blocking': False, 'gradient_accumulation_kwargs': None}, 'deepspeed': 'groot/vla/configs/deepspeed/zero2_offload.json', 'label_smoothing_factor': 0.0, 'optim': 'adamw_torch', 'optim_args': None, 'adafactor': False, 'group_by_length': False, 'length_column_name': 'length', 'report_to': ['wandb'], 'ddp_find_unused_parameters': False, 'ddp_bucket_cap_mb': 100, 'ddp_broadcast_buffers': None, 'dataloader_pin_memory': False, 'dataloader_persistent_workers': True, 'skip_memory_metrics': True, 'use_legacy_prediction_loop': False, 'push_to_hub': False, 'resume_from_checkpoint': None, 'hub_model_id': None, 'hub_strategy': 'every_save', 'hub_token': '<HUB_TOKEN>', 'hub_private_repo': None, 'hub_always_push': False, 'gradient_checkpointing': False, 'gradient_checkpointing_kwargs': None, 'include_inputs_for_metrics': False, 'include_for_metrics': [], 'eval_do_concat_batches': True, 'fp16_backend': 'auto', 'push_to_hub_model_id': None, 'push_to_hub_organization': None, 'push_to_hub_token': '<PUSH_TO_HUB_TOKEN>', 'mp_parameters': '', 'auto_find_batch_size': False, 'full_determinism': False, 'torchdynamo': None, 'ray_scope': 'last', 'ddp_timeout': 1800, 'torch_compile': False, 'torch_compile_backend': None, 'torch_compile_mode': None, 'include_tokens_per_second': False, 'include_num_input_tokens_seen': False, 'neftune_noise_alpha': None, 'optim_target_modules': None, 'batch_eval_metrics': False, 'eval_on_start': False, 'use_liger_kernel': False, 'eval_use_gather_object': False, 'average_tokens_across_devices': False}
 2026-03-30 04:15:10,939 INFO    MainThread:977089 [wandb_config.py:__setitem__():155] [no run ID] config set model/num_parameters = 22924196696 - <bound method Run._config_callback of <wandb.sdk.wandb_run.Run object at 0x7b9f0d23f190>>
 2026-03-30 04:15:10,939 INFO    MainThread:977089 [wandb_run.py:_config_callback():1403] config_cb model/num_parameters 22924196696 None
+2026-04-01 14:24:28,737 INFO    wandb-AsyncioManager-main:977089 [service_client.py:_forward_responses():134] Reached EOF.
+2026-04-01 14:24:28,738 INFO    wandb-AsyncioManager-main:977089 [mailbox.py:close():155] Closing mailbox, abandoning 1 handles.

dreamzero_real_teleop_g1_full_finetune_absolute_Pick_bottle_and_turn_and_pour_into_cup/wandb/run-20260330_041509-ao6lob27/files/config.yaml

@@ -0,0 +1,651 @@
+_attn_implementation_autoset:
+    value: true
+_name_or_path:
+    value: ./checkpoints/dreamzero_real_teleop_g1_full_finetune_absolute
+_wandb:
+    value:
+        cli_version: 0.25.0
+        e:
+            ppsci3bha2i61tuvdhjs2d9rc3ipwu77:
+                args:
+                    - report_to=wandb
+                    - data=dreamzero/real_teleop_g1_relative
+                    - wandb_project=dreamzero
+                    - train_architecture=full
+                    - num_frames=33
+                    - action_horizon=24
+                    - num_views=1
+                    - model=dreamzero/vla
+                    - model/dreamzero/action_head=wan_flow_matching_action_tf
+                    - model/dreamzero/transform=dreamzero_cotrain
+                    - num_frame_per_block=2
+                    - num_action_per_block=24
+                    - num_state_per_block=1
+                    - seed=42
+                    - training_args.learning_rate=1e-5
+                    - training_args.deepspeed=groot/vla/configs/deepspeed/zero2_offload.json
+                    - save_steps=8000
+                    - training_args.warmup_ratio=0.05
+                    - output_dir=./checkpoints/dreamzero_real_teleop_g1_full_finetune_absolute
+                    - per_device_train_batch_size=1
+                    - max_steps=20000
+                    - weight_decay=1e-5
+                    - save_total_limit=10
+                    - upload_checkpoints=false
+                    - bf16=true
+                    - tf32=true
+                    - eval_bf16=true
+                    - dataloader_pin_memory=false
+                    - dataloader_num_workers=1
+                    - image_resolution_width=320
+                    - image_resolution_height=176
+                    - save_lora_only=false
+                    - max_chunk_size=4
+                    - frame_seqlen=220
+                    - save_strategy=steps
+                    - real_teleop_g1_data_root=/hfm/boqian/liboqian_data/data/real_data/gear/g1/Pick_bottle_and_turn_and_pour_into_cup_absolute
+                    - dit_version=/hfm/boqian/liboqian_data/checkpoints/Wan2.1-I2V-14B-480P
+                    - text_encoder_pretrained_path=/hfm/boqian/liboqian_data/checkpoints/Wan2.1-I2V-14B-480P/models_t5_umt5-xxl-enc-bf16.pth
+                    - image_encoder_pretrained_path=/hfm/boqian/liboqian_data/checkpoints/Wan2.1-I2V-14B-480P/models_clip_open-clip-xlm-roberta-large-vit-huge-14.pth
+                    - vae_pretrained_path=/hfm/boqian/liboqian_data/checkpoints/Wan2.1-I2V-14B-480P/Wan2.1_VAE.pth
+                    - tokenizer_path=/hfm/boqian/liboqian_data/checkpoints/umt5-xxl
+                    - pretrained_model_path=/hfm/boqian/liboqian_data/checkpoints/DreamZero-AgiBot
+                    - max_action_dim=36
+                    - ++action_head_cfg.config.skip_component_loading=true
+                codePath: groot/vla/experiment/experiment.py
+                codePathLocal: groot/vla/experiment/experiment.py
+                cpu_count: 128
+                cpu_count_logical: 128
+                cudaVersion: "12.9"
+                disk:
+                    /:
+                        total: "3934171283456"
+                        used: "3686894223360"
+                email: 2074894208@qq.com
+                executable: /hfm/boqian/miniconda3/envs/dreamzero/bin/python3.11
+                git:
+                    commit: 21d1fe7f867c819a9feb12ecbfba6d0cd933d126
+                    remote: git@github.com:boqian-li/DreamZero-private.git
+                gpu: NVIDIA A100-SXM4-80GB
+                gpu_count: 8
+                gpu_nvidia:
+                    - architecture: Ampere
+                      cudaCores: 6912
+                      memoryTotal: "85899345920"
+                      name: NVIDIA A100-SXM4-80GB
+                      uuid: GPU-95155859-2785-a5fa-d493-0c9e9eb3332f
+                    - architecture: Ampere
+                      cudaCores: 6912
+                      memoryTotal: "85899345920"
+                      name: NVIDIA A100-SXM4-80GB
+                      uuid: GPU-3829db0c-5c30-974b-5c88-9ed4f984eddf
+                    - architecture: Ampere
+                      cudaCores: 6912
+                      memoryTotal: "85899345920"
+                      name: NVIDIA A100-SXM4-80GB
+                      uuid: GPU-0e260d0d-d121-6d8c-4c97-6ca0dfce506f
+                    - architecture: Ampere
+                      cudaCores: 6912
+                      memoryTotal: "85899345920"
+                      name: NVIDIA A100-SXM4-80GB
+                      uuid: GPU-b99492cb-2347-0f80-e94f-a95e105eaffc
+                    - architecture: Ampere
+                      cudaCores: 6912
+                      memoryTotal: "85899345920"
+                      name: NVIDIA A100-SXM4-80GB
+                      uuid: GPU-8a143efc-044d-2012-5820-afb258793ffd
+                    - architecture: Ampere
+                      cudaCores: 6912
+                      memoryTotal: "85899345920"
+                      name: NVIDIA A100-SXM4-80GB
+                      uuid: GPU-9fedd571-c6c4-fd5a-1dd7-f82e048386ad
+                    - architecture: Ampere
+                      cudaCores: 6912
+                      memoryTotal: "85899345920"
+                      name: NVIDIA A100-SXM4-80GB
+                      uuid: GPU-8cd44ee0-3344-490d-bb51-ba5b38f2b924
+                    - architecture: Ampere
+                      cudaCores: 6912
+                      memoryTotal: "85899345920"
+                      name: NVIDIA A100-SXM4-80GB
+                      uuid: GPU-55a7cb99-494a-2d02-b9d8-6f85a0614e8c
+                host: nebula100
+                memory:
+                    total: "1082011422720"
+                os: Linux-6.8.0-100-generic-x86_64-with-glibc2.39
+                program: /hfm/boqian/liboqian_code/DreamZero-private/groot/vla/experiment/experiment.py
+                python: CPython 3.11.14
+                root: ./checkpoints/dreamzero_real_teleop_g1_full_finetune_absolute
+                startedAt: "2026-03-30T04:15:09.855560Z"
+                writerId: ppsci3bha2i61tuvdhjs2d9rc3ipwu77
+        m:
+            - "1": train/global_step
+              "6":
+                - 3
+              "7": []
+            - "2": '*'
+              "5": 1
+              "6":
+                - 1
+              "7": []
+        python_version: 3.11.14
+        t:
+            "1":
+                - 1
+                - 11
+                - 41
+                - 49
+                - 50
+                - 51
+                - 71
+                - 80
+                - 83
+                - 98
+            "2":
+                - 1
+                - 11
+                - 41
+                - 49
+                - 50
+                - 51
+                - 71
+                - 80
+                - 83
+                - 98
+            "3":
+                - 7
+                - 13
+                - 19
+                - 62
+                - 66
+            "4": 3.11.14
+            "5": 0.25.0
+            "6": 4.51.3
+            "9":
+                "1": transformers_trainer
+            "12": 0.25.0
+            "13": linux-x86_64
+accelerator_config:
+    value:
+        dispatch_batches: null
+        even_batches: true
+        gradient_accumulation_kwargs: null
+        non_blocking: false
+        split_batches: false
+        use_seedable_sampler: true
+action_dim:
+    value: 36
+action_head_cfg:
+    value:
+        _convert_: object
+        _target_: groot.vla.model.dreamzero.action_head.wan_flow_matching_action_tf.WANPolicyHead
+        config:
+            _recursive_: false
+            _target_: groot.vla.model.dreamzero.action_head.wan_flow_matching_action_tf.WANPolicyHeadConfig
+            action_dim: 36
+            action_horizon: 24
+            action_loss_embodiment_ids:
+                - 26
+                - 17
+                - 32
+            add_pos_embed: true
+            backbone_embedding_dim: 0
+            backbone_features_projector_cfg: null
+            decouple_video_action_noise: false
+            diffusion_model_cfg:
+                _convert_: object
+                _target_: groot.vla.model.dreamzero.modules.wan_video_dit_action_casual_chunk.CausalWanModel
+                action_dim: 36
+                diffusion_model_pretrained_path: /hfm/boqian/liboqian_data/checkpoints/Wan2.1-I2V-14B-480P
+                dim: 5120
+                eps: 1e-06
+                ffn_dim: 13824
+                frame_seqlen: 220
+                freq_dim: 256
+                in_dim: 36
+                max_chunk_size: 4
+                model_type: i2v
+                num_action_per_block: 24
+                num_frame_per_block: 2
+                num_heads: 40
+                num_layers: 40
+                num_state_per_block: 1
+                out_dim: 16
+            expand_batch: null
+            freeze_decode_layer: false
+            hidden_size: 64
+            image_encoder_cfg:
+                _convert_: object
+                _target_: groot.vla.model.dreamzero.modules.wan_video_image_encoder.WanImageEncoder
+                image_encoder_pretrained_path: /hfm/boqian/liboqian_data/checkpoints/Wan2.1-I2V-14B-480P/models_clip_open-clip-xlm-roberta-large-vit-huge-14.pth
+            init_lora_weights: kaiming
+            input_embedding_dim: 1536
+            load_pretrained_det_decode_layer_path: null
+            lora_alpha: 4
+            lora_rank: 4
+            lora_target_modules: q,k,v,o,ffn.0,ffn.2
+            max_action_dim: 36
+            max_state_dim: 64
+            model_dtype: float32
+            noise_beta_alpha: 1.5
+            noise_beta_beta: 1
+            noise_s: 0.999
+            num_frame_per_block: 2
+            num_frames: 33
+            num_inference_timesteps: 4
+            num_timestep_buckets: 1000
+            repa_coeff: 1
+            repa_layer: 8
+            skip_component_loading: true
+            text_encoder_cfg:
+                _convert_: object
+                _target_: groot.vla.model.dreamzero.modules.wan_video_text_encoder.WanTextEncoder
+                text_encoder_pretrained_path: /hfm/boqian/liboqian_data/checkpoints/Wan2.1-I2V-14B-480P/models_t5_umt5-xxl-enc-bf16.pth
+            tile_size_height: 34
+            tile_size_width: 34
+            tile_stride_height: 18
+            tile_stride_width: 16
+            tiled: false
+            train_architecture: full
+            tune_diffusion_model: true
+            tune_projector: true
+            use_gradient_checkpointing: true
+            use_vlln: true
+            vae_cfg:
+                _convert_: object
+                _target_: groot.vla.model.dreamzero.modules.wan_video_vae.WanVideoVAE
+                vae_pretrained_path: /hfm/boqian/liboqian_data/checkpoints/Wan2.1-I2V-14B-480P/Wan2.1_VAE.pth
+            video_noise_beta_alpha: 3
+            video_noise_beta_beta: 1
+            vl_self_attention_cfg:
+                _target_: groot.vla.model.n1_5.modules.cross_attention_dit.SelfAttentionTransformer
+                attention_head_dim: 64
+                dropout: 0.2
+                final_dropout: true
+                num_attention_heads: 24
+                num_layers: 4
+                positional_embeddings: null
+action_horizon:
+    value: 24
+adafactor:
+    value: false
+adam_beta1:
+    value: 0.95
+adam_beta2:
+    value: 0.999
+adam_epsilon:
+    value: 1e-08
+add_cross_attention:
+    value: false
+architectures:
+    value: null
+auto_find_batch_size:
+    value: false
+average_tokens_across_devices:
+    value: false
+backbone_cfg:
+    value:
+        _target_: groot.vla.model.dreamzero.backbone.identity.IdentityBackbone
+bad_words_ids:
+    value: null
+batch_eval_metrics:
+    value: false
+begin_suppress_tokens:
+    value: null
+bf16:
+    value: true
+bf16_full_eval:
+    value: false
+bos_token_id:
+    value: null
+chunk_size_feed_forward:
+    value: 0
+cross_attention_hidden_size:
+    value: null
+data_seed:
+    value: null
+dataloader_drop_last:
+    value: false
+dataloader_num_workers:
+    value: 1
+dataloader_persistent_workers:
+    value: true
+dataloader_pin_memory:
+    value: false
+dataloader_prefetch_factor:
+    value: null
+ddp_backend:
+    value: null
+ddp_broadcast_buffers:
+    value: null
+ddp_bucket_cap_mb:
+    value: 100
+ddp_find_unused_parameters:
+    value: false
+ddp_timeout:
+    value: 1800
+debug:
+    value: []
+decoder_start_token_id:
+    value: null
+deepspeed:
+    value: groot/vla/configs/deepspeed/zero2_offload.json
+disable_tqdm:
+    value: false
+diversity_penalty:
+    value: 0
+do_eval:
+    value: false
+do_predict:
+    value: false
+do_sample:
+    value: false
+do_train:
+    value: false
+early_stopping:
+    value: false
+encoder_no_repeat_ngram_size:
+    value: 0
+eos_token_id:
+    value: null
+eval_accumulation_steps:
+    value: null
+eval_delay:
+    value: 0
+eval_do_concat_batches:
+    value: true
+eval_on_start:
+    value: false
+eval_steps:
+    value: null
+eval_strategy:
+    value: "no"
+eval_use_gather_object:
+    value: false
+exponential_decay_length_penalty:
+    value: null
+finetuning_task:
+    value: null
+forced_bos_token_id:
+    value: null
+forced_eos_token_id:
+    value: null
+fp16:
+    value: false
+fp16_backend:
+    value: auto
+fp16_full_eval:
+    value: false
+fp16_opt_level:
+    value: O1
+fsdp:
+    value: []
+fsdp_config:
+    value:
+        min_num_params: 0
+        xla: false
+        xla_fsdp_grad_ckpt: false
+        xla_fsdp_v2: false
+fsdp_min_num_params:
+    value: 0
+fsdp_transformer_layer_cls_to_wrap:
+    value: null
+full_determinism:
+    value: false
+gradient_accumulation_steps:
+    value: 1
+gradient_checkpointing:
+    value: false
+gradient_checkpointing_kwargs:
+    value: null
+greater_is_better:
+    value: null
+group_by_length:
+    value: false
+half_precision_backend:
+    value: auto
+hidden_size:
+    value: 0
+hub_always_push:
+    value: false
+hub_model_id:
+    value: null
+hub_private_repo:
+    value: null
+hub_strategy:
+    value: every_save
+hub_token:
+    value: <HUB_TOKEN>
+id2label:
+    value:
+        "0": LABEL_0
+        "1": LABEL_1
+ignore_data_skip:
+    value: true
+include_for_metrics:
+    value: []
+include_inputs_for_metrics:
+    value: false
+include_num_input_tokens_seen:
+    value: false
+include_tokens_per_second:
+    value: false
+is_decoder:
+    value: false
+is_encoder_decoder:
+    value: false
+jit_mode_eval:
+    value: false
+label_names:
+    value: null
+label_smoothing_factor:
+    value: 0
+label2id:
+    value:
+        LABEL_0: 0
+        LABEL_1: 1
+learning_rate:
+    value: 1e-05
+length_column_name:
+    value: length
+length_penalty:
+    value: 1
+load_best_model_at_end:
+    value: false
+local_rank:
+    value: 0
+log_level:
+    value: passive
+log_level_replica:
+    value: warning
+log_on_each_node:
+    value: true
+logging_dir:
+    value: ./checkpoints/dreamzero_real_teleop_g1_full_finetune_absolute/runs/Mar30_04-04-13_nebula100
+logging_first_step:
+    value: false
+logging_nan_inf_filter:
+    value: true
+logging_steps:
+    value: 10
+logging_strategy:
+    value: steps
+lr_scheduler_type:
+    value: cosine
+max_grad_norm:
+    value: 1
+max_length:
+    value: 20
+max_steps:
+    value: 20000
+metric_for_best_model:
+    value: null
+min_length:
+    value: 0
+model/num_parameters:
+    value: 22924196696
+model_dtype:
+    value: float32
+model_type:
+    value: vla
+mp_parameters:
+    value: ""
+neftune_noise_alpha:
+    value: null
+no_cuda:
+    value: false
+no_repeat_ngram_size:
+    value: 0
+num_beam_groups:
+    value: 1
+num_beams:
+    value: 1
+num_return_sequences:
+    value: 1
+num_train_epochs:
+    value: 1000
+optim:
+    value: adamw_torch
+optim_args:
+    value: null
+optim_target_modules:
+    value: null
+output_attentions:
+    value: false
+output_dir:
+    value: ./checkpoints/dreamzero_real_teleop_g1_full_finetune_absolute
+output_hidden_states:
+    value: false
+output_scores:
+    value: false
+overwrite_output_dir:
+    value: false
+pad_token_id:
+    value: null
+past_index:
+    value: -1
+per_device_eval_batch_size:
+    value: 64
+per_device_train_batch_size:
+    value: 1
+per_gpu_eval_batch_size:
+    value: null
+per_gpu_train_batch_size:
+    value: null
+prediction_loss_only:
+    value: false
+prefix:
+    value: null
+problem_type:
+    value: null
+push_to_hub:
+    value: false
+push_to_hub_model_id:
+    value: null
+push_to_hub_organization:
+    value: null
+push_to_hub_token:
+    value: <PUSH_TO_HUB_TOKEN>
+ray_scope:
+    value: last
+remove_invalid_values:
+    value: false
+remove_unused_columns:
+    value: false
+repetition_penalty:
+    value: 1
+report_to:
+    value:
+        - wandb
+restore_callback_states_from_checkpoint:
+    value: false
+resume_from_checkpoint:
+    value: null
+resume_path:
+    value: ./checkpoints/dreamzero_real_teleop_g1_full_finetune_absolute
+return_dict:
+    value: true
+return_dict_in_generate:
+    value: false
+run_name:
+    value: dreamzero_real_teleop_g1_full_finetune_absolute
+save_on_each_node:
+    value: false
+save_only_model:
+    value: false
+save_safetensors:
+    value: true
+save_steps:
+    value: 8000
+save_strategy:
+    value: steps
+save_total_limit:
+    value: 10
+seed:
+    value: 42
+sep_token_id:
+    value: null
+skip_memory_metrics:
+    value: true
+suppress_tokens:
+    value: null
+task_specific_params:
+    value: null
+temperature:
+    value: 1
+tf_legacy_loss:
+    value: false
+tf32:
+    value: true
+tie_encoder_decoder:
+    value: false
+tie_word_embeddings:
+    value: true
+tokenizer_class:
+    value: null
+top_k:
+    value: 50
+top_p:
+    value: 1
+torch_compile:
+    value: false
+torch_compile_backend:
+    value: null
+torch_compile_mode:
+    value: null
+torch_dtype:
+    value: null
+torch_empty_cache_steps:
+    value: null
+torchdynamo:
+    value: null
+torchscript:
+    value: false
+tp_size:
+    value: 0
+tpu_metrics_debug:
+    value: false
+tpu_num_cores:
+    value: null
+transformers_version:
+    value: 4.51.3
+typical_p:
+    value: 1
+use_bfloat16:
+    value: false
+use_cpu:
+    value: false
+use_ipex:
+    value: false
+use_legacy_prediction_loop:
+    value: false
+use_liger_kernel:
+    value: false
+use_mps_device:
+    value: false
+warmup_ratio:
+    value: 0.05
+warmup_steps:
+    value: 0
+weight_decay:
+    value: 1e-05

dreamzero_real_teleop_g1_full_finetune_absolute_Pick_bottle_and_turn_and_pour_into_cup/wandb/run-20260330_041509-ao6lob27/files/wandb-summary.json

@@ -0,0 +1 @@
+{"_wandb":{"runtime":209358},"_step":27600,"train/dynamics_loss_avg":0.047976726107299326,"train/grad_norm":0.745741605758667,"_timestamp":1.7750533435776424e+09,"train_samples_per_second":0.765,"train_loss":0.04518256734460592,"train/training_step_time":16.269297122955322,"train/loss":0.0696,"train_runtime":209234.0757,"train/learning_rate":6.834906085551041e-14,"total_flos":0,"train_steps_per_second":0.096,"train/global_step":20000,"train/epoch":1.1444091796875e-05,"train/model_forward_time":1.6608974933624268,"train/action_loss_avg":0.02277223765850067,"_runtime":209358}

dreamzero_real_teleop_g1_full_finetune_absolute_Pick_bottle_and_turn_and_pour_into_cup/wandb/run-20260330_041509-ao6lob27/logs/debug-core.log

@@ -5,3 +5,11 @@
 {"time":"2026-03-30T04:15:10.145622128Z","level":"INFO","msg":"handleInformInit: received","streamId":"ao6lob27","id":"1(@)"}
 {"time":"2026-03-30T04:15:10.413378001Z","level":"INFO","msg":"handleInformInit: stream started","streamId":"ao6lob27","id":"1(@)"}
 {"time":"2026-03-30T04:15:15.935635806Z","level":"INFO","msg":"connection: cancelling request","id":"1(@)","requestId":"x5vnbp8x770d"}
+{"time":"2026-04-01T14:24:28.744616233Z","level":"INFO","msg":"handleInformTeardown: server teardown initiated","id":"1(@)"}
+{"time":"2026-04-01T14:24:28.746553349Z","level":"INFO","msg":"connection: closing","id":"1(@)"}
+{"time":"2026-04-01T14:24:28.74663126Z","level":"INFO","msg":"server is shutting down"}
+{"time":"2026-04-01T14:24:28.74666882Z","level":"INFO","msg":"connection: closed successfully","id":"1(@)"}
+{"time":"2026-04-01T14:24:28.750389742Z","level":"INFO","msg":"server: listener closed","addr":{"Name":"/tmp/wandb-977089-981034-2917860159/socket","Net":"unix"}}
+{"time":"2026-04-01T14:24:29.858907981Z","level":"INFO","msg":"handleInformTeardown: server shutdown complete","id":"1(@)"}
+{"time":"2026-04-01T14:24:29.858971311Z","level":"INFO","msg":"connection: ManageConnectionData: connection closed","id":"1(@)"}
+{"time":"2026-04-01T14:24:29.858997432Z","level":"INFO","msg":"server is closed"}

dreamzero_real_teleop_g1_full_finetune_absolute_Pick_bottle_and_turn_and_pour_into_cup/wandb/run-20260330_041509-ao6lob27/logs/debug-internal.log

@@ -5,3 +5,37 @@
 {"time":"2026-03-30T04:15:10.413395221Z","level":"INFO","msg":"writer: started","stream_id":"ao6lob27"}
 {"time":"2026-03-30T04:15:10.413394431Z","level":"INFO","msg":"sender: started","stream_id":"ao6lob27"}
 {"time":"2026-03-30T22:30:11.007336848Z","level":"INFO","msg":"api: retrying HTTP error","status":502,"url":"https://api.wandb.ai/files/boqianli/dreamzero/ao6lob27/file_stream","body":"\n<html><head>\n<meta http-equiv=\"content-type\" content=\"text/html;charset=utf-8\">\n<title>502 Server Error</title>\n</head>\n<body text=#000000 bgcolor=#ffffff>\n<h1>Error: Server Error</h1>\n<h2>The server encountered a temporary error and could not complete your request.<p>Please try again in 30 seconds.</h2>\n<h2></h2>\n</body></html>\n"}
+{"time":"2026-04-01T10:18:44.020890295Z","level":"INFO","msg":"api: retrying error","error":"Post \"https://api.wandb.ai/files/boqianli/dreamzero/ao6lob27/file_stream\": read tcp 192.168.0.3:41808->35.186.228.49:443: read: connection timed out"}
+{"time":"2026-04-01T10:18:56.087878557Z","level":"INFO","msg":"api: retrying error","error":"Post \"https://api.wandb.ai/files/boqianli/dreamzero/ao6lob27/file_stream\": dial tcp: lookup api.wandb.ai on 127.0.0.53:53: read udp 127.0.0.1:38712->127.0.0.53:53: i/o timeout"}
+{"time":"2026-04-01T10:19:10.908212851Z","level":"INFO","msg":"api: retrying error","error":"Post \"https://api.wandb.ai/files/boqianli/dreamzero/ao6lob27/file_stream\": dial tcp: lookup api.wandb.ai on 127.0.0.53:53: read udp 127.0.0.1:39469->127.0.0.53:53: i/o timeout"}
+{"time":"2026-04-01T10:19:30.190045118Z","level":"INFO","msg":"api: retrying error","error":"Post \"https://api.wandb.ai/files/boqianli/dreamzero/ao6lob27/file_stream\": dial tcp: lookup api.wandb.ai on 127.0.0.53:53: read udp 127.0.0.1:47884->127.0.0.53:53: i/o timeout"}
+{"time":"2026-04-01T10:19:58.216957187Z","level":"INFO","msg":"api: retrying error","error":"Post \"https://api.wandb.ai/files/boqianli/dreamzero/ao6lob27/file_stream\": dial tcp: lookup api.wandb.ai on 127.0.0.53:53: read udp 127.0.0.1:34560->127.0.0.53:53: i/o timeout"}
+{"time":"2026-04-01T10:20:40.445735516Z","level":"INFO","msg":"api: retrying error","error":"Post \"https://api.wandb.ai/files/boqianli/dreamzero/ao6lob27/file_stream\": dial tcp: lookup api.wandb.ai on 127.0.0.53:53: read udp 127.0.0.1:46537->127.0.0.53:53: i/o timeout"}
+{"time":"2026-04-01T10:21:50.450660421Z","level":"INFO","msg":"api: retrying error","error":"Post \"https://api.wandb.ai/files/boqianli/dreamzero/ao6lob27/file_stream\": dial tcp: lookup api.wandb.ai on 127.0.0.53:53: read udp 127.0.0.1:39132->127.0.0.53:53: i/o timeout"}
+{"time":"2026-04-01T10:23:00.454344979Z","level":"INFO","msg":"api: retrying error","error":"Post \"https://api.wandb.ai/files/boqianli/dreamzero/ao6lob27/file_stream\": dial tcp: lookup api.wandb.ai on 127.0.0.53:53: read udp 127.0.0.1:60808->127.0.0.53:53: i/o timeout"}
+{"time":"2026-04-01T10:24:10.457186285Z","level":"INFO","msg":"api: retrying error","error":"Post \"https://api.wandb.ai/files/boqianli/dreamzero/ao6lob27/file_stream\": dial tcp: lookup api.wandb.ai on 127.0.0.53:53: read udp 127.0.0.1:42899->127.0.0.53:53: i/o timeout"}
+{"time":"2026-04-01T10:25:20.462333271Z","level":"INFO","msg":"api: retrying error","error":"Post \"https://api.wandb.ai/files/boqianli/dreamzero/ao6lob27/file_stream\": dial tcp: lookup api.wandb.ai on 127.0.0.53:53: read udp 127.0.0.1:51640->127.0.0.53:53: i/o timeout"}
+{"time":"2026-04-01T10:26:30.466228733Z","level":"INFO","msg":"api: retrying error","error":"Post \"https://api.wandb.ai/files/boqianli/dreamzero/ao6lob27/file_stream\": dial tcp: lookup api.wandb.ai on 127.0.0.53:53: read udp 127.0.0.1:54385->127.0.0.53:53: i/o timeout"}
+{"time":"2026-04-01T10:27:40.469992011Z","level":"INFO","msg":"api: retrying error","error":"Post \"https://api.wandb.ai/files/boqianli/dreamzero/ao6lob27/file_stream\": dial tcp: lookup api.wandb.ai on 127.0.0.53:53: read udp 127.0.0.1:36994->127.0.0.53:53: i/o timeout"}
+{"time":"2026-04-01T10:28:50.474794045Z","level":"INFO","msg":"api: retrying error","error":"Post \"https://api.wandb.ai/files/boqianli/dreamzero/ao6lob27/file_stream\": dial tcp: lookup api.wandb.ai on 127.0.0.53:53: read udp 127.0.0.1:59340->127.0.0.53:53: i/o timeout"}
+{"time":"2026-04-01T10:30:00.479410672Z","level":"INFO","msg":"api: retrying error","error":"Post \"https://api.wandb.ai/files/boqianli/dreamzero/ao6lob27/file_stream\": dial tcp: lookup api.wandb.ai on 127.0.0.53:53: read udp 127.0.0.1:53914->127.0.0.53:53: i/o timeout"}
+{"time":"2026-04-01T10:31:10.483284091Z","level":"INFO","msg":"api: retrying error","error":"Post \"https://api.wandb.ai/files/boqianli/dreamzero/ao6lob27/file_stream\": dial tcp: lookup api.wandb.ai on 127.0.0.53:53: read udp 127.0.0.1:52683->127.0.0.53:53: i/o timeout"}
+{"time":"2026-04-01T10:32:20.487788622Z","level":"INFO","msg":"api: retrying error","error":"Post \"https://api.wandb.ai/files/boqianli/dreamzero/ao6lob27/file_stream\": dial tcp: lookup api.wandb.ai on 127.0.0.53:53: read udp 127.0.0.1:52313->127.0.0.53:53: i/o timeout"}
+{"time":"2026-04-01T10:33:30.492784858Z","level":"INFO","msg":"api: retrying error","error":"Post \"https://api.wandb.ai/files/boqianli/dreamzero/ao6lob27/file_stream\": dial tcp: lookup api.wandb.ai on 127.0.0.53:53: read udp 127.0.0.1:38331->127.0.0.53:53: i/o timeout"}
+{"time":"2026-04-01T10:34:40.497056663Z","level":"INFO","msg":"api: retrying error","error":"Post \"https://api.wandb.ai/files/boqianli/dreamzero/ao6lob27/file_stream\": dial tcp: lookup api.wandb.ai on 127.0.0.53:53: read udp 127.0.0.1:59696->127.0.0.53:53: i/o timeout"}
+{"time":"2026-04-01T10:35:50.500859105Z","level":"INFO","msg":"api: retrying error","error":"Post \"https://api.wandb.ai/files/boqianli/dreamzero/ao6lob27/file_stream\": dial tcp: lookup api.wandb.ai on 127.0.0.53:53: read udp 127.0.0.1:40973->127.0.0.53:53: i/o timeout"}
+{"time":"2026-04-01T10:37:00.5054549Z","level":"INFO","msg":"api: retrying error","error":"Post \"https://api.wandb.ai/files/boqianli/dreamzero/ao6lob27/file_stream\": dial tcp: lookup api.wandb.ai on 127.0.0.53:53: read udp 127.0.0.1:45222->127.0.0.53:53: i/o timeout"}
+{"time":"2026-04-01T10:38:10.508535232Z","level":"INFO","msg":"api: retrying error","error":"Post \"https://api.wandb.ai/files/boqianli/dreamzero/ao6lob27/file_stream\": dial tcp: lookup api.wandb.ai on 127.0.0.53:53: read udp 127.0.0.1:55362->127.0.0.53:53: i/o timeout"}
+{"time":"2026-04-01T10:39:20.512300509Z","level":"INFO","msg":"api: retrying error","error":"Post \"https://api.wandb.ai/files/boqianli/dreamzero/ao6lob27/file_stream\": dial tcp: lookup api.wandb.ai on 127.0.0.53:53: read udp 127.0.0.1:50089->127.0.0.53:53: i/o timeout"}
+{"time":"2026-04-01T10:40:30.517197984Z","level":"INFO","msg":"api: retrying error","error":"Post \"https://api.wandb.ai/files/boqianli/dreamzero/ao6lob27/file_stream\": dial tcp: lookup api.wandb.ai on 127.0.0.53:53: read udp 127.0.0.1:34388->127.0.0.53:53: i/o timeout"}
+{"time":"2026-04-01T10:41:40.521767935Z","level":"INFO","msg":"api: retrying error","error":"Post \"https://api.wandb.ai/files/boqianli/dreamzero/ao6lob27/file_stream\": dial tcp: lookup api.wandb.ai on 127.0.0.53:53: read udp 127.0.0.1:53489->127.0.0.53:53: i/o timeout"}
+{"time":"2026-04-01T10:42:50.525040035Z","level":"INFO","msg":"api: retrying error","error":"Post \"https://api.wandb.ai/files/boqianli/dreamzero/ao6lob27/file_stream\": dial tcp: lookup api.wandb.ai on 127.0.0.53:53: read udp 127.0.0.1:50726->127.0.0.53:53: i/o timeout"}
+{"time":"2026-04-01T10:44:00.527917425Z","level":"INFO","msg":"api: retrying error","error":"Post \"https://api.wandb.ai/files/boqianli/dreamzero/ao6lob27/file_stream\": dial tcp: lookup api.wandb.ai on 127.0.0.53:53: read udp 127.0.0.1:53434->127.0.0.53:53: i/o timeout"}
+{"time":"2026-04-01T10:45:10.532222422Z","level":"INFO","msg":"api: retrying error","error":"Post \"https://api.wandb.ai/files/boqianli/dreamzero/ao6lob27/file_stream\": dial tcp: lookup api.wandb.ai on 127.0.0.53:53: read udp 127.0.0.1:37341->127.0.0.53:53: i/o timeout"}
+{"time":"2026-04-01T10:46:20.535679123Z","level":"INFO","msg":"api: retrying error","error":"Post \"https://api.wandb.ai/files/boqianli/dreamzero/ao6lob27/file_stream\": dial tcp: lookup api.wandb.ai on 127.0.0.53:53: read udp 127.0.0.1:37921->127.0.0.53:53: i/o timeout"}
+{"time":"2026-04-01T10:47:30.539549836Z","level":"INFO","msg":"api: retrying error","error":"Post \"https://api.wandb.ai/files/boqianli/dreamzero/ao6lob27/file_stream\": dial tcp: lookup api.wandb.ai on 127.0.0.53:53: read udp 127.0.0.1:38214->127.0.0.53:53: i/o timeout"}
+{"time":"2026-04-01T14:24:28.755750998Z","level":"INFO","msg":"stream: closing","id":"ao6lob27"}
+{"time":"2026-04-01T14:24:29.342449946Z","level":"INFO","msg":"fileTransfer: Close: file transfer manager closed"}
+{"time":"2026-04-01T14:24:29.828344988Z","level":"INFO","msg":"handler: closed","stream_id":"ao6lob27"}
+{"time":"2026-04-01T14:24:29.82853025Z","level":"INFO","msg":"sender: closed","stream_id":"ao6lob27"}
+{"time":"2026-04-01T14:24:29.8285597Z","level":"INFO","msg":"stream: closed","id":"ao6lob27"}

dreamzero_real_teleop_g1_full_finetune_absolute_Pick_bottle_and_turn_and_pour_into_cup/wandb/run-20260330_041509-ao6lob27/logs/debug.log

@@ -20,3 +20,5 @@ config: {'_wandb': {}}
 2026-03-30 04:15:10,932 INFO    MainThread:977089 [wandb_run.py:_config_callback():1403] config_cb None None {'return_dict': True, 'output_hidden_states': False, 'output_attentions': False, 'torchscript': False, 'torch_dtype': None, 'use_bfloat16': False, 'tf_legacy_loss': False, 'pruned_heads': {}, 'tie_word_embeddings': True, 'chunk_size_feed_forward': 0, 'is_encoder_decoder': False, 'is_decoder': False, 'cross_attention_hidden_size': None, 'add_cross_attention': False, 'tie_encoder_decoder': False, 'max_length': 20, 'min_length': 0, 'do_sample': False, 'early_stopping': False, 'num_beams': 1, 'num_beam_groups': 1, 'diversity_penalty': 0.0, 'temperature': 1.0, 'top_k': 50, 'top_p': 1.0, 'typical_p': 1.0, 'repetition_penalty': 1.0, 'length_penalty': 1.0, 'no_repeat_ngram_size': 0, 'encoder_no_repeat_ngram_size': 0, 'bad_words_ids': None, 'num_return_sequences': 1, 'output_scores': False, 'return_dict_in_generate': False, 'forced_bos_token_id': None, 'forced_eos_token_id': None, 'remove_invalid_values': False, 'exponential_decay_length_penalty': None, 'suppress_tokens': None, 'begin_suppress_tokens': None, 'architectures': None, 'finetuning_task': None, 'id2label': {0: 'LABEL_0', 1: 'LABEL_1'}, 'label2id': {'LABEL_0': 0, 'LABEL_1': 1}, 'tokenizer_class': None, 'prefix': None, 'bos_token_id': None, 'pad_token_id': None, 'eos_token_id': None, 'sep_token_id': None, 'decoder_start_token_id': None, 'task_specific_params': None, 'problem_type': None, '_name_or_path': './checkpoints/dreamzero_real_teleop_g1_full_finetune_absolute', '_attn_implementation_autoset': True, 'transformers_version': '4.51.3', 'model_dtype': 'float32', 'hidden_size': 0, 'action_horizon': 24, 'action_dim': 36, 'backbone_cfg': {'_target_': 'groot.vla.model.dreamzero.backbone.identity.IdentityBackbone'}, 'action_head_cfg': {'config': {'backbone_features_projector_cfg': None, '_target_': 'groot.vla.model.dreamzero.action_head.wan_flow_matching_action_tf.WANPolicyHeadConfig', '_recursive_': False, 'tiled': False, 'tile_size_height': 34, 'tile_size_width': 34, 'tile_stride_height': 18, 'tile_stride_width': 16, 'lora_rank': 4, 'lora_alpha': 4, 'num_frames': 33, 'num_frame_per_block': 2, 'lora_target_modules': 'q,k,v,o,ffn.0,ffn.2', 'init_lora_weights': 'kaiming', 'train_architecture': 'full', 'use_gradient_checkpointing': True, 'add_pos_embed': True, 'model_dtype': 'float32', 'max_state_dim': 64, 'max_action_dim': 36, 'action_loss_embodiment_ids': [26, 17, 32], 'hidden_size': 64, 'input_embedding_dim': 1536, 'backbone_embedding_dim': 0, 'repa_layer': 8, 'repa_coeff': 1.0, 'load_pretrained_det_decode_layer_path': None, 'freeze_decode_layer': False, 'expand_batch': None, 'use_vlln': True, 'vl_self_attention_cfg': {'_target_': 'groot.vla.model.n1_5.modules.cross_attention_dit.SelfAttentionTransformer', 'positional_embeddings': None, 'num_layers': 4, 'num_attention_heads': 24, 'attention_head_dim': 64, 'dropout': 0.2, 'final_dropout': True}, 'diffusion_model_cfg': {'_target_': 'groot.vla.model.dreamzero.modules.wan_video_dit_action_casual_chunk.CausalWanModel', '_convert_': 'object', 'diffusion_model_pretrained_path': '/hfm/boqian/liboqian_data/checkpoints/Wan2.1-I2V-14B-480P', 'model_type': 'i2v', 'frame_seqlen': 220, 'dim': 5120, 'in_dim': 36, 'ffn_dim': 13824, 'out_dim': 16, 'freq_dim': 256, 'eps': 1e-06, 'num_heads': 40, 'num_layers': 40, 'max_chunk_size': 4, 'num_frame_per_block': 2, 'num_action_per_block': 24, 'num_state_per_block': 1, 'action_dim': 36}, 'text_encoder_cfg': {'_target_': 'groot.vla.model.dreamzero.modules.wan_video_text_encoder.WanTextEncoder', '_convert_': 'object', 'text_encoder_pretrained_path': '/hfm/boqian/liboqian_data/checkpoints/Wan2.1-I2V-14B-480P/models_t5_umt5-xxl-enc-bf16.pth'}, 'image_encoder_cfg': {'_target_': 'groot.vla.model.dreamzero.modules.wan_video_image_encoder.WanImageEncoder', '_convert_': 'object', 'image_encoder_pretrained_path': '/hfm/boqian/liboqian_data/checkpoints/Wan2.1-I2V-14B-480P/models_clip_open-clip-xlm-roberta-large-vit-huge-14.pth'}, 'vae_cfg': {'_target_': 'groot.vla.model.dreamzero.modules.wan_video_vae.WanVideoVAE', '_convert_': 'object', 'vae_pretrained_path': '/hfm/boqian/liboqian_data/checkpoints/Wan2.1-I2V-14B-480P/Wan2.1_VAE.pth'}, 'action_dim': 36, 'action_horizon': 24, 'num_inference_timesteps': 4, 'noise_beta_alpha': 1.5, 'noise_beta_beta': 1.0, 'noise_s': 0.999, 'num_timestep_buckets': 1000, 'decouple_video_action_noise': False, 'video_noise_beta_alpha': 3.0, 'video_noise_beta_beta': 1.0, 'tune_projector': True, 'tune_diffusion_model': True, 'skip_component_loading': True}, '_target_': 'groot.vla.model.dreamzero.action_head.wan_flow_matching_action_tf.WANPolicyHead', '_convert_': 'object'}, 'resume_path': './checkpoints/dreamzero_real_teleop_g1_full_finetune_absolute', 'model_type': 'vla', 'output_dir': './checkpoints/dreamzero_real_teleop_g1_full_finetune_absolute', 'overwrite_output_dir': False, 'do_train': False, 'do_eval': False, 'do_predict': False, 'eval_strategy': 'no', 'prediction_loss_only': False, 'per_device_train_batch_size': 1, 'per_device_eval_batch_size': 64, 'per_gpu_train_batch_size': None, 'per_gpu_eval_batch_size': None, 'gradient_accumulation_steps': 1, 'eval_accumulation_steps': None, 'eval_delay': 0, 'torch_empty_cache_steps': None, 'learning_rate': 1e-05, 'weight_decay': 1e-05, 'adam_beta1': 0.95, 'adam_beta2': 0.999, 'adam_epsilon': 1e-08, 'max_grad_norm': 1.0, 'num_train_epochs': 1000, 'max_steps': 20000, 'lr_scheduler_type': 'cosine', 'lr_scheduler_kwargs': {}, 'warmup_ratio': 0.05, 'warmup_steps': 0, 'log_level': 'passive', 'log_level_replica': 'warning', 'log_on_each_node': True, 'logging_dir': './checkpoints/dreamzero_real_teleop_g1_full_finetune_absolute/runs/Mar30_04-04-13_nebula100', 'logging_strategy': 'steps', 'logging_first_step': False, 'logging_steps': 10, 'logging_nan_inf_filter': True, 'save_strategy': 'steps', 'save_steps': 8000, 'save_total_limit': 10, 'save_safetensors': True, 'save_on_each_node': False, 'save_only_model': False, 'restore_callback_states_from_checkpoint': False, 'no_cuda': False, 'use_cpu': False, 'use_mps_device': False, 'seed': 42, 'data_seed': None, 'jit_mode_eval': False, 'use_ipex': False, 'bf16': True, 'fp16': False, 'fp16_opt_level': 'O1', 'half_precision_backend': 'auto', 'bf16_full_eval': False, 'fp16_full_eval': False, 'tf32': True, 'local_rank': 0, 'ddp_backend': None, 'tpu_num_cores': None, 'tpu_metrics_debug': False, 'debug': [], 'dataloader_drop_last': False, 'eval_steps': None, 'dataloader_num_workers': 1, 'dataloader_prefetch_factor': None, 'past_index': -1, 'run_name': 'dreamzero_real_teleop_g1_full_finetune_absolute', 'disable_tqdm': False, 'remove_unused_columns': False, 'label_names': None, 'load_best_model_at_end': False, 'metric_for_best_model': None, 'greater_is_better': None, 'ignore_data_skip': True, 'fsdp': [], 'fsdp_min_num_params': 0, 'fsdp_config': {'min_num_params': 0, 'xla': False, 'xla_fsdp_v2': False, 'xla_fsdp_grad_ckpt': False}, 'tp_size': 0, 'fsdp_transformer_layer_cls_to_wrap': None, 'accelerator_config': {'split_batches': False, 'dispatch_batches': None, 'even_batches': True, 'use_seedable_sampler': True, 'non_blocking': False, 'gradient_accumulation_kwargs': None}, 'deepspeed': 'groot/vla/configs/deepspeed/zero2_offload.json', 'label_smoothing_factor': 0.0, 'optim': 'adamw_torch', 'optim_args': None, 'adafactor': False, 'group_by_length': False, 'length_column_name': 'length', 'report_to': ['wandb'], 'ddp_find_unused_parameters': False, 'ddp_bucket_cap_mb': 100, 'ddp_broadcast_buffers': None, 'dataloader_pin_memory': False, 'dataloader_persistent_workers': True, 'skip_memory_metrics': True, 'use_legacy_prediction_loop': False, 'push_to_hub': False, 'resume_from_checkpoint': None, 'hub_model_id': None, 'hub_strategy': 'every_save', 'hub_token': '<HUB_TOKEN>', 'hub_private_repo': None, 'hub_always_push': False, 'gradient_checkpointing': False, 'gradient_checkpointing_kwargs': None, 'include_inputs_for_metrics': False, 'include_for_metrics': [], 'eval_do_concat_batches': True, 'fp16_backend': 'auto', 'push_to_hub_model_id': None, 'push_to_hub_organization': None, 'push_to_hub_token': '<PUSH_TO_HUB_TOKEN>', 'mp_parameters': '', 'auto_find_batch_size': False, 'full_determinism': False, 'torchdynamo': None, 'ray_scope': 'last', 'ddp_timeout': 1800, 'torch_compile': False, 'torch_compile_backend': None, 'torch_compile_mode': None, 'include_tokens_per_second': False, 'include_num_input_tokens_seen': False, 'neftune_noise_alpha': None, 'optim_target_modules': None, 'batch_eval_metrics': False, 'eval_on_start': False, 'use_liger_kernel': False, 'eval_use_gather_object': False, 'average_tokens_across_devices': False}
 2026-03-30 04:15:10,939 INFO    MainThread:977089 [wandb_config.py:__setitem__():155] [no run ID] config set model/num_parameters = 22924196696 - <bound method Run._config_callback of <wandb.sdk.wandb_run.Run object at 0x7b9f0d23f190>>
 2026-03-30 04:15:10,939 INFO    MainThread:977089 [wandb_run.py:_config_callback():1403] config_cb model/num_parameters 22924196696 None
+2026-04-01 14:24:28,737 INFO    wandb-AsyncioManager-main:977089 [service_client.py:_forward_responses():134] Reached EOF.
+2026-04-01 14:24:28,738 INFO    wandb-AsyncioManager-main:977089 [mailbox.py:close():155] Closing mailbox, abandoning 1 handles.

dreamzero_real_teleop_g1_full_finetune_absolute_Pick_bottle_and_turn_and_pour_into_cup/wandb/run-20260330_041509-ao6lob27/run-ao6lob27.wandb

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:d9a7c8981c4b92aea8ed1d6f0294eea264e6dad8984cdf7b6f78537488b5906b
-size 94961664
+oid sha256:b42c924544506ffebf5d1a80c95d92dd18fa643ba9d19630056da6d28a5df412
+size 116007394