diff --git "a/data_config.json" "b/data_config.json" new file mode 100644--- /dev/null +++ "b/data_config.json" @@ -0,0 +1,16778 @@ +{ + "datasets": [ + { + "dataset": { + "datasets": [ + { + "dataset": { + "datasets": [ + { + "dataset": { + "batch_size": 64, + "num_workers": 8, + "shuffle_buffer_size": 1000, + "normalization_mapping": { + "ACTION": "ACTIONCHUNK_QUANTILE", + "ENV": "QUANTILE", + "STATE": "QUANTILE", + "VISUAL": "IDENTITY" + }, + "features": { + "action": { + "type": "ACTION", + "shape": "(20,)" + }, + "observation.image.0": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.1": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.2": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.state": { + "type": "STATE", + "shape": "(20,)" + } + }, + "stats": null, + "clip_values": { + "ENV": [ + -1.05, + 1.05 + ], + "STATE": [ + -1.05, + 1.05 + ] + }, + "chunk_size": 50, + "observation_mapping": { + "action.ee_rpy_pos": "action", + "observation.ee_rpy_pos": "observation.state", + "observation.images.image": "observation.image.0" + }, + "observation_whitelist": [ + "observation.state", + "observation.image.0", + "observation.image.1", + "observation.image.2", + "action", + "task", + "subtask", + "dataset_index", + "training_mode" + ], + "observation_blacklist": null, + "transform_mapping": { + "action": [ + { + "type": "convert_to_6d_actions", + "action_key": "action", + "action_type": "EE_EULER_POS", + "input_type": "Dict", + "post_norm": false + }, + { + "type": "convert_to_6d_actions", + "action_key": "observation.state", + "action_type": "EE_EULER_POS", + "input_type": "Dict", + "post_norm": false + }, + { + "type": "delta_actions", + "state_key": "observation.state", + "action_key": "action", + "action_type": "EE_6D_POS", + "input_type": "Dict", + "post_norm": false, + "relative_to_state": true + } + ], + "observation.image.0": [ + { + "type": "random_resized_crop", + "height": 224, + "width": 224, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "observation.image.1": [ + { + "type": "random_resized_crop", + "height": 224, + "width": 224, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "observation.image.2": [ + { + "type": "random_resized_crop", + "height": 224, + "width": 224, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ] + }, + "language_instruction": null, + "training_mode": "ROBOT_FLOWMATCH", + "action_type": "EE_EULER_POS", + "serialize_stats": false, + "video_decoder_cache_size": null, + "repo_id": "kuka", + "streaming": false, + "tolerance_s": 0.1, + "use_imagenet_stats": false, + "split_name": "train", + "prefetch_factor": 2, + "max_episodes": null + }, + "weight": 0.8341046294 + }, + { + "dataset": { + "batch_size": 64, + "num_workers": 8, + "shuffle_buffer_size": 1000, + "normalization_mapping": { + "ACTION": "ACTIONCHUNK_QUANTILE", + "ENV": "QUANTILE", + "STATE": "QUANTILE", + "VISUAL": "IDENTITY" + }, + "features": { + "action": { + "type": "ACTION", + "shape": "(20,)" + }, + "observation.image.0": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.1": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.2": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.state": { + "type": "STATE", + "shape": "(20,)" + } + }, + "stats": null, + "clip_values": { + "ENV": [ + -1.05, + 1.05 + ], + "STATE": [ + -1.05, + 1.05 + ] + }, + "chunk_size": 50, + "observation_mapping": { + "action.ee_rpy_pos": "action", + "observation.ee_rpy_pos": "observation.state", + "observation.images.image_0": "observation.image.0" + }, + "observation_whitelist": [ + "observation.state", + "observation.image.0", + "observation.image.1", + "observation.image.2", + "action", + "task", + "subtask", + "dataset_index", + "training_mode" + ], + "observation_blacklist": [ + "observation.images.image_1", + "observation.images.image_2", + "observation.images.image_3" + ], + "transform_mapping": { + "action": [ + { + "type": "convert_to_6d_actions", + "action_key": "action", + "action_type": "EE_EULER_POS", + "input_type": "Dict", + "post_norm": false + }, + { + "type": "convert_to_6d_actions", + "action_key": "observation.state", + "action_type": "EE_EULER_POS", + "input_type": "Dict", + "post_norm": false + }, + { + "type": "delta_actions", + "state_key": "observation.state", + "action_key": "action", + "action_type": "EE_6D_POS", + "input_type": "Dict", + "post_norm": false, + "relative_to_state": true + } + ], + "observation.image.0": [ + { + "type": "random_resized_crop", + "height": 224, + "width": 224, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "observation.image.1": [ + { + "type": "random_resized_crop", + "height": 224, + "width": 224, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "observation.image.2": [ + { + "type": "random_resized_crop", + "height": 224, + "width": 224, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ] + }, + "language_instruction": null, + "training_mode": "ROBOT_FLOWMATCH", + "action_type": "EE_EULER_POS", + "serialize_stats": false, + "video_decoder_cache_size": null, + "repo_id": "bridge_orig", + "streaming": false, + "tolerance_s": 0.1, + "use_imagenet_stats": false, + "split_name": "train", + "prefetch_factor": 2, + "max_episodes": null + }, + "weight": 1.0 + }, + { + "dataset": { + "batch_size": 64, + "num_workers": 8, + "shuffle_buffer_size": 1000, + "normalization_mapping": { + "ACTION": "ACTIONCHUNK_QUANTILE", + "ENV": "QUANTILE", + "STATE": "QUANTILE", + "VISUAL": "IDENTITY" + }, + "features": { + "action": { + "type": "ACTION", + "shape": "(20,)" + }, + "observation.image.0": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.1": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.2": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.state": { + "type": "STATE", + "shape": "(20,)" + } + }, + "stats": null, + "clip_values": { + "ENV": [ + -1.05, + 1.05 + ], + "STATE": [ + -1.05, + 1.05 + ] + }, + "chunk_size": 50, + "observation_mapping": { + "action.ee_rpy_pos": "action", + "observation.ee_rpy_pos": "observation.state", + "observation.images.image": "observation.image.0", + "observation.images.image_wrist": "observation.image.1" + }, + "observation_whitelist": [ + "observation.state", + "observation.image.0", + "observation.image.1", + "observation.image.2", + "action", + "task", + "subtask", + "dataset_index", + "training_mode" + ], + "observation_blacklist": null, + "transform_mapping": { + "action": [ + { + "type": "convert_to_6d_actions", + "action_key": "action", + "action_type": "EE_EULER_POS", + "input_type": "Dict", + "post_norm": false + }, + { + "type": "convert_to_6d_actions", + "action_key": "observation.state", + "action_type": "EE_EULER_POS", + "input_type": "Dict", + "post_norm": false + }, + { + "type": "delta_actions", + "state_key": "observation.state", + "action_key": "action", + "action_type": "EE_6D_POS", + "input_type": "Dict", + "post_norm": false, + "relative_to_state": true + } + ], + "observation.image.0": [ + { + "type": "random_resized_crop", + "height": 224, + "width": 224, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "observation.image.1": [ + { + "type": "random_resized_crop", + "height": 224, + "width": 224, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "observation.image.2": [ + { + "type": "random_resized_crop", + "height": 224, + "width": 224, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ] + }, + "language_instruction": null, + "training_mode": "ROBOT_FLOWMATCH", + "action_type": "EE_EULER_POS", + "serialize_stats": false, + "video_decoder_cache_size": null, + "repo_id": "jaco_play", + "streaming": false, + "tolerance_s": 0.1, + "use_imagenet_stats": false, + "split_name": "train", + "prefetch_factor": 2, + "max_episodes": null + }, + "weight": 1.0 + }, + { + "dataset": { + "batch_size": 64, + "num_workers": 8, + "shuffle_buffer_size": 1000, + "normalization_mapping": { + "ACTION": "ACTIONCHUNK_QUANTILE", + "ENV": "QUANTILE", + "STATE": "QUANTILE", + "VISUAL": "IDENTITY" + }, + "features": { + "action": { + "type": "ACTION", + "shape": "(20,)" + }, + "observation.image.0": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.1": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.2": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.state": { + "type": "STATE", + "shape": "(20,)" + } + }, + "stats": null, + "clip_values": { + "ENV": [ + -1.05, + 1.05 + ], + "STATE": [ + -1.05, + 1.05 + ] + }, + "chunk_size": 50, + "observation_mapping": { + "action.ee_rpy_pos": "action", + "observation.ee_rpy_pos": "observation.state", + "observation.images.hand_image": "observation.image.2", + "observation.images.image": "observation.image.0" + }, + "observation_whitelist": [ + "observation.state", + "observation.image.0", + "observation.image.1", + "observation.image.2", + "action", + "task", + "subtask", + "dataset_index", + "training_mode" + ], + "observation_blacklist": null, + "transform_mapping": { + "action": [ + { + "type": "convert_to_6d_actions", + "action_key": "action", + "action_type": "EE_EULER_POS", + "input_type": "Dict", + "post_norm": false + }, + { + "type": "convert_to_6d_actions", + "action_key": "observation.state", + "action_type": "EE_EULER_POS", + "input_type": "Dict", + "post_norm": false + }, + { + "type": "delta_actions", + "state_key": "observation.state", + "action_key": "action", + "action_type": "EE_6D_POS", + "input_type": "Dict", + "post_norm": false, + "relative_to_state": true + } + ], + "observation.image.0": [ + { + "type": "random_resized_crop", + "height": 224, + "width": 224, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "observation.image.1": [ + { + "type": "random_resized_crop", + "height": 224, + "width": 224, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "observation.image.2": [ + { + "type": "random_resized_crop", + "height": 224, + "width": 224, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ] + }, + "language_instruction": null, + "training_mode": "ROBOT_FLOWMATCH", + "action_type": "EE_EULER_POS", + "serialize_stats": false, + "video_decoder_cache_size": null, + "repo_id": "berkeley_autolab_ur5", + "streaming": false, + "tolerance_s": 0.1, + "use_imagenet_stats": false, + "split_name": "train", + "prefetch_factor": 2, + "max_episodes": null + }, + "weight": 2.0 + }, + { + "dataset": { + "batch_size": 64, + "num_workers": 8, + "shuffle_buffer_size": 1000, + "normalization_mapping": { + "ACTION": "ACTIONCHUNK_QUANTILE", + "ENV": "QUANTILE", + "STATE": "QUANTILE", + "VISUAL": "IDENTITY" + }, + "features": { + "action": { + "type": "ACTION", + "shape": "(20,)" + }, + "observation.image.0": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.1": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.2": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.state": { + "type": "STATE", + "shape": "(20,)" + } + }, + "stats": null, + "clip_values": { + "ENV": [ + -1.05, + 1.05 + ], + "STATE": [ + -1.05, + 1.05 + ] + }, + "chunk_size": 50, + "observation_mapping": { + "action.ee_rpy_pos": "action", + "observation.ee_rpy_pos": "observation.state", + "observation.images.image": "observation.image.0" + }, + "observation_whitelist": [ + "observation.state", + "observation.image.0", + "observation.image.1", + "observation.image.2", + "action", + "task", + "subtask", + "dataset_index", + "training_mode" + ], + "observation_blacklist": null, + "transform_mapping": { + "action": [ + { + "type": "convert_to_6d_actions", + "action_key": "action", + "action_type": "EE_EULER_POS", + "input_type": "Dict", + "post_norm": false + }, + { + "type": "convert_to_6d_actions", + "action_key": "observation.state", + "action_type": "EE_EULER_POS", + "input_type": "Dict", + "post_norm": false + }, + { + "type": "delta_actions", + "state_key": "observation.state", + "action_key": "action", + "action_type": "EE_6D_POS", + "input_type": "Dict", + "post_norm": false, + "relative_to_state": true + } + ], + "observation.image.0": [ + { + "type": "random_resized_crop", + "height": 224, + "width": 224, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "observation.image.1": [ + { + "type": "random_resized_crop", + "height": 224, + "width": 224, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "observation.image.2": [ + { + "type": "random_resized_crop", + "height": 224, + "width": 224, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ] + }, + "language_instruction": null, + "training_mode": "ROBOT_FLOWMATCH", + "action_type": "EE_EULER_POS", + "serialize_stats": false, + "video_decoder_cache_size": null, + "repo_id": "toto", + "streaming": false, + "tolerance_s": 0.1, + "use_imagenet_stats": false, + "split_name": "train", + "prefetch_factor": 2, + "max_episodes": null + }, + "weight": 1.0 + }, + { + "dataset": { + "batch_size": 64, + "num_workers": 8, + "shuffle_buffer_size": 1000, + "normalization_mapping": { + "ACTION": "ACTIONCHUNK_QUANTILE", + "ENV": "QUANTILE", + "STATE": "QUANTILE", + "VISUAL": "IDENTITY" + }, + "features": { + "action": { + "type": "ACTION", + "shape": "(20,)" + }, + "observation.image.0": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.1": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.2": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.state": { + "type": "STATE", + "shape": "(20,)" + } + }, + "stats": null, + "clip_values": { + "ENV": [ + -1.05, + 1.05 + ], + "STATE": [ + -1.05, + 1.05 + ] + }, + "chunk_size": 50, + "observation_mapping": { + "action.ee_rpy_pos": "action", + "observation.ee_rpy_pos": "observation.state", + "observation.images.image": "observation.image.0", + "observation.images.wrist_image": "observation.image.1" + }, + "observation_whitelist": [ + "observation.state", + "observation.image.0", + "observation.image.1", + "observation.image.2", + "action", + "task", + "subtask", + "dataset_index", + "training_mode" + ], + "observation_blacklist": null, + "transform_mapping": { + "action": [ + { + "type": "convert_to_6d_actions", + "action_key": "action", + "action_type": "EE_EULER_POS", + "input_type": "Dict", + "post_norm": false + }, + { + "type": "convert_to_6d_actions", + "action_key": "observation.state", + "action_type": "EE_EULER_POS", + "input_type": "Dict", + "post_norm": false + }, + { + "type": "delta_actions", + "state_key": "observation.state", + "action_key": "action", + "action_type": "EE_6D_POS", + "input_type": "Dict", + "post_norm": false, + "relative_to_state": true + } + ], + "observation.image.0": [ + { + "type": "random_resized_crop", + "height": 224, + "width": 224, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "observation.image.1": [ + { + "type": "random_resized_crop", + "height": 224, + "width": 224, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "observation.image.2": [ + { + "type": "random_resized_crop", + "height": 224, + "width": 224, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ] + }, + "language_instruction": null, + "training_mode": "ROBOT_FLOWMATCH", + "action_type": "EE_EULER_POS", + "serialize_stats": false, + "video_decoder_cache_size": null, + "repo_id": "austin_buds", + "streaming": false, + "tolerance_s": 0.1, + "use_imagenet_stats": false, + "split_name": "train", + "prefetch_factor": 2, + "max_episodes": null + }, + "weight": 1.0 + }, + { + "dataset": { + "batch_size": 64, + "num_workers": 8, + "shuffle_buffer_size": 1000, + "normalization_mapping": { + "ACTION": "ACTIONCHUNK_QUANTILE", + "ENV": "QUANTILE", + "STATE": "QUANTILE", + "VISUAL": "IDENTITY" + }, + "features": { + "action": { + "type": "ACTION", + "shape": "(20,)" + }, + "observation.image.0": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.1": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.2": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.state": { + "type": "STATE", + "shape": "(20,)" + } + }, + "stats": null, + "clip_values": { + "ENV": [ + -1.05, + 1.05 + ], + "STATE": [ + -1.05, + 1.05 + ] + }, + "chunk_size": 50, + "observation_mapping": { + "action.ee_rpy_pos": "action", + "observation.ee_rpy_pos": "observation.state", + "observation.images.image_additional_view": "observation.image.0" + }, + "observation_whitelist": [ + "observation.state", + "observation.image.0", + "observation.image.1", + "observation.image.2", + "action", + "task", + "subtask", + "dataset_index", + "training_mode" + ], + "observation_blacklist": [ + "observation.images.image" + ], + "transform_mapping": { + "action": [ + { + "type": "convert_to_6d_actions", + "action_key": "action", + "action_type": "EE_EULER_POS", + "input_type": "Dict", + "post_norm": false + }, + { + "type": "convert_to_6d_actions", + "action_key": "observation.state", + "action_type": "EE_EULER_POS", + "input_type": "Dict", + "post_norm": false + }, + { + "type": "delta_actions", + "state_key": "observation.state", + "action_key": "action", + "action_type": "EE_6D_POS", + "input_type": "Dict", + "post_norm": false, + "relative_to_state": true + } + ], + "observation.image.0": [ + { + "type": "random_resized_crop", + "height": 224, + "width": 224, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "observation.image.1": [ + { + "type": "random_resized_crop", + "height": 224, + "width": 224, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "observation.image.2": [ + { + "type": "random_resized_crop", + "height": 224, + "width": 224, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ] + }, + "language_instruction": null, + "training_mode": "ROBOT_FLOWMATCH", + "action_type": "EE_EULER_POS", + "serialize_stats": false, + "video_decoder_cache_size": null, + "repo_id": "nyu_franka_play", + "streaming": false, + "tolerance_s": 0.1, + "use_imagenet_stats": false, + "split_name": "train", + "prefetch_factor": 2, + "max_episodes": null + }, + "weight": 1.5 + }, + { + "dataset": { + "batch_size": 64, + "num_workers": 8, + "shuffle_buffer_size": 1000, + "normalization_mapping": { + "ACTION": "ACTIONCHUNK_QUANTILE", + "ENV": "QUANTILE", + "STATE": "QUANTILE", + "VISUAL": "IDENTITY" + }, + "features": { + "action": { + "type": "ACTION", + "shape": "(20,)" + }, + "observation.image.0": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.1": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.2": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.state": { + "type": "STATE", + "shape": "(20,)" + } + }, + "stats": null, + "clip_values": { + "ENV": [ + -1.05, + 1.05 + ], + "STATE": [ + -1.05, + 1.05 + ] + }, + "chunk_size": 50, + "observation_mapping": { + "action.ee_rpy_pos": "action", + "observation.ee_rpy_pos": "observation.state", + "observation.images.image": "observation.image.0" + }, + "observation_whitelist": [ + "observation.state", + "observation.image.0", + "observation.image.1", + "observation.image.2", + "action", + "task", + "subtask", + "dataset_index", + "training_mode" + ], + "observation_blacklist": [ + "observation.images.image_additional_view" + ], + "transform_mapping": { + "action": [ + { + "type": "convert_to_6d_actions", + "action_key": "action", + "action_type": "EE_EULER_POS", + "input_type": "Dict", + "post_norm": false + }, + { + "type": "convert_to_6d_actions", + "action_key": "observation.state", + "action_type": "EE_EULER_POS", + "input_type": "Dict", + "post_norm": false + }, + { + "type": "delta_actions", + "state_key": "observation.state", + "action_key": "action", + "action_type": "EE_6D_POS", + "input_type": "Dict", + "post_norm": false, + "relative_to_state": true + } + ], + "observation.image.0": [ + { + "type": "random_resized_crop", + "height": 224, + "width": 224, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "observation.image.1": [ + { + "type": "random_resized_crop", + "height": 224, + "width": 224, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "observation.image.2": [ + { + "type": "random_resized_crop", + "height": 224, + "width": 224, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ] + }, + "language_instruction": null, + "training_mode": "ROBOT_FLOWMATCH", + "action_type": "EE_EULER_POS", + "serialize_stats": false, + "video_decoder_cache_size": null, + "repo_id": "nyu_franka_play", + "streaming": false, + "tolerance_s": 0.1, + "use_imagenet_stats": false, + "split_name": "train", + "prefetch_factor": 2, + "max_episodes": null + }, + "weight": 1.5 + }, + { + "dataset": { + "batch_size": 64, + "num_workers": 8, + "shuffle_buffer_size": 1000, + "normalization_mapping": { + "ACTION": "ACTIONCHUNK_QUANTILE", + "ENV": "QUANTILE", + "STATE": "QUANTILE", + "VISUAL": "IDENTITY" + }, + "features": { + "action": { + "type": "ACTION", + "shape": "(20,)" + }, + "observation.image.0": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.1": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.2": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.state": { + "type": "STATE", + "shape": "(20,)" + } + }, + "stats": null, + "clip_values": { + "ENV": [ + -1.05, + 1.05 + ], + "STATE": [ + -1.05, + 1.05 + ] + }, + "chunk_size": 50, + "observation_mapping": { + "action.ee_rpy_pos": "action", + "observation.ee_rpy_pos": "observation.state", + "observation.images.image": "observation.image.0" + }, + "observation_whitelist": [ + "observation.state", + "observation.image.0", + "observation.image.1", + "observation.image.2", + "action", + "task", + "subtask", + "dataset_index", + "training_mode" + ], + "observation_blacklist": null, + "transform_mapping": { + "action": [ + { + "type": "convert_to_6d_actions", + "action_key": "action", + "action_type": "EE_EULER_POS", + "input_type": "Dict", + "post_norm": false + }, + { + "type": "convert_to_6d_actions", + "action_key": "observation.state", + "action_type": "EE_EULER_POS", + "input_type": "Dict", + "post_norm": false + }, + { + "type": "delta_actions", + "state_key": "observation.state", + "action_key": "action", + "action_type": "EE_6D_POS", + "input_type": "Dict", + "post_norm": false, + "relative_to_state": true + } + ], + "observation.image.0": [ + { + "type": "random_resized_crop", + "height": 224, + "width": 224, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "observation.image.1": [ + { + "type": "random_resized_crop", + "height": 224, + "width": 224, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "observation.image.2": [ + { + "type": "random_resized_crop", + "height": 224, + "width": 224, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ] + }, + "language_instruction": null, + "training_mode": "ROBOT_FLOWMATCH", + "action_type": "EE_EULER_POS", + "serialize_stats": false, + "video_decoder_cache_size": null, + "repo_id": "dlr_edan_shared_control", + "streaming": false, + "tolerance_s": 0.1, + "use_imagenet_stats": false, + "split_name": "train", + "prefetch_factor": 2, + "max_episodes": null + }, + "weight": 1.0 + }, + { + "dataset": { + "batch_size": 64, + "num_workers": 8, + "shuffle_buffer_size": 1000, + "normalization_mapping": { + "ACTION": "ACTIONCHUNK_QUANTILE", + "ENV": "QUANTILE", + "STATE": "QUANTILE", + "VISUAL": "IDENTITY" + }, + "features": { + "action": { + "type": "ACTION", + "shape": "(20,)" + }, + "observation.image.0": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.1": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.2": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.state": { + "type": "STATE", + "shape": "(20,)" + } + }, + "stats": null, + "clip_values": { + "ENV": [ + -1.05, + 1.05 + ], + "STATE": [ + -1.05, + 1.05 + ] + }, + "chunk_size": 50, + "observation_mapping": { + "action.ee_rpy_pos": "action", + "observation.ee_rpy_pos": "observation.state", + "observation.images.image": "observation.image.0", + "observation.images.wrist_image": "observation.image.1" + }, + "observation_whitelist": [ + "observation.state", + "observation.image.0", + "observation.image.1", + "observation.image.2", + "action", + "task", + "subtask", + "dataset_index", + "training_mode" + ], + "observation_blacklist": null, + "transform_mapping": { + "action": [ + { + "type": "convert_to_6d_actions", + "action_key": "action", + "action_type": "EE_EULER_POS", + "input_type": "Dict", + "post_norm": false + }, + { + "type": "convert_to_6d_actions", + "action_key": "observation.state", + "action_type": "EE_EULER_POS", + "input_type": "Dict", + "post_norm": false + }, + { + "type": "delta_actions", + "state_key": "observation.state", + "action_key": "action", + "action_type": "EE_6D_POS", + "input_type": "Dict", + "post_norm": false, + "relative_to_state": true + } + ], + "observation.image.0": [ + { + "type": "random_resized_crop", + "height": 224, + "width": 224, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "observation.image.1": [ + { + "type": "random_resized_crop", + "height": 224, + "width": 224, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "observation.image.2": [ + { + "type": "random_resized_crop", + "height": 224, + "width": 224, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ] + }, + "language_instruction": null, + "training_mode": "ROBOT_FLOWMATCH", + "action_type": "EE_EULER_POS", + "serialize_stats": false, + "video_decoder_cache_size": null, + "repo_id": "iamlab_cmu_pickup_insert", + "streaming": false, + "tolerance_s": 0.1, + "use_imagenet_stats": false, + "split_name": "train", + "prefetch_factor": 2, + "max_episodes": null + }, + "weight": 1.0 + }, + { + "dataset": { + "batch_size": 64, + "num_workers": 8, + "shuffle_buffer_size": 1000, + "normalization_mapping": { + "ACTION": "ACTIONCHUNK_QUANTILE", + "ENV": "QUANTILE", + "STATE": "QUANTILE", + "VISUAL": "IDENTITY" + }, + "features": { + "action": { + "type": "ACTION", + "shape": "(20,)" + }, + "observation.image.0": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.1": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.2": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.state": { + "type": "STATE", + "shape": "(20,)" + } + }, + "stats": null, + "clip_values": { + "ENV": [ + -1.05, + 1.05 + ], + "STATE": [ + -1.05, + 1.05 + ] + }, + "chunk_size": 50, + "observation_mapping": { + "action.ee_rpy_pos": "action", + "observation.ee_rpy_pos": "observation.state", + "observation.images.image": "observation.image.0", + "observation.images.wrist_image": "observation.image.2" + }, + "observation_whitelist": [ + "observation.state", + "observation.image.0", + "observation.image.1", + "observation.image.2", + "action", + "task", + "subtask", + "dataset_index", + "training_mode" + ], + "observation_blacklist": null, + "transform_mapping": { + "action": [ + { + "type": "convert_to_6d_actions", + "action_key": "action", + "action_type": "EE_EULER_POS", + "input_type": "Dict", + "post_norm": false + }, + { + "type": "convert_to_6d_actions", + "action_key": "observation.state", + "action_type": "EE_EULER_POS", + "input_type": "Dict", + "post_norm": false + }, + { + "type": "delta_actions", + "state_key": "observation.state", + "action_key": "action", + "action_type": "EE_6D_POS", + "input_type": "Dict", + "post_norm": false, + "relative_to_state": true + } + ], + "observation.image.0": [ + { + "type": "random_resized_crop", + "height": 224, + "width": 224, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "observation.image.1": [ + { + "type": "random_resized_crop", + "height": 224, + "width": 224, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "observation.image.2": [ + { + "type": "random_resized_crop", + "height": 224, + "width": 224, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ] + }, + "language_instruction": null, + "training_mode": "ROBOT_FLOWMATCH", + "action_type": "EE_EULER_POS", + "serialize_stats": false, + "video_decoder_cache_size": null, + "repo_id": "utaustin_mutex", + "streaming": false, + "tolerance_s": 0.1, + "use_imagenet_stats": false, + "split_name": "train", + "prefetch_factor": 2, + "max_episodes": null + }, + "weight": 1.0 + }, + { + "dataset": { + "batch_size": 64, + "num_workers": 8, + "shuffle_buffer_size": 1000, + "normalization_mapping": { + "ACTION": "ACTIONCHUNK_QUANTILE", + "ENV": "QUANTILE", + "STATE": "QUANTILE", + "VISUAL": "IDENTITY" + }, + "features": { + "action": { + "type": "ACTION", + "shape": "(20,)" + }, + "observation.image.0": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.1": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.2": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.state": { + "type": "STATE", + "shape": "(20,)" + } + }, + "stats": null, + "clip_values": { + "ENV": [ + -1.05, + 1.05 + ], + "STATE": [ + -1.05, + 1.05 + ] + }, + "chunk_size": 50, + "observation_mapping": { + "action.ee_rpy_pos": "action", + "observation.ee_rpy_pos": "observation.state", + "observation.images.image_side_1": "observation.image.0", + "observation.images.image_wrist_1": "observation.image.1" + }, + "observation_whitelist": [ + "observation.state", + "observation.image.0", + "observation.image.1", + "observation.image.2", + "action", + "task", + "subtask", + "dataset_index", + "training_mode" + ], + "observation_blacklist": [ + "observation.images.image_side_2", + "observation.images.image_wrist_2" + ], + "transform_mapping": { + "action": [ + { + "type": "convert_to_6d_actions", + "action_key": "action", + "action_type": "EE_EULER_POS", + "input_type": "Dict", + "post_norm": false + }, + { + "type": "convert_to_6d_actions", + "action_key": "observation.state", + "action_type": "EE_EULER_POS", + "input_type": "Dict", + "post_norm": false + }, + { + "type": "delta_actions", + "state_key": "observation.state", + "action_key": "action", + "action_type": "EE_6D_POS", + "input_type": "Dict", + "post_norm": false, + "relative_to_state": true + } + ], + "observation.image.0": [ + { + "type": "random_resized_crop", + "height": 224, + "width": 224, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "observation.image.1": [ + { + "type": "random_resized_crop", + "height": 224, + "width": 224, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "observation.image.2": [ + { + "type": "random_resized_crop", + "height": 224, + "width": 224, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ] + }, + "language_instruction": null, + "training_mode": "ROBOT_FLOWMATCH", + "action_type": "EE_EULER_POS", + "serialize_stats": false, + "video_decoder_cache_size": null, + "repo_id": "fmb", + "streaming": false, + "tolerance_s": 0.1, + "use_imagenet_stats": false, + "split_name": "train", + "prefetch_factor": 2, + "max_episodes": null + }, + "weight": 1.0 + }, + { + "dataset": { + "batch_size": 64, + "num_workers": 8, + "shuffle_buffer_size": 1000, + "normalization_mapping": { + "ACTION": "ACTIONCHUNK_QUANTILE", + "ENV": "QUANTILE", + "STATE": "QUANTILE", + "VISUAL": "IDENTITY" + }, + "features": { + "action": { + "type": "ACTION", + "shape": "(20,)" + }, + "observation.image.0": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.1": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.2": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.state": { + "type": "STATE", + "shape": "(20,)" + } + }, + "stats": null, + "clip_values": { + "ENV": [ + -1.05, + 1.05 + ], + "STATE": [ + -1.05, + 1.05 + ] + }, + "chunk_size": 50, + "observation_mapping": { + "action.ee_rpy_pos": "action", + "observation.ee_rpy_pos": "observation.state", + "observation.images.wrist_image": "observation.image.1" + }, + "observation_whitelist": [ + "observation.state", + "observation.image.0", + "observation.image.1", + "observation.image.2", + "action", + "task", + "subtask", + "dataset_index", + "training_mode" + ], + "observation_blacklist": null, + "transform_mapping": { + "action": [ + { + "type": "convert_to_6d_actions", + "action_key": "action", + "action_type": "EE_EULER_POS", + "input_type": "Dict", + "post_norm": false + }, + { + "type": "convert_to_6d_actions", + "action_key": "observation.state", + "action_type": "EE_EULER_POS", + "input_type": "Dict", + "post_norm": false + }, + { + "type": "delta_actions", + "state_key": "observation.state", + "action_key": "action", + "action_type": "EE_6D_POS", + "input_type": "Dict", + "post_norm": false, + "relative_to_state": true + } + ], + "observation.image.0": [ + { + "type": "random_resized_crop", + "height": 224, + "width": 224, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "observation.image.1": [ + { + "type": "random_resized_crop", + "height": 224, + "width": 224, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "observation.image.2": [ + { + "type": "random_resized_crop", + "height": 224, + "width": 224, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ] + }, + "language_instruction": null, + "training_mode": "ROBOT_FLOWMATCH", + "action_type": "EE_EULER_POS", + "serialize_stats": false, + "video_decoder_cache_size": null, + "repo_id": "dobbe", + "streaming": false, + "tolerance_s": 0.1, + "use_imagenet_stats": false, + "split_name": "train", + "prefetch_factor": 2, + "max_episodes": null + }, + "weight": 0.1 + }, + { + "dataset": { + "batch_size": 64, + "num_workers": 8, + "shuffle_buffer_size": 1000, + "normalization_mapping": { + "ACTION": "ACTIONCHUNK_QUANTILE", + "ENV": "QUANTILE", + "STATE": "QUANTILE", + "VISUAL": "IDENTITY" + }, + "features": { + "action": { + "type": "ACTION", + "shape": "(20,)" + }, + "observation.image.0": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.1": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.2": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.state": { + "type": "STATE", + "shape": "(20,)" + } + }, + "stats": null, + "clip_values": { + "ENV": [ + -1.05, + 1.05 + ], + "STATE": [ + -1.05, + 1.05 + ] + }, + "chunk_size": 50, + "observation_mapping": { + "action.ee_rpy_pos": "action", + "observation.ee_rpy_pos": "observation.state", + "observation.images.wrist_image": "observation.image.2" + }, + "observation_whitelist": [ + "observation.state", + "observation.image.0", + "observation.image.1", + "observation.image.2", + "action", + "task", + "subtask", + "dataset_index", + "training_mode" + ], + "observation_blacklist": null, + "transform_mapping": { + "action": [ + { + "type": "convert_to_6d_actions", + "action_key": "action", + "action_type": "EE_EULER_POS", + "input_type": "Dict", + "post_norm": false + }, + { + "type": "convert_to_6d_actions", + "action_key": "observation.state", + "action_type": "EE_EULER_POS", + "input_type": "Dict", + "post_norm": false + }, + { + "type": "delta_actions", + "state_key": "observation.state", + "action_key": "action", + "action_type": "EE_6D_POS", + "input_type": "Dict", + "post_norm": false, + "relative_to_state": true + } + ], + "observation.image.0": [ + { + "type": "random_resized_crop", + "height": 224, + "width": 224, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "observation.image.1": [ + { + "type": "random_resized_crop", + "height": 224, + "width": 224, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "observation.image.2": [ + { + "type": "random_resized_crop", + "height": 224, + "width": 224, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ] + }, + "language_instruction": null, + "training_mode": "ROBOT_FLOWMATCH", + "action_type": "EE_EULER_POS", + "serialize_stats": false, + "video_decoder_cache_size": null, + "repo_id": "dobbe", + "streaming": false, + "tolerance_s": 0.1, + "use_imagenet_stats": false, + "split_name": "train", + "prefetch_factor": 2, + "max_episodes": null + }, + "weight": 0.1 + } + ], + "split_name": null, + "shuffle_type": "FRAME", + "seed": 12345, + "num_workers": 1, + "prefetch_factor": 2, + "batch_size": 64, + "streaming": false, + "shuffle_buffer_size": null, + "vl_batch_size": null, + "observation_whitelist": [ + "observation.state", + "observation.image.0", + "observation.image.1", + "observation.image.2", + "action", + "task", + "subtask", + "task_index", + "frame_index", + "episode_index" + ], + "transform_mapping": null, + "chunk_size": null, + "flatten_nested": true, + "serialize_stats": true, + "video_decoder_cache_size": null, + "training_mode_override": null + }, + "weight": 0.072628397212 + }, + { + "dataset": { + "datasets": [ + { + "dataset": { + "datasets": [ + { + "dataset": { + "datasets": [ + { + "dataset": { + "batch_size": 16, + "num_workers": 8, + "shuffle_buffer_size": 100, + "normalization_mapping": { + "ACTION": "ACTIONCHUNK_QUANTILE", + "ENV": "QUANTILE", + "STATE": "QUANTILE", + "VISUAL": "IDENTITY" + }, + "features": { + "action": { + "type": "ACTION", + "shape": "(20,)" + }, + "observation.image.0": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.1": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.2": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.state": { + "type": "STATE", + "shape": "(20,)" + } + }, + "stats": null, + "clip_values": { + "ENV": [ + -1.05, + 1.05 + ], + "STATE": [ + -1.05, + 1.05 + ] + }, + "chunk_size": 50, + "observation_mapping": { + "action.ee_quat_pos": "action", + "observations.ee_quat_pos": "observation.state", + "observations.images.cam_left": "observation.image.1", + "observations.images.cam_right": "observation.image.2", + "observations.images.cam_top": "observation.image.0" + }, + "observation_whitelist": [ + "observation.state", + "observation.image.0", + "observation.image.1", + "observation.image.2", + "action", + "task", + "subtask", + "dataset_index", + "training_mode" + ], + "observation_blacklist": null, + "transform_mapping": { + "action": [ + { + "type": "convert_to_6d_actions", + "action_key": "action", + "action_type": "EE_QUAT_POS_XYZW", + "input_type": "Dict", + "post_norm": false + }, + { + "type": "convert_to_6d_actions", + "action_key": "observation.state", + "action_type": "EE_QUAT_POS_XYZW", + "input_type": "Dict", + "post_norm": false + }, + { + "type": "delta_actions", + "state_key": "observation.state", + "action_key": "action", + "action_type": "EE_6D_POS", + "input_type": "Dict", + "post_norm": false, + "relative_to_state": true + } + ], + "observation.image.0": [ + { + "type": "random_resized_crop", + "height": 448, + "width": 448, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "observation.image.1": [ + { + "type": "random_resized_crop", + "height": 448, + "width": 448, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "observation.image.2": [ + { + "type": "random_resized_crop", + "height": 448, + "width": 448, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "task": [ + { + "type": "task_description_selector", + "task_key": "task", + "subtask_key": "subtask", + "output_key": "task", + "subtask_probability": 0.98, + "drop_subtask_key": true, + "has_subtask_key": "__task_description_has_subtask__", + "used_subtask_key": "__task_description_used_subtask__", + "input_type": "Dict", + "post_norm": false + } + ] + }, + "language_instruction": null, + "training_mode": "ROBOT_FLOWMATCH", + "action_type": "EE_QUAT_POS_XYZW", + "serialize_stats": false, + "video_decoder_cache_size": null, + "repo_id": "xdof/xmi_merged/split_00", + "streaming": false, + "tolerance_s": 0.1, + "use_imagenet_stats": false, + "split_name": "train", + "prefetch_factor": 2, + "max_episodes": null + }, + "weight": 0.062442955152 + }, + { + "dataset": { + "batch_size": 16, + "num_workers": 8, + "shuffle_buffer_size": 100, + "normalization_mapping": { + "ACTION": "ACTIONCHUNK_QUANTILE", + "ENV": "QUANTILE", + "STATE": "QUANTILE", + "VISUAL": "IDENTITY" + }, + "features": { + "action": { + "type": "ACTION", + "shape": "(20,)" + }, + "observation.image.0": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.1": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.2": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.state": { + "type": "STATE", + "shape": "(20,)" + } + }, + "stats": null, + "clip_values": { + "ENV": [ + -1.05, + 1.05 + ], + "STATE": [ + -1.05, + 1.05 + ] + }, + "chunk_size": 50, + "observation_mapping": { + "action.ee_quat_pos": "action", + "observations.ee_quat_pos": "observation.state", + "observations.images.cam_left": "observation.image.1", + "observations.images.cam_right": "observation.image.2", + "observations.images.cam_top": "observation.image.0" + }, + "observation_whitelist": [ + "observation.state", + "observation.image.0", + "observation.image.1", + "observation.image.2", + "action", + "task", + "subtask", + "dataset_index", + "training_mode" + ], + "observation_blacklist": null, + "transform_mapping": { + "action": [ + { + "type": "convert_to_6d_actions", + "action_key": "action", + "action_type": "EE_QUAT_POS_XYZW", + "input_type": "Dict", + "post_norm": false + }, + { + "type": "convert_to_6d_actions", + "action_key": "observation.state", + "action_type": "EE_QUAT_POS_XYZW", + "input_type": "Dict", + "post_norm": false + }, + { + "type": "delta_actions", + "state_key": "observation.state", + "action_key": "action", + "action_type": "EE_6D_POS", + "input_type": "Dict", + "post_norm": false, + "relative_to_state": true + } + ], + "observation.image.0": [ + { + "type": "random_resized_crop", + "height": 448, + "width": 448, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "observation.image.1": [ + { + "type": "random_resized_crop", + "height": 448, + "width": 448, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "observation.image.2": [ + { + "type": "random_resized_crop", + "height": 448, + "width": 448, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "task": [ + { + "type": "task_description_selector", + "task_key": "task", + "subtask_key": "subtask", + "output_key": "task", + "subtask_probability": 0.98, + "drop_subtask_key": true, + "has_subtask_key": "__task_description_has_subtask__", + "used_subtask_key": "__task_description_used_subtask__", + "input_type": "Dict", + "post_norm": false + } + ] + }, + "language_instruction": null, + "training_mode": "ROBOT_FLOWMATCH", + "action_type": "EE_QUAT_POS_XYZW", + "serialize_stats": false, + "video_decoder_cache_size": null, + "repo_id": "xdof/xmi_merged/split_01", + "streaming": false, + "tolerance_s": 0.1, + "use_imagenet_stats": false, + "split_name": "train", + "prefetch_factor": 2, + "max_episodes": null + }, + "weight": 0.061575011137 + }, + { + "dataset": { + "batch_size": 16, + "num_workers": 8, + "shuffle_buffer_size": 100, + "normalization_mapping": { + "ACTION": "ACTIONCHUNK_QUANTILE", + "ENV": "QUANTILE", + "STATE": "QUANTILE", + "VISUAL": "IDENTITY" + }, + "features": { + "action": { + "type": "ACTION", + "shape": "(20,)" + }, + "observation.image.0": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.1": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.2": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.state": { + "type": "STATE", + "shape": "(20,)" + } + }, + "stats": null, + "clip_values": { + "ENV": [ + -1.05, + 1.05 + ], + "STATE": [ + -1.05, + 1.05 + ] + }, + "chunk_size": 50, + "observation_mapping": { + "action.ee_quat_pos": "action", + "observations.ee_quat_pos": "observation.state", + "observations.images.cam_left": "observation.image.1", + "observations.images.cam_right": "observation.image.2", + "observations.images.cam_top": "observation.image.0" + }, + "observation_whitelist": [ + "observation.state", + "observation.image.0", + "observation.image.1", + "observation.image.2", + "action", + "task", + "subtask", + "dataset_index", + "training_mode" + ], + "observation_blacklist": null, + "transform_mapping": { + "action": [ + { + "type": "convert_to_6d_actions", + "action_key": "action", + "action_type": "EE_QUAT_POS_XYZW", + "input_type": "Dict", + "post_norm": false + }, + { + "type": "convert_to_6d_actions", + "action_key": "observation.state", + "action_type": "EE_QUAT_POS_XYZW", + "input_type": "Dict", + "post_norm": false + }, + { + "type": "delta_actions", + "state_key": "observation.state", + "action_key": "action", + "action_type": "EE_6D_POS", + "input_type": "Dict", + "post_norm": false, + "relative_to_state": true + } + ], + "observation.image.0": [ + { + "type": "random_resized_crop", + "height": 448, + "width": 448, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "observation.image.1": [ + { + "type": "random_resized_crop", + "height": 448, + "width": 448, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "observation.image.2": [ + { + "type": "random_resized_crop", + "height": 448, + "width": 448, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "task": [ + { + "type": "task_description_selector", + "task_key": "task", + "subtask_key": "subtask", + "output_key": "task", + "subtask_probability": 0.98, + "drop_subtask_key": true, + "has_subtask_key": "__task_description_has_subtask__", + "used_subtask_key": "__task_description_used_subtask__", + "input_type": "Dict", + "post_norm": false + } + ] + }, + "language_instruction": null, + "training_mode": "ROBOT_FLOWMATCH", + "action_type": "EE_QUAT_POS_XYZW", + "serialize_stats": false, + "video_decoder_cache_size": null, + "repo_id": "xdof/xmi_merged/split_02", + "streaming": false, + "tolerance_s": 0.1, + "use_imagenet_stats": false, + "split_name": "train", + "prefetch_factor": 2, + "max_episodes": null + }, + "weight": 0.061211691536 + }, + { + "dataset": { + "batch_size": 16, + "num_workers": 8, + "shuffle_buffer_size": 100, + "normalization_mapping": { + "ACTION": "ACTIONCHUNK_QUANTILE", + "ENV": "QUANTILE", + "STATE": "QUANTILE", + "VISUAL": "IDENTITY" + }, + "features": { + "action": { + "type": "ACTION", + "shape": "(20,)" + }, + "observation.image.0": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.1": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.2": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.state": { + "type": "STATE", + "shape": "(20,)" + } + }, + "stats": null, + "clip_values": { + "ENV": [ + -1.05, + 1.05 + ], + "STATE": [ + -1.05, + 1.05 + ] + }, + "chunk_size": 50, + "observation_mapping": { + "action.ee_quat_pos": "action", + "observations.ee_quat_pos": "observation.state", + "observations.images.cam_left": "observation.image.1", + "observations.images.cam_right": "observation.image.2", + "observations.images.cam_top": "observation.image.0" + }, + "observation_whitelist": [ + "observation.state", + "observation.image.0", + "observation.image.1", + "observation.image.2", + "action", + "task", + "subtask", + "dataset_index", + "training_mode" + ], + "observation_blacklist": null, + "transform_mapping": { + "action": [ + { + "type": "convert_to_6d_actions", + "action_key": "action", + "action_type": "EE_QUAT_POS_XYZW", + "input_type": "Dict", + "post_norm": false + }, + { + "type": "convert_to_6d_actions", + "action_key": "observation.state", + "action_type": "EE_QUAT_POS_XYZW", + "input_type": "Dict", + "post_norm": false + }, + { + "type": "delta_actions", + "state_key": "observation.state", + "action_key": "action", + "action_type": "EE_6D_POS", + "input_type": "Dict", + "post_norm": false, + "relative_to_state": true + } + ], + "observation.image.0": [ + { + "type": "random_resized_crop", + "height": 448, + "width": 448, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "observation.image.1": [ + { + "type": "random_resized_crop", + "height": 448, + "width": 448, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "observation.image.2": [ + { + "type": "random_resized_crop", + "height": 448, + "width": 448, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "task": [ + { + "type": "task_description_selector", + "task_key": "task", + "subtask_key": "subtask", + "output_key": "task", + "subtask_probability": 0.98, + "drop_subtask_key": true, + "has_subtask_key": "__task_description_has_subtask__", + "used_subtask_key": "__task_description_used_subtask__", + "input_type": "Dict", + "post_norm": false + } + ] + }, + "language_instruction": null, + "training_mode": "ROBOT_FLOWMATCH", + "action_type": "EE_QUAT_POS_XYZW", + "serialize_stats": false, + "video_decoder_cache_size": null, + "repo_id": "xdof/xmi_merged/split_03", + "streaming": false, + "tolerance_s": 0.1, + "use_imagenet_stats": false, + "split_name": "train", + "prefetch_factor": 2, + "max_episodes": null + }, + "weight": 0.061824178684 + }, + { + "dataset": { + "batch_size": 16, + "num_workers": 8, + "shuffle_buffer_size": 100, + "normalization_mapping": { + "ACTION": "ACTIONCHUNK_QUANTILE", + "ENV": "QUANTILE", + "STATE": "QUANTILE", + "VISUAL": "IDENTITY" + }, + "features": { + "action": { + "type": "ACTION", + "shape": "(20,)" + }, + "observation.image.0": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.1": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.2": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.state": { + "type": "STATE", + "shape": "(20,)" + } + }, + "stats": null, + "clip_values": { + "ENV": [ + -1.05, + 1.05 + ], + "STATE": [ + -1.05, + 1.05 + ] + }, + "chunk_size": 50, + "observation_mapping": { + "action.ee_quat_pos": "action", + "observations.ee_quat_pos": "observation.state", + "observations.images.cam_left": "observation.image.1", + "observations.images.cam_right": "observation.image.2", + "observations.images.cam_top": "observation.image.0" + }, + "observation_whitelist": [ + "observation.state", + "observation.image.0", + "observation.image.1", + "observation.image.2", + "action", + "task", + "subtask", + "dataset_index", + "training_mode" + ], + "observation_blacklist": null, + "transform_mapping": { + "action": [ + { + "type": "convert_to_6d_actions", + "action_key": "action", + "action_type": "EE_QUAT_POS_XYZW", + "input_type": "Dict", + "post_norm": false + }, + { + "type": "convert_to_6d_actions", + "action_key": "observation.state", + "action_type": "EE_QUAT_POS_XYZW", + "input_type": "Dict", + "post_norm": false + }, + { + "type": "delta_actions", + "state_key": "observation.state", + "action_key": "action", + "action_type": "EE_6D_POS", + "input_type": "Dict", + "post_norm": false, + "relative_to_state": true + } + ], + "observation.image.0": [ + { + "type": "random_resized_crop", + "height": 448, + "width": 448, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "observation.image.1": [ + { + "type": "random_resized_crop", + "height": 448, + "width": 448, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "observation.image.2": [ + { + "type": "random_resized_crop", + "height": 448, + "width": 448, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "task": [ + { + "type": "task_description_selector", + "task_key": "task", + "subtask_key": "subtask", + "output_key": "task", + "subtask_probability": 0.98, + "drop_subtask_key": true, + "has_subtask_key": "__task_description_has_subtask__", + "used_subtask_key": "__task_description_used_subtask__", + "input_type": "Dict", + "post_norm": false + } + ] + }, + "language_instruction": null, + "training_mode": "ROBOT_FLOWMATCH", + "action_type": "EE_QUAT_POS_XYZW", + "serialize_stats": false, + "video_decoder_cache_size": null, + "repo_id": "xdof/xmi_merged/split_04", + "streaming": false, + "tolerance_s": 0.1, + "use_imagenet_stats": false, + "split_name": "train", + "prefetch_factor": 2, + "max_episodes": null + }, + "weight": 0.062086690993 + }, + { + "dataset": { + "batch_size": 16, + "num_workers": 8, + "shuffle_buffer_size": 100, + "normalization_mapping": { + "ACTION": "ACTIONCHUNK_QUANTILE", + "ENV": "QUANTILE", + "STATE": "QUANTILE", + "VISUAL": "IDENTITY" + }, + "features": { + "action": { + "type": "ACTION", + "shape": "(20,)" + }, + "observation.image.0": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.1": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.2": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.state": { + "type": "STATE", + "shape": "(20,)" + } + }, + "stats": null, + "clip_values": { + "ENV": [ + -1.05, + 1.05 + ], + "STATE": [ + -1.05, + 1.05 + ] + }, + "chunk_size": 50, + "observation_mapping": { + "action.ee_quat_pos": "action", + "observations.ee_quat_pos": "observation.state", + "observations.images.cam_left": "observation.image.1", + "observations.images.cam_right": "observation.image.2", + "observations.images.cam_top": "observation.image.0" + }, + "observation_whitelist": [ + "observation.state", + "observation.image.0", + "observation.image.1", + "observation.image.2", + "action", + "task", + "subtask", + "dataset_index", + "training_mode" + ], + "observation_blacklist": null, + "transform_mapping": { + "action": [ + { + "type": "convert_to_6d_actions", + "action_key": "action", + "action_type": "EE_QUAT_POS_XYZW", + "input_type": "Dict", + "post_norm": false + }, + { + "type": "convert_to_6d_actions", + "action_key": "observation.state", + "action_type": "EE_QUAT_POS_XYZW", + "input_type": "Dict", + "post_norm": false + }, + { + "type": "delta_actions", + "state_key": "observation.state", + "action_key": "action", + "action_type": "EE_6D_POS", + "input_type": "Dict", + "post_norm": false, + "relative_to_state": true + } + ], + "observation.image.0": [ + { + "type": "random_resized_crop", + "height": 448, + "width": 448, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "observation.image.1": [ + { + "type": "random_resized_crop", + "height": 448, + "width": 448, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "observation.image.2": [ + { + "type": "random_resized_crop", + "height": 448, + "width": 448, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "task": [ + { + "type": "task_description_selector", + "task_key": "task", + "subtask_key": "subtask", + "output_key": "task", + "subtask_probability": 0.98, + "drop_subtask_key": true, + "has_subtask_key": "__task_description_has_subtask__", + "used_subtask_key": "__task_description_used_subtask__", + "input_type": "Dict", + "post_norm": false + } + ] + }, + "language_instruction": null, + "training_mode": "ROBOT_FLOWMATCH", + "action_type": "EE_QUAT_POS_XYZW", + "serialize_stats": false, + "video_decoder_cache_size": null, + "repo_id": "xdof/xmi_merged/split_05", + "streaming": false, + "tolerance_s": 0.1, + "use_imagenet_stats": false, + "split_name": "train", + "prefetch_factor": 2, + "max_episodes": null + }, + "weight": 0.061719344802 + }, + { + "dataset": { + "batch_size": 16, + "num_workers": 8, + "shuffle_buffer_size": 100, + "normalization_mapping": { + "ACTION": "ACTIONCHUNK_QUANTILE", + "ENV": "QUANTILE", + "STATE": "QUANTILE", + "VISUAL": "IDENTITY" + }, + "features": { + "action": { + "type": "ACTION", + "shape": "(20,)" + }, + "observation.image.0": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.1": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.2": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.state": { + "type": "STATE", + "shape": "(20,)" + } + }, + "stats": null, + "clip_values": { + "ENV": [ + -1.05, + 1.05 + ], + "STATE": [ + -1.05, + 1.05 + ] + }, + "chunk_size": 50, + "observation_mapping": { + "action.ee_quat_pos": "action", + "observations.ee_quat_pos": "observation.state", + "observations.images.cam_left": "observation.image.1", + "observations.images.cam_right": "observation.image.2", + "observations.images.cam_top": "observation.image.0" + }, + "observation_whitelist": [ + "observation.state", + "observation.image.0", + "observation.image.1", + "observation.image.2", + "action", + "task", + "subtask", + "dataset_index", + "training_mode" + ], + "observation_blacklist": null, + "transform_mapping": { + "action": [ + { + "type": "convert_to_6d_actions", + "action_key": "action", + "action_type": "EE_QUAT_POS_XYZW", + "input_type": "Dict", + "post_norm": false + }, + { + "type": "convert_to_6d_actions", + "action_key": "observation.state", + "action_type": "EE_QUAT_POS_XYZW", + "input_type": "Dict", + "post_norm": false + }, + { + "type": "delta_actions", + "state_key": "observation.state", + "action_key": "action", + "action_type": "EE_6D_POS", + "input_type": "Dict", + "post_norm": false, + "relative_to_state": true + } + ], + "observation.image.0": [ + { + "type": "random_resized_crop", + "height": 448, + "width": 448, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "observation.image.1": [ + { + "type": "random_resized_crop", + "height": 448, + "width": 448, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "observation.image.2": [ + { + "type": "random_resized_crop", + "height": 448, + "width": 448, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "task": [ + { + "type": "task_description_selector", + "task_key": "task", + "subtask_key": "subtask", + "output_key": "task", + "subtask_probability": 0.98, + "drop_subtask_key": true, + "has_subtask_key": "__task_description_has_subtask__", + "used_subtask_key": "__task_description_used_subtask__", + "input_type": "Dict", + "post_norm": false + } + ] + }, + "language_instruction": null, + "training_mode": "ROBOT_FLOWMATCH", + "action_type": "EE_QUAT_POS_XYZW", + "serialize_stats": false, + "video_decoder_cache_size": null, + "repo_id": "xdof/xmi_merged/split_06", + "streaming": false, + "tolerance_s": 0.1, + "use_imagenet_stats": false, + "split_name": "train", + "prefetch_factor": 2, + "max_episodes": null + }, + "weight": 0.063928945148 + }, + { + "dataset": { + "batch_size": 16, + "num_workers": 8, + "shuffle_buffer_size": 100, + "normalization_mapping": { + "ACTION": "ACTIONCHUNK_QUANTILE", + "ENV": "QUANTILE", + "STATE": "QUANTILE", + "VISUAL": "IDENTITY" + }, + "features": { + "action": { + "type": "ACTION", + "shape": "(20,)" + }, + "observation.image.0": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.1": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.2": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.state": { + "type": "STATE", + "shape": "(20,)" + } + }, + "stats": null, + "clip_values": { + "ENV": [ + -1.05, + 1.05 + ], + "STATE": [ + -1.05, + 1.05 + ] + }, + "chunk_size": 50, + "observation_mapping": { + "action.ee_quat_pos": "action", + "observations.ee_quat_pos": "observation.state", + "observations.images.cam_left": "observation.image.1", + "observations.images.cam_right": "observation.image.2", + "observations.images.cam_top": "observation.image.0" + }, + "observation_whitelist": [ + "observation.state", + "observation.image.0", + "observation.image.1", + "observation.image.2", + "action", + "task", + "subtask", + "dataset_index", + "training_mode" + ], + "observation_blacklist": null, + "transform_mapping": { + "action": [ + { + "type": "convert_to_6d_actions", + "action_key": "action", + "action_type": "EE_QUAT_POS_XYZW", + "input_type": "Dict", + "post_norm": false + }, + { + "type": "convert_to_6d_actions", + "action_key": "observation.state", + "action_type": "EE_QUAT_POS_XYZW", + "input_type": "Dict", + "post_norm": false + }, + { + "type": "delta_actions", + "state_key": "observation.state", + "action_key": "action", + "action_type": "EE_6D_POS", + "input_type": "Dict", + "post_norm": false, + "relative_to_state": true + } + ], + "observation.image.0": [ + { + "type": "random_resized_crop", + "height": 448, + "width": 448, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "observation.image.1": [ + { + "type": "random_resized_crop", + "height": 448, + "width": 448, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "observation.image.2": [ + { + "type": "random_resized_crop", + "height": 448, + "width": 448, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "task": [ + { + "type": "task_description_selector", + "task_key": "task", + "subtask_key": "subtask", + "output_key": "task", + "subtask_probability": 0.98, + "drop_subtask_key": true, + "has_subtask_key": "__task_description_has_subtask__", + "used_subtask_key": "__task_description_used_subtask__", + "input_type": "Dict", + "post_norm": false + } + ] + }, + "language_instruction": null, + "training_mode": "ROBOT_FLOWMATCH", + "action_type": "EE_QUAT_POS_XYZW", + "serialize_stats": false, + "video_decoder_cache_size": null, + "repo_id": "xdof/xmi_merged/split_07", + "streaming": false, + "tolerance_s": 0.1, + "use_imagenet_stats": false, + "split_name": "train", + "prefetch_factor": 2, + "max_episodes": null + }, + "weight": 0.06259129759 + }, + { + "dataset": { + "batch_size": 16, + "num_workers": 8, + "shuffle_buffer_size": 100, + "normalization_mapping": { + "ACTION": "ACTIONCHUNK_QUANTILE", + "ENV": "QUANTILE", + "STATE": "QUANTILE", + "VISUAL": "IDENTITY" + }, + "features": { + "action": { + "type": "ACTION", + "shape": "(20,)" + }, + "observation.image.0": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.1": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.2": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.state": { + "type": "STATE", + "shape": "(20,)" + } + }, + "stats": null, + "clip_values": { + "ENV": [ + -1.05, + 1.05 + ], + "STATE": [ + -1.05, + 1.05 + ] + }, + "chunk_size": 50, + "observation_mapping": { + "action.ee_quat_pos": "action", + "observations.ee_quat_pos": "observation.state", + "observations.images.cam_left": "observation.image.1", + "observations.images.cam_right": "observation.image.2", + "observations.images.cam_top": "observation.image.0" + }, + "observation_whitelist": [ + "observation.state", + "observation.image.0", + "observation.image.1", + "observation.image.2", + "action", + "task", + "subtask", + "dataset_index", + "training_mode" + ], + "observation_blacklist": null, + "transform_mapping": { + "action": [ + { + "type": "convert_to_6d_actions", + "action_key": "action", + "action_type": "EE_QUAT_POS_XYZW", + "input_type": "Dict", + "post_norm": false + }, + { + "type": "convert_to_6d_actions", + "action_key": "observation.state", + "action_type": "EE_QUAT_POS_XYZW", + "input_type": "Dict", + "post_norm": false + }, + { + "type": "delta_actions", + "state_key": "observation.state", + "action_key": "action", + "action_type": "EE_6D_POS", + "input_type": "Dict", + "post_norm": false, + "relative_to_state": true + } + ], + "observation.image.0": [ + { + "type": "random_resized_crop", + "height": 448, + "width": 448, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "observation.image.1": [ + { + "type": "random_resized_crop", + "height": 448, + "width": 448, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "observation.image.2": [ + { + "type": "random_resized_crop", + "height": 448, + "width": 448, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "task": [ + { + "type": "task_description_selector", + "task_key": "task", + "subtask_key": "subtask", + "output_key": "task", + "subtask_probability": 0.98, + "drop_subtask_key": true, + "has_subtask_key": "__task_description_has_subtask__", + "used_subtask_key": "__task_description_used_subtask__", + "input_type": "Dict", + "post_norm": false + } + ] + }, + "language_instruction": null, + "training_mode": "ROBOT_FLOWMATCH", + "action_type": "EE_QUAT_POS_XYZW", + "serialize_stats": false, + "video_decoder_cache_size": null, + "repo_id": "xdof/xmi_merged/split_08", + "streaming": false, + "tolerance_s": 0.1, + "use_imagenet_stats": false, + "split_name": "train", + "prefetch_factor": 2, + "max_episodes": null + }, + "weight": 0.064312166083 + }, + { + "dataset": { + "batch_size": 16, + "num_workers": 8, + "shuffle_buffer_size": 100, + "normalization_mapping": { + "ACTION": "ACTIONCHUNK_QUANTILE", + "ENV": "QUANTILE", + "STATE": "QUANTILE", + "VISUAL": "IDENTITY" + }, + "features": { + "action": { + "type": "ACTION", + "shape": "(20,)" + }, + "observation.image.0": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.1": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.2": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.state": { + "type": "STATE", + "shape": "(20,)" + } + }, + "stats": null, + "clip_values": { + "ENV": [ + -1.05, + 1.05 + ], + "STATE": [ + -1.05, + 1.05 + ] + }, + "chunk_size": 50, + "observation_mapping": { + "action.ee_quat_pos": "action", + "observations.ee_quat_pos": "observation.state", + "observations.images.cam_left": "observation.image.1", + "observations.images.cam_right": "observation.image.2", + "observations.images.cam_top": "observation.image.0" + }, + "observation_whitelist": [ + "observation.state", + "observation.image.0", + "observation.image.1", + "observation.image.2", + "action", + "task", + "subtask", + "dataset_index", + "training_mode" + ], + "observation_blacklist": null, + "transform_mapping": { + "action": [ + { + "type": "convert_to_6d_actions", + "action_key": "action", + "action_type": "EE_QUAT_POS_XYZW", + "input_type": "Dict", + "post_norm": false + }, + { + "type": "convert_to_6d_actions", + "action_key": "observation.state", + "action_type": "EE_QUAT_POS_XYZW", + "input_type": "Dict", + "post_norm": false + }, + { + "type": "delta_actions", + "state_key": "observation.state", + "action_key": "action", + "action_type": "EE_6D_POS", + "input_type": "Dict", + "post_norm": false, + "relative_to_state": true + } + ], + "observation.image.0": [ + { + "type": "random_resized_crop", + "height": 448, + "width": 448, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "observation.image.1": [ + { + "type": "random_resized_crop", + "height": 448, + "width": 448, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "observation.image.2": [ + { + "type": "random_resized_crop", + "height": 448, + "width": 448, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "task": [ + { + "type": "task_description_selector", + "task_key": "task", + "subtask_key": "subtask", + "output_key": "task", + "subtask_probability": 0.98, + "drop_subtask_key": true, + "has_subtask_key": "__task_description_has_subtask__", + "used_subtask_key": "__task_description_used_subtask__", + "input_type": "Dict", + "post_norm": false + } + ] + }, + "language_instruction": null, + "training_mode": "ROBOT_FLOWMATCH", + "action_type": "EE_QUAT_POS_XYZW", + "serialize_stats": false, + "video_decoder_cache_size": null, + "repo_id": "xdof/xmi_merged/split_09", + "streaming": false, + "tolerance_s": 0.1, + "use_imagenet_stats": false, + "split_name": "train", + "prefetch_factor": 2, + "max_episodes": null + }, + "weight": 0.063727865067 + }, + { + "dataset": { + "batch_size": 16, + "num_workers": 8, + "shuffle_buffer_size": 100, + "normalization_mapping": { + "ACTION": "ACTIONCHUNK_QUANTILE", + "ENV": "QUANTILE", + "STATE": "QUANTILE", + "VISUAL": "IDENTITY" + }, + "features": { + "action": { + "type": "ACTION", + "shape": "(20,)" + }, + "observation.image.0": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.1": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.2": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.state": { + "type": "STATE", + "shape": "(20,)" + } + }, + "stats": null, + "clip_values": { + "ENV": [ + -1.05, + 1.05 + ], + "STATE": [ + -1.05, + 1.05 + ] + }, + "chunk_size": 50, + "observation_mapping": { + "action.ee_quat_pos": "action", + "observations.ee_quat_pos": "observation.state", + "observations.images.cam_left": "observation.image.1", + "observations.images.cam_right": "observation.image.2", + "observations.images.cam_top": "observation.image.0" + }, + "observation_whitelist": [ + "observation.state", + "observation.image.0", + "observation.image.1", + "observation.image.2", + "action", + "task", + "subtask", + "dataset_index", + "training_mode" + ], + "observation_blacklist": null, + "transform_mapping": { + "action": [ + { + "type": "convert_to_6d_actions", + "action_key": "action", + "action_type": "EE_QUAT_POS_XYZW", + "input_type": "Dict", + "post_norm": false + }, + { + "type": "convert_to_6d_actions", + "action_key": "observation.state", + "action_type": "EE_QUAT_POS_XYZW", + "input_type": "Dict", + "post_norm": false + }, + { + "type": "delta_actions", + "state_key": "observation.state", + "action_key": "action", + "action_type": "EE_6D_POS", + "input_type": "Dict", + "post_norm": false, + "relative_to_state": true + } + ], + "observation.image.0": [ + { + "type": "random_resized_crop", + "height": 448, + "width": 448, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "observation.image.1": [ + { + "type": "random_resized_crop", + "height": 448, + "width": 448, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "observation.image.2": [ + { + "type": "random_resized_crop", + "height": 448, + "width": 448, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "task": [ + { + "type": "task_description_selector", + "task_key": "task", + "subtask_key": "subtask", + "output_key": "task", + "subtask_probability": 0.98, + "drop_subtask_key": true, + "has_subtask_key": "__task_description_has_subtask__", + "used_subtask_key": "__task_description_used_subtask__", + "input_type": "Dict", + "post_norm": false + } + ] + }, + "language_instruction": null, + "training_mode": "ROBOT_FLOWMATCH", + "action_type": "EE_QUAT_POS_XYZW", + "serialize_stats": false, + "video_decoder_cache_size": null, + "repo_id": "xdof/xmi_merged/split_10", + "streaming": false, + "tolerance_s": 0.1, + "use_imagenet_stats": false, + "split_name": "train", + "prefetch_factor": 2, + "max_episodes": null + }, + "weight": 0.064219002185 + }, + { + "dataset": { + "batch_size": 16, + "num_workers": 8, + "shuffle_buffer_size": 100, + "normalization_mapping": { + "ACTION": "ACTIONCHUNK_QUANTILE", + "ENV": "QUANTILE", + "STATE": "QUANTILE", + "VISUAL": "IDENTITY" + }, + "features": { + "action": { + "type": "ACTION", + "shape": "(20,)" + }, + "observation.image.0": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.1": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.2": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.state": { + "type": "STATE", + "shape": "(20,)" + } + }, + "stats": null, + "clip_values": { + "ENV": [ + -1.05, + 1.05 + ], + "STATE": [ + -1.05, + 1.05 + ] + }, + "chunk_size": 50, + "observation_mapping": { + "action.ee_quat_pos": "action", + "observations.ee_quat_pos": "observation.state", + "observations.images.cam_left": "observation.image.1", + "observations.images.cam_right": "observation.image.2", + "observations.images.cam_top": "observation.image.0" + }, + "observation_whitelist": [ + "observation.state", + "observation.image.0", + "observation.image.1", + "observation.image.2", + "action", + "task", + "subtask", + "dataset_index", + "training_mode" + ], + "observation_blacklist": null, + "transform_mapping": { + "action": [ + { + "type": "convert_to_6d_actions", + "action_key": "action", + "action_type": "EE_QUAT_POS_XYZW", + "input_type": "Dict", + "post_norm": false + }, + { + "type": "convert_to_6d_actions", + "action_key": "observation.state", + "action_type": "EE_QUAT_POS_XYZW", + "input_type": "Dict", + "post_norm": false + }, + { + "type": "delta_actions", + "state_key": "observation.state", + "action_key": "action", + "action_type": "EE_6D_POS", + "input_type": "Dict", + "post_norm": false, + "relative_to_state": true + } + ], + "observation.image.0": [ + { + "type": "random_resized_crop", + "height": 448, + "width": 448, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "observation.image.1": [ + { + "type": "random_resized_crop", + "height": 448, + "width": 448, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "observation.image.2": [ + { + "type": "random_resized_crop", + "height": 448, + "width": 448, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "task": [ + { + "type": "task_description_selector", + "task_key": "task", + "subtask_key": "subtask", + "output_key": "task", + "subtask_probability": 0.98, + "drop_subtask_key": true, + "has_subtask_key": "__task_description_has_subtask__", + "used_subtask_key": "__task_description_used_subtask__", + "input_type": "Dict", + "post_norm": false + } + ] + }, + "language_instruction": null, + "training_mode": "ROBOT_FLOWMATCH", + "action_type": "EE_QUAT_POS_XYZW", + "serialize_stats": false, + "video_decoder_cache_size": null, + "repo_id": "xdof/xmi_merged/split_11", + "streaming": false, + "tolerance_s": 0.1, + "use_imagenet_stats": false, + "split_name": "train", + "prefetch_factor": 2, + "max_episodes": null + }, + "weight": 0.062235781735 + }, + { + "dataset": { + "batch_size": 16, + "num_workers": 8, + "shuffle_buffer_size": 100, + "normalization_mapping": { + "ACTION": "ACTIONCHUNK_QUANTILE", + "ENV": "QUANTILE", + "STATE": "QUANTILE", + "VISUAL": "IDENTITY" + }, + "features": { + "action": { + "type": "ACTION", + "shape": "(20,)" + }, + "observation.image.0": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.1": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.2": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.state": { + "type": "STATE", + "shape": "(20,)" + } + }, + "stats": null, + "clip_values": { + "ENV": [ + -1.05, + 1.05 + ], + "STATE": [ + -1.05, + 1.05 + ] + }, + "chunk_size": 50, + "observation_mapping": { + "action.ee_quat_pos": "action", + "observations.ee_quat_pos": "observation.state", + "observations.images.cam_left": "observation.image.1", + "observations.images.cam_right": "observation.image.2", + "observations.images.cam_top": "observation.image.0" + }, + "observation_whitelist": [ + "observation.state", + "observation.image.0", + "observation.image.1", + "observation.image.2", + "action", + "task", + "subtask", + "dataset_index", + "training_mode" + ], + "observation_blacklist": null, + "transform_mapping": { + "action": [ + { + "type": "convert_to_6d_actions", + "action_key": "action", + "action_type": "EE_QUAT_POS_XYZW", + "input_type": "Dict", + "post_norm": false + }, + { + "type": "convert_to_6d_actions", + "action_key": "observation.state", + "action_type": "EE_QUAT_POS_XYZW", + "input_type": "Dict", + "post_norm": false + }, + { + "type": "delta_actions", + "state_key": "observation.state", + "action_key": "action", + "action_type": "EE_6D_POS", + "input_type": "Dict", + "post_norm": false, + "relative_to_state": true + } + ], + "observation.image.0": [ + { + "type": "random_resized_crop", + "height": 448, + "width": 448, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "observation.image.1": [ + { + "type": "random_resized_crop", + "height": 448, + "width": 448, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "observation.image.2": [ + { + "type": "random_resized_crop", + "height": 448, + "width": 448, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "task": [ + { + "type": "task_description_selector", + "task_key": "task", + "subtask_key": "subtask", + "output_key": "task", + "subtask_probability": 0.98, + "drop_subtask_key": true, + "has_subtask_key": "__task_description_has_subtask__", + "used_subtask_key": "__task_description_used_subtask__", + "input_type": "Dict", + "post_norm": false + } + ] + }, + "language_instruction": null, + "training_mode": "ROBOT_FLOWMATCH", + "action_type": "EE_QUAT_POS_XYZW", + "serialize_stats": false, + "video_decoder_cache_size": null, + "repo_id": "xdof/xmi_merged/split_12", + "streaming": false, + "tolerance_s": 0.1, + "use_imagenet_stats": false, + "split_name": "train", + "prefetch_factor": 2, + "max_episodes": null + }, + "weight": 0.06088029959 + }, + { + "dataset": { + "batch_size": 16, + "num_workers": 8, + "shuffle_buffer_size": 100, + "normalization_mapping": { + "ACTION": "ACTIONCHUNK_QUANTILE", + "ENV": "QUANTILE", + "STATE": "QUANTILE", + "VISUAL": "IDENTITY" + }, + "features": { + "action": { + "type": "ACTION", + "shape": "(20,)" + }, + "observation.image.0": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.1": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.2": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.state": { + "type": "STATE", + "shape": "(20,)" + } + }, + "stats": null, + "clip_values": { + "ENV": [ + -1.05, + 1.05 + ], + "STATE": [ + -1.05, + 1.05 + ] + }, + "chunk_size": 50, + "observation_mapping": { + "action.ee_quat_pos": "action", + "observations.ee_quat_pos": "observation.state", + "observations.images.cam_left": "observation.image.1", + "observations.images.cam_right": "observation.image.2", + "observations.images.cam_top": "observation.image.0" + }, + "observation_whitelist": [ + "observation.state", + "observation.image.0", + "observation.image.1", + "observation.image.2", + "action", + "task", + "subtask", + "dataset_index", + "training_mode" + ], + "observation_blacklist": null, + "transform_mapping": { + "action": [ + { + "type": "convert_to_6d_actions", + "action_key": "action", + "action_type": "EE_QUAT_POS_XYZW", + "input_type": "Dict", + "post_norm": false + }, + { + "type": "convert_to_6d_actions", + "action_key": "observation.state", + "action_type": "EE_QUAT_POS_XYZW", + "input_type": "Dict", + "post_norm": false + }, + { + "type": "delta_actions", + "state_key": "observation.state", + "action_key": "action", + "action_type": "EE_6D_POS", + "input_type": "Dict", + "post_norm": false, + "relative_to_state": true + } + ], + "observation.image.0": [ + { + "type": "random_resized_crop", + "height": 448, + "width": 448, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "observation.image.1": [ + { + "type": "random_resized_crop", + "height": 448, + "width": 448, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "observation.image.2": [ + { + "type": "random_resized_crop", + "height": 448, + "width": 448, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "task": [ + { + "type": "task_description_selector", + "task_key": "task", + "subtask_key": "subtask", + "output_key": "task", + "subtask_probability": 0.98, + "drop_subtask_key": true, + "has_subtask_key": "__task_description_has_subtask__", + "used_subtask_key": "__task_description_used_subtask__", + "input_type": "Dict", + "post_norm": false + } + ] + }, + "language_instruction": null, + "training_mode": "ROBOT_FLOWMATCH", + "action_type": "EE_QUAT_POS_XYZW", + "serialize_stats": false, + "video_decoder_cache_size": null, + "repo_id": "xdof/xmi_merged/split_13", + "streaming": false, + "tolerance_s": 0.1, + "use_imagenet_stats": false, + "split_name": "train", + "prefetch_factor": 2, + "max_episodes": null + }, + "weight": 0.062857319803 + }, + { + "dataset": { + "batch_size": 16, + "num_workers": 8, + "shuffle_buffer_size": 100, + "normalization_mapping": { + "ACTION": "ACTIONCHUNK_QUANTILE", + "ENV": "QUANTILE", + "STATE": "QUANTILE", + "VISUAL": "IDENTITY" + }, + "features": { + "action": { + "type": "ACTION", + "shape": "(20,)" + }, + "observation.image.0": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.1": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.2": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.state": { + "type": "STATE", + "shape": "(20,)" + } + }, + "stats": null, + "clip_values": { + "ENV": [ + -1.05, + 1.05 + ], + "STATE": [ + -1.05, + 1.05 + ] + }, + "chunk_size": 50, + "observation_mapping": { + "action.ee_quat_pos": "action", + "observations.ee_quat_pos": "observation.state", + "observations.images.cam_left": "observation.image.1", + "observations.images.cam_right": "observation.image.2", + "observations.images.cam_top": "observation.image.0" + }, + "observation_whitelist": [ + "observation.state", + "observation.image.0", + "observation.image.1", + "observation.image.2", + "action", + "task", + "subtask", + "dataset_index", + "training_mode" + ], + "observation_blacklist": null, + "transform_mapping": { + "action": [ + { + "type": "convert_to_6d_actions", + "action_key": "action", + "action_type": "EE_QUAT_POS_XYZW", + "input_type": "Dict", + "post_norm": false + }, + { + "type": "convert_to_6d_actions", + "action_key": "observation.state", + "action_type": "EE_QUAT_POS_XYZW", + "input_type": "Dict", + "post_norm": false + }, + { + "type": "delta_actions", + "state_key": "observation.state", + "action_key": "action", + "action_type": "EE_6D_POS", + "input_type": "Dict", + "post_norm": false, + "relative_to_state": true + } + ], + "observation.image.0": [ + { + "type": "random_resized_crop", + "height": 448, + "width": 448, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "observation.image.1": [ + { + "type": "random_resized_crop", + "height": 448, + "width": 448, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "observation.image.2": [ + { + "type": "random_resized_crop", + "height": 448, + "width": 448, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "task": [ + { + "type": "task_description_selector", + "task_key": "task", + "subtask_key": "subtask", + "output_key": "task", + "subtask_probability": 0.98, + "drop_subtask_key": true, + "has_subtask_key": "__task_description_has_subtask__", + "used_subtask_key": "__task_description_used_subtask__", + "input_type": "Dict", + "post_norm": false + } + ] + }, + "language_instruction": null, + "training_mode": "ROBOT_FLOWMATCH", + "action_type": "EE_QUAT_POS_XYZW", + "serialize_stats": false, + "video_decoder_cache_size": null, + "repo_id": "xdof/xmi_merged/split_14", + "streaming": false, + "tolerance_s": 0.1, + "use_imagenet_stats": false, + "split_name": "train", + "prefetch_factor": 2, + "max_episodes": null + }, + "weight": 0.063241930446 + }, + { + "dataset": { + "batch_size": 16, + "num_workers": 8, + "shuffle_buffer_size": 100, + "normalization_mapping": { + "ACTION": "ACTIONCHUNK_QUANTILE", + "ENV": "QUANTILE", + "STATE": "QUANTILE", + "VISUAL": "IDENTITY" + }, + "features": { + "action": { + "type": "ACTION", + "shape": "(20,)" + }, + "observation.image.0": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.1": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.2": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.state": { + "type": "STATE", + "shape": "(20,)" + } + }, + "stats": null, + "clip_values": { + "ENV": [ + -1.05, + 1.05 + ], + "STATE": [ + -1.05, + 1.05 + ] + }, + "chunk_size": 50, + "observation_mapping": { + "action.ee_quat_pos": "action", + "observations.ee_quat_pos": "observation.state", + "observations.images.cam_left": "observation.image.1", + "observations.images.cam_right": "observation.image.2", + "observations.images.cam_top": "observation.image.0" + }, + "observation_whitelist": [ + "observation.state", + "observation.image.0", + "observation.image.1", + "observation.image.2", + "action", + "task", + "subtask", + "dataset_index", + "training_mode" + ], + "observation_blacklist": null, + "transform_mapping": { + "action": [ + { + "type": "convert_to_6d_actions", + "action_key": "action", + "action_type": "EE_QUAT_POS_XYZW", + "input_type": "Dict", + "post_norm": false + }, + { + "type": "convert_to_6d_actions", + "action_key": "observation.state", + "action_type": "EE_QUAT_POS_XYZW", + "input_type": "Dict", + "post_norm": false + }, + { + "type": "delta_actions", + "state_key": "observation.state", + "action_key": "action", + "action_type": "EE_6D_POS", + "input_type": "Dict", + "post_norm": false, + "relative_to_state": true + } + ], + "observation.image.0": [ + { + "type": "random_resized_crop", + "height": 448, + "width": 448, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "observation.image.1": [ + { + "type": "random_resized_crop", + "height": 448, + "width": 448, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "observation.image.2": [ + { + "type": "random_resized_crop", + "height": 448, + "width": 448, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "task": [ + { + "type": "task_description_selector", + "task_key": "task", + "subtask_key": "subtask", + "output_key": "task", + "subtask_probability": 0.98, + "drop_subtask_key": true, + "has_subtask_key": "__task_description_has_subtask__", + "used_subtask_key": "__task_description_used_subtask__", + "input_type": "Dict", + "post_norm": false + } + ] + }, + "language_instruction": null, + "training_mode": "ROBOT_FLOWMATCH", + "action_type": "EE_QUAT_POS_XYZW", + "serialize_stats": false, + "video_decoder_cache_size": null, + "repo_id": "xdof/xmi_merged/split_15", + "streaming": false, + "tolerance_s": 0.1, + "use_imagenet_stats": false, + "split_name": "train", + "prefetch_factor": 2, + "max_episodes": null + }, + "weight": 0.061145520048 + } + ], + "split_name": null, + "shuffle_type": "FRAME", + "seed": 12345, + "num_workers": 4, + "prefetch_factor": 2, + "batch_size": 64, + "streaming": false, + "shuffle_buffer_size": null, + "vl_batch_size": null, + "observation_whitelist": null, + "transform_mapping": null, + "chunk_size": null, + "flatten_nested": true, + "serialize_stats": true, + "video_decoder_cache_size": null, + "training_mode_override": null + }, + "weight": 0.399030792742 + }, + { + "dataset": { + "datasets": [ + { + "dataset": { + "batch_size": 16, + "num_workers": 8, + "shuffle_buffer_size": 100, + "normalization_mapping": { + "ACTION": "ACTIONCHUNK_QUANTILE", + "ENV": "QUANTILE", + "STATE": "QUANTILE", + "VISUAL": "IDENTITY" + }, + "features": { + "action": { + "type": "ACTION", + "shape": "(20,)" + }, + "observation.image.0": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.1": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.2": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.state": { + "type": "STATE", + "shape": "(20,)" + } + }, + "stats": null, + "clip_values": { + "ENV": [ + -1.05, + 1.05 + ], + "STATE": [ + -1.05, + 1.05 + ] + }, + "chunk_size": 50, + "observation_mapping": { + "action.ee_quat_pos": "action", + "observations.ee_quat_pos": "observation.state", + "observations.images.cam_left": "observation.image.1", + "observations.images.cam_right": "observation.image.2", + "observations.images.cam_top": "observation.image.0" + }, + "observation_whitelist": [ + "observation.state", + "observation.image.0", + "observation.image.1", + "observation.image.2", + "action", + "task", + "subtask", + "dataset_index", + "training_mode" + ], + "observation_blacklist": null, + "transform_mapping": { + "action": [ + { + "type": "convert_to_6d_actions", + "action_key": "action", + "action_type": "EE_QUAT_POS_XYZW", + "input_type": "Dict", + "post_norm": false + }, + { + "type": "convert_to_6d_actions", + "action_key": "observation.state", + "action_type": "EE_QUAT_POS_XYZW", + "input_type": "Dict", + "post_norm": false + }, + { + "type": "delta_actions", + "state_key": "observation.state", + "action_key": "action", + "action_type": "EE_6D_POS", + "input_type": "Dict", + "post_norm": false, + "relative_to_state": true + } + ], + "observation.image.0": [ + { + "type": "random_resized_crop", + "height": 448, + "width": 448, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "observation.image.1": [ + { + "type": "random_resized_crop", + "height": 448, + "width": 448, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "observation.image.2": [ + { + "type": "random_resized_crop", + "height": 448, + "width": 448, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "task": [ + { + "type": "task_description_selector", + "task_key": "task", + "subtask_key": "subtask", + "output_key": "task", + "subtask_probability": 0.98, + "drop_subtask_key": true, + "has_subtask_key": "__task_description_has_subtask__", + "used_subtask_key": "__task_description_used_subtask__", + "input_type": "Dict", + "post_norm": false + } + ] + }, + "language_instruction": null, + "training_mode": "ROBOT_FLOWMATCH", + "action_type": "EE_QUAT_POS_XYZW", + "serialize_stats": false, + "video_decoder_cache_size": null, + "repo_id": "xdof/yam_box_merged/split_00", + "streaming": false, + "tolerance_s": 0.1, + "use_imagenet_stats": false, + "split_name": "train", + "prefetch_factor": 2, + "max_episodes": null + }, + "weight": 0.084604590545 + }, + { + "dataset": { + "batch_size": 16, + "num_workers": 8, + "shuffle_buffer_size": 100, + "normalization_mapping": { + "ACTION": "ACTIONCHUNK_QUANTILE", + "ENV": "QUANTILE", + "STATE": "QUANTILE", + "VISUAL": "IDENTITY" + }, + "features": { + "action": { + "type": "ACTION", + "shape": "(20,)" + }, + "observation.image.0": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.1": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.2": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.state": { + "type": "STATE", + "shape": "(20,)" + } + }, + "stats": null, + "clip_values": { + "ENV": [ + -1.05, + 1.05 + ], + "STATE": [ + -1.05, + 1.05 + ] + }, + "chunk_size": 50, + "observation_mapping": { + "action.ee_quat_pos": "action", + "observations.ee_quat_pos": "observation.state", + "observations.images.cam_left": "observation.image.1", + "observations.images.cam_right": "observation.image.2", + "observations.images.cam_top": "observation.image.0" + }, + "observation_whitelist": [ + "observation.state", + "observation.image.0", + "observation.image.1", + "observation.image.2", + "action", + "task", + "subtask", + "dataset_index", + "training_mode" + ], + "observation_blacklist": null, + "transform_mapping": { + "action": [ + { + "type": "convert_to_6d_actions", + "action_key": "action", + "action_type": "EE_QUAT_POS_XYZW", + "input_type": "Dict", + "post_norm": false + }, + { + "type": "convert_to_6d_actions", + "action_key": "observation.state", + "action_type": "EE_QUAT_POS_XYZW", + "input_type": "Dict", + "post_norm": false + }, + { + "type": "delta_actions", + "state_key": "observation.state", + "action_key": "action", + "action_type": "EE_6D_POS", + "input_type": "Dict", + "post_norm": false, + "relative_to_state": true + } + ], + "observation.image.0": [ + { + "type": "random_resized_crop", + "height": 448, + "width": 448, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "observation.image.1": [ + { + "type": "random_resized_crop", + "height": 448, + "width": 448, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "observation.image.2": [ + { + "type": "random_resized_crop", + "height": 448, + "width": 448, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "task": [ + { + "type": "task_description_selector", + "task_key": "task", + "subtask_key": "subtask", + "output_key": "task", + "subtask_probability": 0.98, + "drop_subtask_key": true, + "has_subtask_key": "__task_description_has_subtask__", + "used_subtask_key": "__task_description_used_subtask__", + "input_type": "Dict", + "post_norm": false + } + ] + }, + "language_instruction": null, + "training_mode": "ROBOT_FLOWMATCH", + "action_type": "EE_QUAT_POS_XYZW", + "serialize_stats": false, + "video_decoder_cache_size": null, + "repo_id": "xdof/yam_box_merged/split_01", + "streaming": false, + "tolerance_s": 0.1, + "use_imagenet_stats": false, + "split_name": "train", + "prefetch_factor": 2, + "max_episodes": null + }, + "weight": 0.082449558605 + }, + { + "dataset": { + "batch_size": 16, + "num_workers": 8, + "shuffle_buffer_size": 100, + "normalization_mapping": { + "ACTION": "ACTIONCHUNK_QUANTILE", + "ENV": "QUANTILE", + "STATE": "QUANTILE", + "VISUAL": "IDENTITY" + }, + "features": { + "action": { + "type": "ACTION", + "shape": "(20,)" + }, + "observation.image.0": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.1": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.2": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.state": { + "type": "STATE", + "shape": "(20,)" + } + }, + "stats": null, + "clip_values": { + "ENV": [ + -1.05, + 1.05 + ], + "STATE": [ + -1.05, + 1.05 + ] + }, + "chunk_size": 50, + "observation_mapping": { + "action.ee_quat_pos": "action", + "observations.ee_quat_pos": "observation.state", + "observations.images.cam_left": "observation.image.1", + "observations.images.cam_right": "observation.image.2", + "observations.images.cam_top": "observation.image.0" + }, + "observation_whitelist": [ + "observation.state", + "observation.image.0", + "observation.image.1", + "observation.image.2", + "action", + "task", + "subtask", + "dataset_index", + "training_mode" + ], + "observation_blacklist": null, + "transform_mapping": { + "action": [ + { + "type": "convert_to_6d_actions", + "action_key": "action", + "action_type": "EE_QUAT_POS_XYZW", + "input_type": "Dict", + "post_norm": false + }, + { + "type": "convert_to_6d_actions", + "action_key": "observation.state", + "action_type": "EE_QUAT_POS_XYZW", + "input_type": "Dict", + "post_norm": false + }, + { + "type": "delta_actions", + "state_key": "observation.state", + "action_key": "action", + "action_type": "EE_6D_POS", + "input_type": "Dict", + "post_norm": false, + "relative_to_state": true + } + ], + "observation.image.0": [ + { + "type": "random_resized_crop", + "height": 448, + "width": 448, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "observation.image.1": [ + { + "type": "random_resized_crop", + "height": 448, + "width": 448, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "observation.image.2": [ + { + "type": "random_resized_crop", + "height": 448, + "width": 448, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "task": [ + { + "type": "task_description_selector", + "task_key": "task", + "subtask_key": "subtask", + "output_key": "task", + "subtask_probability": 0.98, + "drop_subtask_key": true, + "has_subtask_key": "__task_description_has_subtask__", + "used_subtask_key": "__task_description_used_subtask__", + "input_type": "Dict", + "post_norm": false + } + ] + }, + "language_instruction": null, + "training_mode": "ROBOT_FLOWMATCH", + "action_type": "EE_QUAT_POS_XYZW", + "serialize_stats": false, + "video_decoder_cache_size": null, + "repo_id": "xdof/yam_box_merged/split_02", + "streaming": false, + "tolerance_s": 0.1, + "use_imagenet_stats": false, + "split_name": "train", + "prefetch_factor": 2, + "max_episodes": null + }, + "weight": 0.083861401471 + }, + { + "dataset": { + "batch_size": 16, + "num_workers": 8, + "shuffle_buffer_size": 100, + "normalization_mapping": { + "ACTION": "ACTIONCHUNK_QUANTILE", + "ENV": "QUANTILE", + "STATE": "QUANTILE", + "VISUAL": "IDENTITY" + }, + "features": { + "action": { + "type": "ACTION", + "shape": "(20,)" + }, + "observation.image.0": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.1": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.2": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.state": { + "type": "STATE", + "shape": "(20,)" + } + }, + "stats": null, + "clip_values": { + "ENV": [ + -1.05, + 1.05 + ], + "STATE": [ + -1.05, + 1.05 + ] + }, + "chunk_size": 50, + "observation_mapping": { + "action.ee_quat_pos": "action", + "observations.ee_quat_pos": "observation.state", + "observations.images.cam_left": "observation.image.1", + "observations.images.cam_right": "observation.image.2", + "observations.images.cam_top": "observation.image.0" + }, + "observation_whitelist": [ + "observation.state", + "observation.image.0", + "observation.image.1", + "observation.image.2", + "action", + "task", + "subtask", + "dataset_index", + "training_mode" + ], + "observation_blacklist": null, + "transform_mapping": { + "action": [ + { + "type": "convert_to_6d_actions", + "action_key": "action", + "action_type": "EE_QUAT_POS_XYZW", + "input_type": "Dict", + "post_norm": false + }, + { + "type": "convert_to_6d_actions", + "action_key": "observation.state", + "action_type": "EE_QUAT_POS_XYZW", + "input_type": "Dict", + "post_norm": false + }, + { + "type": "delta_actions", + "state_key": "observation.state", + "action_key": "action", + "action_type": "EE_6D_POS", + "input_type": "Dict", + "post_norm": false, + "relative_to_state": true + } + ], + "observation.image.0": [ + { + "type": "random_resized_crop", + "height": 448, + "width": 448, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "observation.image.1": [ + { + "type": "random_resized_crop", + "height": 448, + "width": 448, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "observation.image.2": [ + { + "type": "random_resized_crop", + "height": 448, + "width": 448, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "task": [ + { + "type": "task_description_selector", + "task_key": "task", + "subtask_key": "subtask", + "output_key": "task", + "subtask_probability": 0.98, + "drop_subtask_key": true, + "has_subtask_key": "__task_description_has_subtask__", + "used_subtask_key": "__task_description_used_subtask__", + "input_type": "Dict", + "post_norm": false + } + ] + }, + "language_instruction": null, + "training_mode": "ROBOT_FLOWMATCH", + "action_type": "EE_QUAT_POS_XYZW", + "serialize_stats": false, + "video_decoder_cache_size": null, + "repo_id": "xdof/yam_box_merged/split_03", + "streaming": false, + "tolerance_s": 0.1, + "use_imagenet_stats": false, + "split_name": "train", + "prefetch_factor": 2, + "max_episodes": null + }, + "weight": 0.082710462996 + }, + { + "dataset": { + "batch_size": 16, + "num_workers": 8, + "shuffle_buffer_size": 100, + "normalization_mapping": { + "ACTION": "ACTIONCHUNK_QUANTILE", + "ENV": "QUANTILE", + "STATE": "QUANTILE", + "VISUAL": "IDENTITY" + }, + "features": { + "action": { + "type": "ACTION", + "shape": "(20,)" + }, + "observation.image.0": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.1": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.2": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.state": { + "type": "STATE", + "shape": "(20,)" + } + }, + "stats": null, + "clip_values": { + "ENV": [ + -1.05, + 1.05 + ], + "STATE": [ + -1.05, + 1.05 + ] + }, + "chunk_size": 50, + "observation_mapping": { + "action.ee_quat_pos": "action", + "observations.ee_quat_pos": "observation.state", + "observations.images.cam_left": "observation.image.1", + "observations.images.cam_right": "observation.image.2", + "observations.images.cam_top": "observation.image.0" + }, + "observation_whitelist": [ + "observation.state", + "observation.image.0", + "observation.image.1", + "observation.image.2", + "action", + "task", + "subtask", + "dataset_index", + "training_mode" + ], + "observation_blacklist": null, + "transform_mapping": { + "action": [ + { + "type": "convert_to_6d_actions", + "action_key": "action", + "action_type": "EE_QUAT_POS_XYZW", + "input_type": "Dict", + "post_norm": false + }, + { + "type": "convert_to_6d_actions", + "action_key": "observation.state", + "action_type": "EE_QUAT_POS_XYZW", + "input_type": "Dict", + "post_norm": false + }, + { + "type": "delta_actions", + "state_key": "observation.state", + "action_key": "action", + "action_type": "EE_6D_POS", + "input_type": "Dict", + "post_norm": false, + "relative_to_state": true + } + ], + "observation.image.0": [ + { + "type": "random_resized_crop", + "height": 448, + "width": 448, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "observation.image.1": [ + { + "type": "random_resized_crop", + "height": 448, + "width": 448, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "observation.image.2": [ + { + "type": "random_resized_crop", + "height": 448, + "width": 448, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "task": [ + { + "type": "task_description_selector", + "task_key": "task", + "subtask_key": "subtask", + "output_key": "task", + "subtask_probability": 0.98, + "drop_subtask_key": true, + "has_subtask_key": "__task_description_has_subtask__", + "used_subtask_key": "__task_description_used_subtask__", + "input_type": "Dict", + "post_norm": false + } + ] + }, + "language_instruction": null, + "training_mode": "ROBOT_FLOWMATCH", + "action_type": "EE_QUAT_POS_XYZW", + "serialize_stats": false, + "video_decoder_cache_size": null, + "repo_id": "xdof/yam_box_merged/split_04", + "streaming": false, + "tolerance_s": 0.1, + "use_imagenet_stats": false, + "split_name": "train", + "prefetch_factor": 2, + "max_episodes": null + }, + "weight": 0.084379665861 + }, + { + "dataset": { + "batch_size": 16, + "num_workers": 8, + "shuffle_buffer_size": 100, + "normalization_mapping": { + "ACTION": "ACTIONCHUNK_QUANTILE", + "ENV": "QUANTILE", + "STATE": "QUANTILE", + "VISUAL": "IDENTITY" + }, + "features": { + "action": { + "type": "ACTION", + "shape": "(20,)" + }, + "observation.image.0": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.1": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.2": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.state": { + "type": "STATE", + "shape": "(20,)" + } + }, + "stats": null, + "clip_values": { + "ENV": [ + -1.05, + 1.05 + ], + "STATE": [ + -1.05, + 1.05 + ] + }, + "chunk_size": 50, + "observation_mapping": { + "action.ee_quat_pos": "action", + "observations.ee_quat_pos": "observation.state", + "observations.images.cam_left": "observation.image.1", + "observations.images.cam_right": "observation.image.2", + "observations.images.cam_top": "observation.image.0" + }, + "observation_whitelist": [ + "observation.state", + "observation.image.0", + "observation.image.1", + "observation.image.2", + "action", + "task", + "subtask", + "dataset_index", + "training_mode" + ], + "observation_blacklist": null, + "transform_mapping": { + "action": [ + { + "type": "convert_to_6d_actions", + "action_key": "action", + "action_type": "EE_QUAT_POS_XYZW", + "input_type": "Dict", + "post_norm": false + }, + { + "type": "convert_to_6d_actions", + "action_key": "observation.state", + "action_type": "EE_QUAT_POS_XYZW", + "input_type": "Dict", + "post_norm": false + }, + { + "type": "delta_actions", + "state_key": "observation.state", + "action_key": "action", + "action_type": "EE_6D_POS", + "input_type": "Dict", + "post_norm": false, + "relative_to_state": true + } + ], + "observation.image.0": [ + { + "type": "random_resized_crop", + "height": 448, + "width": 448, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "observation.image.1": [ + { + "type": "random_resized_crop", + "height": 448, + "width": 448, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "observation.image.2": [ + { + "type": "random_resized_crop", + "height": 448, + "width": 448, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "task": [ + { + "type": "task_description_selector", + "task_key": "task", + "subtask_key": "subtask", + "output_key": "task", + "subtask_probability": 0.98, + "drop_subtask_key": true, + "has_subtask_key": "__task_description_has_subtask__", + "used_subtask_key": "__task_description_used_subtask__", + "input_type": "Dict", + "post_norm": false + } + ] + }, + "language_instruction": null, + "training_mode": "ROBOT_FLOWMATCH", + "action_type": "EE_QUAT_POS_XYZW", + "serialize_stats": false, + "video_decoder_cache_size": null, + "repo_id": "xdof/yam_box_merged/split_05", + "streaming": false, + "tolerance_s": 0.1, + "use_imagenet_stats": false, + "split_name": "train", + "prefetch_factor": 2, + "max_episodes": null + }, + "weight": 0.085936725791 + }, + { + "dataset": { + "batch_size": 16, + "num_workers": 8, + "shuffle_buffer_size": 100, + "normalization_mapping": { + "ACTION": "ACTIONCHUNK_QUANTILE", + "ENV": "QUANTILE", + "STATE": "QUANTILE", + "VISUAL": "IDENTITY" + }, + "features": { + "action": { + "type": "ACTION", + "shape": "(20,)" + }, + "observation.image.0": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.1": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.2": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.state": { + "type": "STATE", + "shape": "(20,)" + } + }, + "stats": null, + "clip_values": { + "ENV": [ + -1.05, + 1.05 + ], + "STATE": [ + -1.05, + 1.05 + ] + }, + "chunk_size": 50, + "observation_mapping": { + "action.ee_quat_pos": "action", + "observations.ee_quat_pos": "observation.state", + "observations.images.cam_left": "observation.image.1", + "observations.images.cam_right": "observation.image.2", + "observations.images.cam_top": "observation.image.0" + }, + "observation_whitelist": [ + "observation.state", + "observation.image.0", + "observation.image.1", + "observation.image.2", + "action", + "task", + "subtask", + "dataset_index", + "training_mode" + ], + "observation_blacklist": null, + "transform_mapping": { + "action": [ + { + "type": "convert_to_6d_actions", + "action_key": "action", + "action_type": "EE_QUAT_POS_XYZW", + "input_type": "Dict", + "post_norm": false + }, + { + "type": "convert_to_6d_actions", + "action_key": "observation.state", + "action_type": "EE_QUAT_POS_XYZW", + "input_type": "Dict", + "post_norm": false + }, + { + "type": "delta_actions", + "state_key": "observation.state", + "action_key": "action", + "action_type": "EE_6D_POS", + "input_type": "Dict", + "post_norm": false, + "relative_to_state": true + } + ], + "observation.image.0": [ + { + "type": "random_resized_crop", + "height": 448, + "width": 448, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "observation.image.1": [ + { + "type": "random_resized_crop", + "height": 448, + "width": 448, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "observation.image.2": [ + { + "type": "random_resized_crop", + "height": 448, + "width": 448, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "task": [ + { + "type": "task_description_selector", + "task_key": "task", + "subtask_key": "subtask", + "output_key": "task", + "subtask_probability": 0.98, + "drop_subtask_key": true, + "has_subtask_key": "__task_description_has_subtask__", + "used_subtask_key": "__task_description_used_subtask__", + "input_type": "Dict", + "post_norm": false + } + ] + }, + "language_instruction": null, + "training_mode": "ROBOT_FLOWMATCH", + "action_type": "EE_QUAT_POS_XYZW", + "serialize_stats": false, + "video_decoder_cache_size": null, + "repo_id": "xdof/yam_box_merged/split_06", + "streaming": false, + "tolerance_s": 0.1, + "use_imagenet_stats": false, + "split_name": "train", + "prefetch_factor": 2, + "max_episodes": null + }, + "weight": 0.080828184433 + }, + { + "dataset": { + "batch_size": 16, + "num_workers": 8, + "shuffle_buffer_size": 100, + "normalization_mapping": { + "ACTION": "ACTIONCHUNK_QUANTILE", + "ENV": "QUANTILE", + "STATE": "QUANTILE", + "VISUAL": "IDENTITY" + }, + "features": { + "action": { + "type": "ACTION", + "shape": "(20,)" + }, + "observation.image.0": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.1": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.2": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.state": { + "type": "STATE", + "shape": "(20,)" + } + }, + "stats": null, + "clip_values": { + "ENV": [ + -1.05, + 1.05 + ], + "STATE": [ + -1.05, + 1.05 + ] + }, + "chunk_size": 50, + "observation_mapping": { + "action.ee_quat_pos": "action", + "observations.ee_quat_pos": "observation.state", + "observations.images.cam_left": "observation.image.1", + "observations.images.cam_right": "observation.image.2", + "observations.images.cam_top": "observation.image.0" + }, + "observation_whitelist": [ + "observation.state", + "observation.image.0", + "observation.image.1", + "observation.image.2", + "action", + "task", + "subtask", + "dataset_index", + "training_mode" + ], + "observation_blacklist": null, + "transform_mapping": { + "action": [ + { + "type": "convert_to_6d_actions", + "action_key": "action", + "action_type": "EE_QUAT_POS_XYZW", + "input_type": "Dict", + "post_norm": false + }, + { + "type": "convert_to_6d_actions", + "action_key": "observation.state", + "action_type": "EE_QUAT_POS_XYZW", + "input_type": "Dict", + "post_norm": false + }, + { + "type": "delta_actions", + "state_key": "observation.state", + "action_key": "action", + "action_type": "EE_6D_POS", + "input_type": "Dict", + "post_norm": false, + "relative_to_state": true + } + ], + "observation.image.0": [ + { + "type": "random_resized_crop", + "height": 448, + "width": 448, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "observation.image.1": [ + { + "type": "random_resized_crop", + "height": 448, + "width": 448, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "observation.image.2": [ + { + "type": "random_resized_crop", + "height": 448, + "width": 448, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "task": [ + { + "type": "task_description_selector", + "task_key": "task", + "subtask_key": "subtask", + "output_key": "task", + "subtask_probability": 0.98, + "drop_subtask_key": true, + "has_subtask_key": "__task_description_has_subtask__", + "used_subtask_key": "__task_description_used_subtask__", + "input_type": "Dict", + "post_norm": false + } + ] + }, + "language_instruction": null, + "training_mode": "ROBOT_FLOWMATCH", + "action_type": "EE_QUAT_POS_XYZW", + "serialize_stats": false, + "video_decoder_cache_size": null, + "repo_id": "xdof/yam_box_merged/split_07", + "streaming": false, + "tolerance_s": 0.1, + "use_imagenet_stats": false, + "split_name": "train", + "prefetch_factor": 2, + "max_episodes": null + }, + "weight": 0.082485991664 + }, + { + "dataset": { + "batch_size": 16, + "num_workers": 8, + "shuffle_buffer_size": 100, + "normalization_mapping": { + "ACTION": "ACTIONCHUNK_QUANTILE", + "ENV": "QUANTILE", + "STATE": "QUANTILE", + "VISUAL": "IDENTITY" + }, + "features": { + "action": { + "type": "ACTION", + "shape": "(20,)" + }, + "observation.image.0": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.1": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.2": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.state": { + "type": "STATE", + "shape": "(20,)" + } + }, + "stats": null, + "clip_values": { + "ENV": [ + -1.05, + 1.05 + ], + "STATE": [ + -1.05, + 1.05 + ] + }, + "chunk_size": 50, + "observation_mapping": { + "action.ee_quat_pos": "action", + "observations.ee_quat_pos": "observation.state", + "observations.images.cam_left": "observation.image.1", + "observations.images.cam_right": "observation.image.2", + "observations.images.cam_top": "observation.image.0" + }, + "observation_whitelist": [ + "observation.state", + "observation.image.0", + "observation.image.1", + "observation.image.2", + "action", + "task", + "subtask", + "dataset_index", + "training_mode" + ], + "observation_blacklist": null, + "transform_mapping": { + "action": [ + { + "type": "convert_to_6d_actions", + "action_key": "action", + "action_type": "EE_QUAT_POS_XYZW", + "input_type": "Dict", + "post_norm": false + }, + { + "type": "convert_to_6d_actions", + "action_key": "observation.state", + "action_type": "EE_QUAT_POS_XYZW", + "input_type": "Dict", + "post_norm": false + }, + { + "type": "delta_actions", + "state_key": "observation.state", + "action_key": "action", + "action_type": "EE_6D_POS", + "input_type": "Dict", + "post_norm": false, + "relative_to_state": true + } + ], + "observation.image.0": [ + { + "type": "random_resized_crop", + "height": 448, + "width": 448, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "observation.image.1": [ + { + "type": "random_resized_crop", + "height": 448, + "width": 448, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "observation.image.2": [ + { + "type": "random_resized_crop", + "height": 448, + "width": 448, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "task": [ + { + "type": "task_description_selector", + "task_key": "task", + "subtask_key": "subtask", + "output_key": "task", + "subtask_probability": 0.98, + "drop_subtask_key": true, + "has_subtask_key": "__task_description_has_subtask__", + "used_subtask_key": "__task_description_used_subtask__", + "input_type": "Dict", + "post_norm": false + } + ] + }, + "language_instruction": null, + "training_mode": "ROBOT_FLOWMATCH", + "action_type": "EE_QUAT_POS_XYZW", + "serialize_stats": false, + "video_decoder_cache_size": null, + "repo_id": "xdof/yam_box_merged/split_08", + "streaming": false, + "tolerance_s": 0.1, + "use_imagenet_stats": false, + "split_name": "train", + "prefetch_factor": 2, + "max_episodes": null + }, + "weight": 0.082934584009 + }, + { + "dataset": { + "batch_size": 16, + "num_workers": 8, + "shuffle_buffer_size": 100, + "normalization_mapping": { + "ACTION": "ACTIONCHUNK_QUANTILE", + "ENV": "QUANTILE", + "STATE": "QUANTILE", + "VISUAL": "IDENTITY" + }, + "features": { + "action": { + "type": "ACTION", + "shape": "(20,)" + }, + "observation.image.0": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.1": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.2": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.state": { + "type": "STATE", + "shape": "(20,)" + } + }, + "stats": null, + "clip_values": { + "ENV": [ + -1.05, + 1.05 + ], + "STATE": [ + -1.05, + 1.05 + ] + }, + "chunk_size": 50, + "observation_mapping": { + "action.ee_quat_pos": "action", + "observations.ee_quat_pos": "observation.state", + "observations.images.cam_left": "observation.image.1", + "observations.images.cam_right": "observation.image.2", + "observations.images.cam_top": "observation.image.0" + }, + "observation_whitelist": [ + "observation.state", + "observation.image.0", + "observation.image.1", + "observation.image.2", + "action", + "task", + "subtask", + "dataset_index", + "training_mode" + ], + "observation_blacklist": null, + "transform_mapping": { + "action": [ + { + "type": "convert_to_6d_actions", + "action_key": "action", + "action_type": "EE_QUAT_POS_XYZW", + "input_type": "Dict", + "post_norm": false + }, + { + "type": "convert_to_6d_actions", + "action_key": "observation.state", + "action_type": "EE_QUAT_POS_XYZW", + "input_type": "Dict", + "post_norm": false + }, + { + "type": "delta_actions", + "state_key": "observation.state", + "action_key": "action", + "action_type": "EE_6D_POS", + "input_type": "Dict", + "post_norm": false, + "relative_to_state": true + } + ], + "observation.image.0": [ + { + "type": "random_resized_crop", + "height": 448, + "width": 448, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "observation.image.1": [ + { + "type": "random_resized_crop", + "height": 448, + "width": 448, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "observation.image.2": [ + { + "type": "random_resized_crop", + "height": 448, + "width": 448, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "task": [ + { + "type": "task_description_selector", + "task_key": "task", + "subtask_key": "subtask", + "output_key": "task", + "subtask_probability": 0.98, + "drop_subtask_key": true, + "has_subtask_key": "__task_description_has_subtask__", + "used_subtask_key": "__task_description_used_subtask__", + "input_type": "Dict", + "post_norm": false + } + ] + }, + "language_instruction": null, + "training_mode": "ROBOT_FLOWMATCH", + "action_type": "EE_QUAT_POS_XYZW", + "serialize_stats": false, + "video_decoder_cache_size": null, + "repo_id": "xdof/yam_box_merged/split_09", + "streaming": false, + "tolerance_s": 0.1, + "use_imagenet_stats": false, + "split_name": "train", + "prefetch_factor": 2, + "max_episodes": null + }, + "weight": 0.083525796945 + }, + { + "dataset": { + "batch_size": 16, + "num_workers": 8, + "shuffle_buffer_size": 100, + "normalization_mapping": { + "ACTION": "ACTIONCHUNK_QUANTILE", + "ENV": "QUANTILE", + "STATE": "QUANTILE", + "VISUAL": "IDENTITY" + }, + "features": { + "action": { + "type": "ACTION", + "shape": "(20,)" + }, + "observation.image.0": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.1": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.2": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.state": { + "type": "STATE", + "shape": "(20,)" + } + }, + "stats": null, + "clip_values": { + "ENV": [ + -1.05, + 1.05 + ], + "STATE": [ + -1.05, + 1.05 + ] + }, + "chunk_size": 50, + "observation_mapping": { + "action.ee_quat_pos": "action", + "observations.ee_quat_pos": "observation.state", + "observations.images.cam_left": "observation.image.1", + "observations.images.cam_right": "observation.image.2", + "observations.images.cam_top": "observation.image.0" + }, + "observation_whitelist": [ + "observation.state", + "observation.image.0", + "observation.image.1", + "observation.image.2", + "action", + "task", + "subtask", + "dataset_index", + "training_mode" + ], + "observation_blacklist": null, + "transform_mapping": { + "action": [ + { + "type": "convert_to_6d_actions", + "action_key": "action", + "action_type": "EE_QUAT_POS_XYZW", + "input_type": "Dict", + "post_norm": false + }, + { + "type": "convert_to_6d_actions", + "action_key": "observation.state", + "action_type": "EE_QUAT_POS_XYZW", + "input_type": "Dict", + "post_norm": false + }, + { + "type": "delta_actions", + "state_key": "observation.state", + "action_key": "action", + "action_type": "EE_6D_POS", + "input_type": "Dict", + "post_norm": false, + "relative_to_state": true + } + ], + "observation.image.0": [ + { + "type": "random_resized_crop", + "height": 448, + "width": 448, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "observation.image.1": [ + { + "type": "random_resized_crop", + "height": 448, + "width": 448, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "observation.image.2": [ + { + "type": "random_resized_crop", + "height": 448, + "width": 448, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "task": [ + { + "type": "task_description_selector", + "task_key": "task", + "subtask_key": "subtask", + "output_key": "task", + "subtask_probability": 0.98, + "drop_subtask_key": true, + "has_subtask_key": "__task_description_has_subtask__", + "used_subtask_key": "__task_description_used_subtask__", + "input_type": "Dict", + "post_norm": false + } + ] + }, + "language_instruction": null, + "training_mode": "ROBOT_FLOWMATCH", + "action_type": "EE_QUAT_POS_XYZW", + "serialize_stats": false, + "video_decoder_cache_size": null, + "repo_id": "xdof/yam_box_merged/split_10", + "streaming": false, + "tolerance_s": 0.1, + "use_imagenet_stats": false, + "split_name": "train", + "prefetch_factor": 2, + "max_episodes": null + }, + "weight": 0.082555230962 + }, + { + "dataset": { + "batch_size": 16, + "num_workers": 8, + "shuffle_buffer_size": 100, + "normalization_mapping": { + "ACTION": "ACTIONCHUNK_QUANTILE", + "ENV": "QUANTILE", + "STATE": "QUANTILE", + "VISUAL": "IDENTITY" + }, + "features": { + "action": { + "type": "ACTION", + "shape": "(20,)" + }, + "observation.image.0": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.1": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.2": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.state": { + "type": "STATE", + "shape": "(20,)" + } + }, + "stats": null, + "clip_values": { + "ENV": [ + -1.05, + 1.05 + ], + "STATE": [ + -1.05, + 1.05 + ] + }, + "chunk_size": 50, + "observation_mapping": { + "action.ee_quat_pos": "action", + "observations.ee_quat_pos": "observation.state", + "observations.images.cam_left": "observation.image.1", + "observations.images.cam_right": "observation.image.2", + "observations.images.cam_top": "observation.image.0" + }, + "observation_whitelist": [ + "observation.state", + "observation.image.0", + "observation.image.1", + "observation.image.2", + "action", + "task", + "subtask", + "dataset_index", + "training_mode" + ], + "observation_blacklist": null, + "transform_mapping": { + "action": [ + { + "type": "convert_to_6d_actions", + "action_key": "action", + "action_type": "EE_QUAT_POS_XYZW", + "input_type": "Dict", + "post_norm": false + }, + { + "type": "convert_to_6d_actions", + "action_key": "observation.state", + "action_type": "EE_QUAT_POS_XYZW", + "input_type": "Dict", + "post_norm": false + }, + { + "type": "delta_actions", + "state_key": "observation.state", + "action_key": "action", + "action_type": "EE_6D_POS", + "input_type": "Dict", + "post_norm": false, + "relative_to_state": true + } + ], + "observation.image.0": [ + { + "type": "random_resized_crop", + "height": 448, + "width": 448, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "observation.image.1": [ + { + "type": "random_resized_crop", + "height": 448, + "width": 448, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "observation.image.2": [ + { + "type": "random_resized_crop", + "height": 448, + "width": 448, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "task": [ + { + "type": "task_description_selector", + "task_key": "task", + "subtask_key": "subtask", + "output_key": "task", + "subtask_probability": 0.98, + "drop_subtask_key": true, + "has_subtask_key": "__task_description_has_subtask__", + "used_subtask_key": "__task_description_used_subtask__", + "input_type": "Dict", + "post_norm": false + } + ] + }, + "language_instruction": null, + "training_mode": "ROBOT_FLOWMATCH", + "action_type": "EE_QUAT_POS_XYZW", + "serialize_stats": false, + "video_decoder_cache_size": null, + "repo_id": "xdof/yam_box_merged/split_11", + "streaming": false, + "tolerance_s": 0.1, + "use_imagenet_stats": false, + "split_name": "train", + "prefetch_factor": 2, + "max_episodes": null + }, + "weight": 0.083727806718 + } + ], + "split_name": null, + "shuffle_type": "FRAME", + "seed": 12345, + "num_workers": 4, + "prefetch_factor": 2, + "batch_size": 64, + "streaming": false, + "shuffle_buffer_size": null, + "vl_batch_size": null, + "observation_whitelist": null, + "transform_mapping": null, + "chunk_size": null, + "flatten_nested": true, + "serialize_stats": true, + "video_decoder_cache_size": null, + "training_mode_override": null + }, + "weight": 0.172501168738 + }, + { + "dataset": { + "batch_size": 16, + "num_workers": 8, + "shuffle_buffer_size": 100, + "normalization_mapping": { + "ACTION": "ACTIONCHUNK_QUANTILE", + "ENV": "QUANTILE", + "STATE": "QUANTILE", + "VISUAL": "IDENTITY" + }, + "features": { + "action": { + "type": "ACTION", + "shape": "(20,)" + }, + "observation.image.0": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.1": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.2": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.state": { + "type": "STATE", + "shape": "(20,)" + } + }, + "stats": null, + "clip_values": { + "ENV": [ + -1.05, + 1.05 + ], + "STATE": [ + -1.05, + 1.05 + ] + }, + "chunk_size": 50, + "observation_mapping": { + "action.ee_quat_pos": "action", + "observations.ee_quat_pos": "observation.state", + "observations.images.cam_left": "observation.image.1", + "observations.images.cam_right": "observation.image.2", + "observations.images.cam_top": "observation.image.0" + }, + "observation_whitelist": [ + "observation.state", + "observation.image.0", + "observation.image.1", + "observation.image.2", + "action", + "task", + "subtask", + "dataset_index", + "training_mode" + ], + "observation_blacklist": null, + "transform_mapping": { + "action": [ + { + "type": "convert_to_6d_actions", + "action_key": "action", + "action_type": "EE_QUAT_POS_XYZW", + "input_type": "Dict", + "post_norm": false + }, + { + "type": "convert_to_6d_actions", + "action_key": "observation.state", + "action_type": "EE_QUAT_POS_XYZW", + "input_type": "Dict", + "post_norm": false + }, + { + "type": "delta_actions", + "state_key": "observation.state", + "action_key": "action", + "action_type": "EE_6D_POS", + "input_type": "Dict", + "post_norm": false, + "relative_to_state": true + } + ], + "observation.image.0": [ + { + "type": "random_resized_crop", + "height": 448, + "width": 448, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "observation.image.1": [ + { + "type": "random_resized_crop", + "height": 448, + "width": 448, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "observation.image.2": [ + { + "type": "random_resized_crop", + "height": 448, + "width": 448, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "task": [ + { + "type": "task_description_selector", + "task_key": "task", + "subtask_key": "subtask", + "output_key": "task", + "subtask_probability": 0.98, + "drop_subtask_key": true, + "has_subtask_key": "__task_description_has_subtask__", + "used_subtask_key": "__task_description_used_subtask__", + "input_type": "Dict", + "post_norm": false + } + ] + }, + "language_instruction": null, + "training_mode": "ROBOT_FLOWMATCH", + "action_type": "EE_QUAT_POS_XYZW", + "serialize_stats": false, + "video_decoder_cache_size": null, + "repo_id": "xdof/yam_station_merged", + "streaming": false, + "tolerance_s": 0.1, + "use_imagenet_stats": false, + "split_name": "train", + "prefetch_factor": 2, + "max_episodes": null + }, + "weight": 0.111242935144 + }, + { + "dataset": { + "datasets": [ + { + "dataset": { + "batch_size": 16, + "num_workers": 8, + "shuffle_buffer_size": 100, + "normalization_mapping": { + "ACTION": "ACTIONCHUNK_QUANTILE", + "ENV": "QUANTILE", + "STATE": "QUANTILE", + "VISUAL": "IDENTITY" + }, + "features": { + "action": { + "type": "ACTION", + "shape": "(20,)" + }, + "observation.image.0": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.1": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.2": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.state": { + "type": "STATE", + "shape": "(20,)" + } + }, + "stats": null, + "clip_values": { + "ENV": [ + -1.05, + 1.05 + ], + "STATE": [ + -1.05, + 1.05 + ] + }, + "chunk_size": 50, + "observation_mapping": { + "action.ee_quat_pos": "action", + "observations.ee_quat_pos": "observation.state", + "observations.images.cam_left": "observation.image.1", + "observations.images.cam_right": "observation.image.2", + "observations.images.cam_top": "observation.image.0" + }, + "observation_whitelist": [ + "observation.state", + "observation.image.0", + "observation.image.1", + "observation.image.2", + "action", + "task", + "subtask", + "dataset_index", + "training_mode" + ], + "observation_blacklist": null, + "transform_mapping": { + "action": [ + { + "type": "convert_to_6d_actions", + "action_key": "action", + "action_type": "EE_QUAT_POS_XYZW", + "input_type": "Dict", + "post_norm": false + }, + { + "type": "convert_to_6d_actions", + "action_key": "observation.state", + "action_type": "EE_QUAT_POS_XYZW", + "input_type": "Dict", + "post_norm": false + }, + { + "type": "delta_actions", + "state_key": "observation.state", + "action_key": "action", + "action_type": "EE_6D_POS", + "input_type": "Dict", + "post_norm": false, + "relative_to_state": true + } + ], + "observation.image.0": [ + { + "type": "random_resized_crop", + "height": 448, + "width": 448, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "observation.image.1": [ + { + "type": "random_resized_crop", + "height": 448, + "width": 448, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "observation.image.2": [ + { + "type": "random_resized_crop", + "height": 448, + "width": 448, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "task": [ + { + "type": "task_description_selector", + "task_key": "task", + "subtask_key": "subtask", + "output_key": "task", + "subtask_probability": 0.98, + "drop_subtask_key": true, + "has_subtask_key": "__task_description_has_subtask__", + "used_subtask_key": "__task_description_used_subtask__", + "input_type": "Dict", + "post_norm": false + } + ] + }, + "language_instruction": null, + "training_mode": "ROBOT_FLOWMATCH", + "action_type": "EE_QUAT_POS_XYZW", + "serialize_stats": false, + "video_decoder_cache_size": null, + "repo_id": "xdof/yam_box_wild/arrange_the_aromatherapy_jars_by_size_on_the_bedside_table", + "streaming": false, + "tolerance_s": 0.1, + "use_imagenet_stats": false, + "split_name": "train", + "prefetch_factor": 2, + "max_episodes": null + }, + "weight": 0.057120684641 + }, + { + "dataset": { + "batch_size": 16, + "num_workers": 8, + "shuffle_buffer_size": 100, + "normalization_mapping": { + "ACTION": "ACTIONCHUNK_QUANTILE", + "ENV": "QUANTILE", + "STATE": "QUANTILE", + "VISUAL": "IDENTITY" + }, + "features": { + "action": { + "type": "ACTION", + "shape": "(20,)" + }, + "observation.image.0": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.1": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.2": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.state": { + "type": "STATE", + "shape": "(20,)" + } + }, + "stats": null, + "clip_values": { + "ENV": [ + -1.05, + 1.05 + ], + "STATE": [ + -1.05, + 1.05 + ] + }, + "chunk_size": 50, + "observation_mapping": { + "action.ee_quat_pos": "action", + "observations.ee_quat_pos": "observation.state", + "observations.images.cam_left": "observation.image.1", + "observations.images.cam_right": "observation.image.2", + "observations.images.cam_top": "observation.image.0" + }, + "observation_whitelist": [ + "observation.state", + "observation.image.0", + "observation.image.1", + "observation.image.2", + "action", + "task", + "subtask", + "dataset_index", + "training_mode" + ], + "observation_blacklist": null, + "transform_mapping": { + "action": [ + { + "type": "convert_to_6d_actions", + "action_key": "action", + "action_type": "EE_QUAT_POS_XYZW", + "input_type": "Dict", + "post_norm": false + }, + { + "type": "convert_to_6d_actions", + "action_key": "observation.state", + "action_type": "EE_QUAT_POS_XYZW", + "input_type": "Dict", + "post_norm": false + }, + { + "type": "delta_actions", + "state_key": "observation.state", + "action_key": "action", + "action_type": "EE_6D_POS", + "input_type": "Dict", + "post_norm": false, + "relative_to_state": true + } + ], + "observation.image.0": [ + { + "type": "random_resized_crop", + "height": 448, + "width": 448, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "observation.image.1": [ + { + "type": "random_resized_crop", + "height": 448, + "width": 448, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "observation.image.2": [ + { + "type": "random_resized_crop", + "height": 448, + "width": 448, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "task": [ + { + "type": "task_description_selector", + "task_key": "task", + "subtask_key": "subtask", + "output_key": "task", + "subtask_probability": 0.98, + "drop_subtask_key": true, + "has_subtask_key": "__task_description_has_subtask__", + "used_subtask_key": "__task_description_used_subtask__", + "input_type": "Dict", + "post_norm": false + } + ] + }, + "language_instruction": null, + "training_mode": "ROBOT_FLOWMATCH", + "action_type": "EE_QUAT_POS_XYZW", + "serialize_stats": false, + "video_decoder_cache_size": null, + "repo_id": "xdof/yam_box_wild/arrange_the_lipsticks_in_the_organizer", + "streaming": false, + "tolerance_s": 0.1, + "use_imagenet_stats": false, + "split_name": "train", + "prefetch_factor": 2, + "max_episodes": null + }, + "weight": 0.057747014425 + }, + { + "dataset": { + "batch_size": 16, + "num_workers": 8, + "shuffle_buffer_size": 100, + "normalization_mapping": { + "ACTION": "ACTIONCHUNK_QUANTILE", + "ENV": "QUANTILE", + "STATE": "QUANTILE", + "VISUAL": "IDENTITY" + }, + "features": { + "action": { + "type": "ACTION", + "shape": "(20,)" + }, + "observation.image.0": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.1": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.2": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.state": { + "type": "STATE", + "shape": "(20,)" + } + }, + "stats": null, + "clip_values": { + "ENV": [ + -1.05, + 1.05 + ], + "STATE": [ + -1.05, + 1.05 + ] + }, + "chunk_size": 50, + "observation_mapping": { + "action.ee_quat_pos": "action", + "observations.ee_quat_pos": "observation.state", + "observations.images.cam_left": "observation.image.1", + "observations.images.cam_right": "observation.image.2", + "observations.images.cam_top": "observation.image.0" + }, + "observation_whitelist": [ + "observation.state", + "observation.image.0", + "observation.image.1", + "observation.image.2", + "action", + "task", + "subtask", + "dataset_index", + "training_mode" + ], + "observation_blacklist": null, + "transform_mapping": { + "action": [ + { + "type": "convert_to_6d_actions", + "action_key": "action", + "action_type": "EE_QUAT_POS_XYZW", + "input_type": "Dict", + "post_norm": false + }, + { + "type": "convert_to_6d_actions", + "action_key": "observation.state", + "action_type": "EE_QUAT_POS_XYZW", + "input_type": "Dict", + "post_norm": false + }, + { + "type": "delta_actions", + "state_key": "observation.state", + "action_key": "action", + "action_type": "EE_6D_POS", + "input_type": "Dict", + "post_norm": false, + "relative_to_state": true + } + ], + "observation.image.0": [ + { + "type": "random_resized_crop", + "height": 448, + "width": 448, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "observation.image.1": [ + { + "type": "random_resized_crop", + "height": 448, + "width": 448, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "observation.image.2": [ + { + "type": "random_resized_crop", + "height": 448, + "width": 448, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "task": [ + { + "type": "task_description_selector", + "task_key": "task", + "subtask_key": "subtask", + "output_key": "task", + "subtask_probability": 0.98, + "drop_subtask_key": true, + "has_subtask_key": "__task_description_has_subtask__", + "used_subtask_key": "__task_description_used_subtask__", + "input_type": "Dict", + "post_norm": false + } + ] + }, + "language_instruction": null, + "training_mode": "ROBOT_FLOWMATCH", + "action_type": "EE_QUAT_POS_XYZW", + "serialize_stats": false, + "video_decoder_cache_size": null, + "repo_id": "xdof/yam_box_wild/assemble_a_heart_with_lego", + "streaming": false, + "tolerance_s": 0.1, + "use_imagenet_stats": false, + "split_name": "train", + "prefetch_factor": 2, + "max_episodes": null + }, + "weight": 0.053378449905 + }, + { + "dataset": { + "batch_size": 16, + "num_workers": 8, + "shuffle_buffer_size": 100, + "normalization_mapping": { + "ACTION": "ACTIONCHUNK_QUANTILE", + "ENV": "QUANTILE", + "STATE": "QUANTILE", + "VISUAL": "IDENTITY" + }, + "features": { + "action": { + "type": "ACTION", + "shape": "(20,)" + }, + "observation.image.0": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.1": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.2": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.state": { + "type": "STATE", + "shape": "(20,)" + } + }, + "stats": null, + "clip_values": { + "ENV": [ + -1.05, + 1.05 + ], + "STATE": [ + -1.05, + 1.05 + ] + }, + "chunk_size": 50, + "observation_mapping": { + "action.ee_quat_pos": "action", + "observations.ee_quat_pos": "observation.state", + "observations.images.cam_left": "observation.image.1", + "observations.images.cam_right": "observation.image.2", + "observations.images.cam_top": "observation.image.0" + }, + "observation_whitelist": [ + "observation.state", + "observation.image.0", + "observation.image.1", + "observation.image.2", + "action", + "task", + "subtask", + "dataset_index", + "training_mode" + ], + "observation_blacklist": null, + "transform_mapping": { + "action": [ + { + "type": "convert_to_6d_actions", + "action_key": "action", + "action_type": "EE_QUAT_POS_XYZW", + "input_type": "Dict", + "post_norm": false + }, + { + "type": "convert_to_6d_actions", + "action_key": "observation.state", + "action_type": "EE_QUAT_POS_XYZW", + "input_type": "Dict", + "post_norm": false + }, + { + "type": "delta_actions", + "state_key": "observation.state", + "action_key": "action", + "action_type": "EE_6D_POS", + "input_type": "Dict", + "post_norm": false, + "relative_to_state": true + } + ], + "observation.image.0": [ + { + "type": "random_resized_crop", + "height": 448, + "width": 448, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "observation.image.1": [ + { + "type": "random_resized_crop", + "height": 448, + "width": 448, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "observation.image.2": [ + { + "type": "random_resized_crop", + "height": 448, + "width": 448, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "task": [ + { + "type": "task_description_selector", + "task_key": "task", + "subtask_key": "subtask", + "output_key": "task", + "subtask_probability": 0.98, + "drop_subtask_key": true, + "has_subtask_key": "__task_description_has_subtask__", + "used_subtask_key": "__task_description_used_subtask__", + "input_type": "Dict", + "post_norm": false + } + ] + }, + "language_instruction": null, + "training_mode": "ROBOT_FLOWMATCH", + "action_type": "EE_QUAT_POS_XYZW", + "serialize_stats": false, + "video_decoder_cache_size": null, + "repo_id": "xdof/yam_box_wild/assemble_the_bento_box", + "streaming": false, + "tolerance_s": 0.1, + "use_imagenet_stats": false, + "split_name": "train", + "prefetch_factor": 2, + "max_episodes": null + }, + "weight": 0.005016846214 + }, + { + "dataset": { + "batch_size": 16, + "num_workers": 8, + "shuffle_buffer_size": 100, + "normalization_mapping": { + "ACTION": "ACTIONCHUNK_QUANTILE", + "ENV": "QUANTILE", + "STATE": "QUANTILE", + "VISUAL": "IDENTITY" + }, + "features": { + "action": { + "type": "ACTION", + "shape": "(20,)" + }, + "observation.image.0": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.1": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.2": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.state": { + "type": "STATE", + "shape": "(20,)" + } + }, + "stats": null, + "clip_values": { + "ENV": [ + -1.05, + 1.05 + ], + "STATE": [ + -1.05, + 1.05 + ] + }, + "chunk_size": 50, + "observation_mapping": { + "action.ee_quat_pos": "action", + "observations.ee_quat_pos": "observation.state", + "observations.images.cam_left": "observation.image.1", + "observations.images.cam_right": "observation.image.2", + "observations.images.cam_top": "observation.image.0" + }, + "observation_whitelist": [ + "observation.state", + "observation.image.0", + "observation.image.1", + "observation.image.2", + "action", + "task", + "subtask", + "dataset_index", + "training_mode" + ], + "observation_blacklist": null, + "transform_mapping": { + "action": [ + { + "type": "convert_to_6d_actions", + "action_key": "action", + "action_type": "EE_QUAT_POS_XYZW", + "input_type": "Dict", + "post_norm": false + }, + { + "type": "convert_to_6d_actions", + "action_key": "observation.state", + "action_type": "EE_QUAT_POS_XYZW", + "input_type": "Dict", + "post_norm": false + }, + { + "type": "delta_actions", + "state_key": "observation.state", + "action_key": "action", + "action_type": "EE_6D_POS", + "input_type": "Dict", + "post_norm": false, + "relative_to_state": true + } + ], + "observation.image.0": [ + { + "type": "random_resized_crop", + "height": 448, + "width": 448, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "observation.image.1": [ + { + "type": "random_resized_crop", + "height": 448, + "width": 448, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "observation.image.2": [ + { + "type": "random_resized_crop", + "height": 448, + "width": 448, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "task": [ + { + "type": "task_description_selector", + "task_key": "task", + "subtask_key": "subtask", + "output_key": "task", + "subtask_probability": 0.98, + "drop_subtask_key": true, + "has_subtask_key": "__task_description_has_subtask__", + "used_subtask_key": "__task_description_used_subtask__", + "input_type": "Dict", + "post_norm": false + } + ] + }, + "language_instruction": null, + "training_mode": "ROBOT_FLOWMATCH", + "action_type": "EE_QUAT_POS_XYZW", + "serialize_stats": false, + "video_decoder_cache_size": null, + "repo_id": "xdof/yam_box_wild/clean_the_keyboard_with_a_cloth", + "streaming": false, + "tolerance_s": 0.1, + "use_imagenet_stats": false, + "split_name": "train", + "prefetch_factor": 2, + "max_episodes": null + }, + "weight": 0.056450215719 + }, + { + "dataset": { + "batch_size": 16, + "num_workers": 8, + "shuffle_buffer_size": 100, + "normalization_mapping": { + "ACTION": "ACTIONCHUNK_QUANTILE", + "ENV": "QUANTILE", + "STATE": "QUANTILE", + "VISUAL": "IDENTITY" + }, + "features": { + "action": { + "type": "ACTION", + "shape": "(20,)" + }, + "observation.image.0": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.1": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.2": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.state": { + "type": "STATE", + "shape": "(20,)" + } + }, + "stats": null, + "clip_values": { + "ENV": [ + -1.05, + 1.05 + ], + "STATE": [ + -1.05, + 1.05 + ] + }, + "chunk_size": 50, + "observation_mapping": { + "action.ee_quat_pos": "action", + "observations.ee_quat_pos": "observation.state", + "observations.images.cam_left": "observation.image.1", + "observations.images.cam_right": "observation.image.2", + "observations.images.cam_top": "observation.image.0" + }, + "observation_whitelist": [ + "observation.state", + "observation.image.0", + "observation.image.1", + "observation.image.2", + "action", + "task", + "subtask", + "dataset_index", + "training_mode" + ], + "observation_blacklist": null, + "transform_mapping": { + "action": [ + { + "type": "convert_to_6d_actions", + "action_key": "action", + "action_type": "EE_QUAT_POS_XYZW", + "input_type": "Dict", + "post_norm": false + }, + { + "type": "convert_to_6d_actions", + "action_key": "observation.state", + "action_type": "EE_QUAT_POS_XYZW", + "input_type": "Dict", + "post_norm": false + }, + { + "type": "delta_actions", + "state_key": "observation.state", + "action_key": "action", + "action_type": "EE_6D_POS", + "input_type": "Dict", + "post_norm": false, + "relative_to_state": true + } + ], + "observation.image.0": [ + { + "type": "random_resized_crop", + "height": 448, + "width": 448, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "observation.image.1": [ + { + "type": "random_resized_crop", + "height": 448, + "width": 448, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "observation.image.2": [ + { + "type": "random_resized_crop", + "height": 448, + "width": 448, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "task": [ + { + "type": "task_description_selector", + "task_key": "task", + "subtask_key": "subtask", + "output_key": "task", + "subtask_probability": 0.98, + "drop_subtask_key": true, + "has_subtask_key": "__task_description_has_subtask__", + "used_subtask_key": "__task_description_used_subtask__", + "input_type": "Dict", + "post_norm": false + } + ] + }, + "language_instruction": null, + "training_mode": "ROBOT_FLOWMATCH", + "action_type": "EE_QUAT_POS_XYZW", + "serialize_stats": false, + "video_decoder_cache_size": null, + "repo_id": "xdof/yam_box_wild/clean_the_monitor_with_a_duster", + "streaming": false, + "tolerance_s": 0.1, + "use_imagenet_stats": false, + "split_name": "train", + "prefetch_factor": 2, + "max_episodes": null + }, + "weight": 0.052462352671 + }, + { + "dataset": { + "batch_size": 16, + "num_workers": 8, + "shuffle_buffer_size": 100, + "normalization_mapping": { + "ACTION": "ACTIONCHUNK_QUANTILE", + "ENV": "QUANTILE", + "STATE": "QUANTILE", + "VISUAL": "IDENTITY" + }, + "features": { + "action": { + "type": "ACTION", + "shape": "(20,)" + }, + "observation.image.0": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.1": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.2": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.state": { + "type": "STATE", + "shape": "(20,)" + } + }, + "stats": null, + "clip_values": { + "ENV": [ + -1.05, + 1.05 + ], + "STATE": [ + -1.05, + 1.05 + ] + }, + "chunk_size": 50, + "observation_mapping": { + "action.ee_quat_pos": "action", + "observations.ee_quat_pos": "observation.state", + "observations.images.cam_left": "observation.image.1", + "observations.images.cam_right": "observation.image.2", + "observations.images.cam_top": "observation.image.0" + }, + "observation_whitelist": [ + "observation.state", + "observation.image.0", + "observation.image.1", + "observation.image.2", + "action", + "task", + "subtask", + "dataset_index", + "training_mode" + ], + "observation_blacklist": null, + "transform_mapping": { + "action": [ + { + "type": "convert_to_6d_actions", + "action_key": "action", + "action_type": "EE_QUAT_POS_XYZW", + "input_type": "Dict", + "post_norm": false + }, + { + "type": "convert_to_6d_actions", + "action_key": "observation.state", + "action_type": "EE_QUAT_POS_XYZW", + "input_type": "Dict", + "post_norm": false + }, + { + "type": "delta_actions", + "state_key": "observation.state", + "action_key": "action", + "action_type": "EE_6D_POS", + "input_type": "Dict", + "post_norm": false, + "relative_to_state": true + } + ], + "observation.image.0": [ + { + "type": "random_resized_crop", + "height": 448, + "width": 448, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "observation.image.1": [ + { + "type": "random_resized_crop", + "height": 448, + "width": 448, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "observation.image.2": [ + { + "type": "random_resized_crop", + "height": 448, + "width": 448, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "task": [ + { + "type": "task_description_selector", + "task_key": "task", + "subtask_key": "subtask", + "output_key": "task", + "subtask_probability": 0.98, + "drop_subtask_key": true, + "has_subtask_key": "__task_description_has_subtask__", + "used_subtask_key": "__task_description_used_subtask__", + "input_type": "Dict", + "post_norm": false + } + ] + }, + "language_instruction": null, + "training_mode": "ROBOT_FLOWMATCH", + "action_type": "EE_QUAT_POS_XYZW", + "serialize_stats": false, + "video_decoder_cache_size": null, + "repo_id": "xdof/yam_box_wild/clip_the_small_worksheets_together", + "streaming": false, + "tolerance_s": 0.1, + "use_imagenet_stats": false, + "split_name": "train", + "prefetch_factor": 2, + "max_episodes": null + }, + "weight": 0.047476759074 + }, + { + "dataset": { + "batch_size": 16, + "num_workers": 8, + "shuffle_buffer_size": 100, + "normalization_mapping": { + "ACTION": "ACTIONCHUNK_QUANTILE", + "ENV": "QUANTILE", + "STATE": "QUANTILE", + "VISUAL": "IDENTITY" + }, + "features": { + "action": { + "type": "ACTION", + "shape": "(20,)" + }, + "observation.image.0": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.1": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.2": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.state": { + "type": "STATE", + "shape": "(20,)" + } + }, + "stats": null, + "clip_values": { + "ENV": [ + -1.05, + 1.05 + ], + "STATE": [ + -1.05, + 1.05 + ] + }, + "chunk_size": 50, + "observation_mapping": { + "action.ee_quat_pos": "action", + "observations.ee_quat_pos": "observation.state", + "observations.images.cam_left": "observation.image.1", + "observations.images.cam_right": "observation.image.2", + "observations.images.cam_top": "observation.image.0" + }, + "observation_whitelist": [ + "observation.state", + "observation.image.0", + "observation.image.1", + "observation.image.2", + "action", + "task", + "subtask", + "dataset_index", + "training_mode" + ], + "observation_blacklist": null, + "transform_mapping": { + "action": [ + { + "type": "convert_to_6d_actions", + "action_key": "action", + "action_type": "EE_QUAT_POS_XYZW", + "input_type": "Dict", + "post_norm": false + }, + { + "type": "convert_to_6d_actions", + "action_key": "observation.state", + "action_type": "EE_QUAT_POS_XYZW", + "input_type": "Dict", + "post_norm": false + }, + { + "type": "delta_actions", + "state_key": "observation.state", + "action_key": "action", + "action_type": "EE_6D_POS", + "input_type": "Dict", + "post_norm": false, + "relative_to_state": true + } + ], + "observation.image.0": [ + { + "type": "random_resized_crop", + "height": 448, + "width": 448, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "observation.image.1": [ + { + "type": "random_resized_crop", + "height": 448, + "width": 448, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "observation.image.2": [ + { + "type": "random_resized_crop", + "height": 448, + "width": 448, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "task": [ + { + "type": "task_description_selector", + "task_key": "task", + "subtask_key": "subtask", + "output_key": "task", + "subtask_probability": 0.98, + "drop_subtask_key": true, + "has_subtask_key": "__task_description_has_subtask__", + "used_subtask_key": "__task_description_used_subtask__", + "input_type": "Dict", + "post_norm": false + } + ] + }, + "language_instruction": null, + "training_mode": "ROBOT_FLOWMATCH", + "action_type": "EE_QUAT_POS_XYZW", + "serialize_stats": false, + "video_decoder_cache_size": null, + "repo_id": "xdof/yam_box_wild/close_the_laptop_stand", + "streaming": false, + "tolerance_s": 0.1, + "use_imagenet_stats": false, + "split_name": "train", + "prefetch_factor": 2, + "max_episodes": null + }, + "weight": 0.055362425191 + }, + { + "dataset": { + "batch_size": 16, + "num_workers": 8, + "shuffle_buffer_size": 100, + "normalization_mapping": { + "ACTION": "ACTIONCHUNK_QUANTILE", + "ENV": "QUANTILE", + "STATE": "QUANTILE", + "VISUAL": "IDENTITY" + }, + "features": { + "action": { + "type": "ACTION", + "shape": "(20,)" + }, + "observation.image.0": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.1": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.2": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.state": { + "type": "STATE", + "shape": "(20,)" + } + }, + "stats": null, + "clip_values": { + "ENV": [ + -1.05, + 1.05 + ], + "STATE": [ + -1.05, + 1.05 + ] + }, + "chunk_size": 50, + "observation_mapping": { + "action.ee_quat_pos": "action", + "observations.ee_quat_pos": "observation.state", + "observations.images.cam_left": "observation.image.1", + "observations.images.cam_right": "observation.image.2", + "observations.images.cam_top": "observation.image.0" + }, + "observation_whitelist": [ + "observation.state", + "observation.image.0", + "observation.image.1", + "observation.image.2", + "action", + "task", + "subtask", + "dataset_index", + "training_mode" + ], + "observation_blacklist": null, + "transform_mapping": { + "action": [ + { + "type": "convert_to_6d_actions", + "action_key": "action", + "action_type": "EE_QUAT_POS_XYZW", + "input_type": "Dict", + "post_norm": false + }, + { + "type": "convert_to_6d_actions", + "action_key": "observation.state", + "action_type": "EE_QUAT_POS_XYZW", + "input_type": "Dict", + "post_norm": false + }, + { + "type": "delta_actions", + "state_key": "observation.state", + "action_key": "action", + "action_type": "EE_6D_POS", + "input_type": "Dict", + "post_norm": false, + "relative_to_state": true + } + ], + "observation.image.0": [ + { + "type": "random_resized_crop", + "height": 448, + "width": 448, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "observation.image.1": [ + { + "type": "random_resized_crop", + "height": 448, + "width": 448, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "observation.image.2": [ + { + "type": "random_resized_crop", + "height": 448, + "width": 448, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "task": [ + { + "type": "task_description_selector", + "task_key": "task", + "subtask_key": "subtask", + "output_key": "task", + "subtask_probability": 0.98, + "drop_subtask_key": true, + "has_subtask_key": "__task_description_has_subtask__", + "used_subtask_key": "__task_description_used_subtask__", + "input_type": "Dict", + "post_norm": false + } + ] + }, + "language_instruction": null, + "training_mode": "ROBOT_FLOWMATCH", + "action_type": "EE_QUAT_POS_XYZW", + "serialize_stats": false, + "video_decoder_cache_size": null, + "repo_id": "xdof/yam_box_wild/cut_the_label_with_scissors", + "streaming": false, + "tolerance_s": 0.1, + "use_imagenet_stats": false, + "split_name": "train", + "prefetch_factor": 2, + "max_episodes": null + }, + "weight": 0.051068837816 + }, + { + "dataset": { + "batch_size": 16, + "num_workers": 8, + "shuffle_buffer_size": 100, + "normalization_mapping": { + "ACTION": "ACTIONCHUNK_QUANTILE", + "ENV": "QUANTILE", + "STATE": "QUANTILE", + "VISUAL": "IDENTITY" + }, + "features": { + "action": { + "type": "ACTION", + "shape": "(20,)" + }, + "observation.image.0": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.1": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.2": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.state": { + "type": "STATE", + "shape": "(20,)" + } + }, + "stats": null, + "clip_values": { + "ENV": [ + -1.05, + 1.05 + ], + "STATE": [ + -1.05, + 1.05 + ] + }, + "chunk_size": 50, + "observation_mapping": { + "action.ee_quat_pos": "action", + "observations.ee_quat_pos": "observation.state", + "observations.images.cam_left": "observation.image.1", + "observations.images.cam_right": "observation.image.2", + "observations.images.cam_top": "observation.image.0" + }, + "observation_whitelist": [ + "observation.state", + "observation.image.0", + "observation.image.1", + "observation.image.2", + "action", + "task", + "subtask", + "dataset_index", + "training_mode" + ], + "observation_blacklist": null, + "transform_mapping": { + "action": [ + { + "type": "convert_to_6d_actions", + "action_key": "action", + "action_type": "EE_QUAT_POS_XYZW", + "input_type": "Dict", + "post_norm": false + }, + { + "type": "convert_to_6d_actions", + "action_key": "observation.state", + "action_type": "EE_QUAT_POS_XYZW", + "input_type": "Dict", + "post_norm": false + }, + { + "type": "delta_actions", + "state_key": "observation.state", + "action_key": "action", + "action_type": "EE_6D_POS", + "input_type": "Dict", + "post_norm": false, + "relative_to_state": true + } + ], + "observation.image.0": [ + { + "type": "random_resized_crop", + "height": 448, + "width": 448, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "observation.image.1": [ + { + "type": "random_resized_crop", + "height": 448, + "width": 448, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "observation.image.2": [ + { + "type": "random_resized_crop", + "height": 448, + "width": 448, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "task": [ + { + "type": "task_description_selector", + "task_key": "task", + "subtask_key": "subtask", + "output_key": "task", + "subtask_probability": 0.98, + "drop_subtask_key": true, + "has_subtask_key": "__task_description_has_subtask__", + "used_subtask_key": "__task_description_used_subtask__", + "input_type": "Dict", + "post_norm": false + } + ] + }, + "language_instruction": null, + "training_mode": "ROBOT_FLOWMATCH", + "action_type": "EE_QUAT_POS_XYZW", + "serialize_stats": false, + "video_decoder_cache_size": null, + "repo_id": "xdof/yam_box_wild/distribute_batteries_near_wireless_mice", + "streaming": false, + "tolerance_s": 0.1, + "use_imagenet_stats": false, + "split_name": "train", + "prefetch_factor": 2, + "max_episodes": null + }, + "weight": 0.053098722736 + }, + { + "dataset": { + "batch_size": 16, + "num_workers": 8, + "shuffle_buffer_size": 100, + "normalization_mapping": { + "ACTION": "ACTIONCHUNK_QUANTILE", + "ENV": "QUANTILE", + "STATE": "QUANTILE", + "VISUAL": "IDENTITY" + }, + "features": { + "action": { + "type": "ACTION", + "shape": "(20,)" + }, + "observation.image.0": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.1": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.2": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.state": { + "type": "STATE", + "shape": "(20,)" + } + }, + "stats": null, + "clip_values": { + "ENV": [ + -1.05, + 1.05 + ], + "STATE": [ + -1.05, + 1.05 + ] + }, + "chunk_size": 50, + "observation_mapping": { + "action.ee_quat_pos": "action", + "observations.ee_quat_pos": "observation.state", + "observations.images.cam_left": "observation.image.1", + "observations.images.cam_right": "observation.image.2", + "observations.images.cam_top": "observation.image.0" + }, + "observation_whitelist": [ + "observation.state", + "observation.image.0", + "observation.image.1", + "observation.image.2", + "action", + "task", + "subtask", + "dataset_index", + "training_mode" + ], + "observation_blacklist": null, + "transform_mapping": { + "action": [ + { + "type": "convert_to_6d_actions", + "action_key": "action", + "action_type": "EE_QUAT_POS_XYZW", + "input_type": "Dict", + "post_norm": false + }, + { + "type": "convert_to_6d_actions", + "action_key": "observation.state", + "action_type": "EE_QUAT_POS_XYZW", + "input_type": "Dict", + "post_norm": false + }, + { + "type": "delta_actions", + "state_key": "observation.state", + "action_key": "action", + "action_type": "EE_6D_POS", + "input_type": "Dict", + "post_norm": false, + "relative_to_state": true + } + ], + "observation.image.0": [ + { + "type": "random_resized_crop", + "height": 448, + "width": 448, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "observation.image.1": [ + { + "type": "random_resized_crop", + "height": 448, + "width": 448, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "observation.image.2": [ + { + "type": "random_resized_crop", + "height": 448, + "width": 448, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "task": [ + { + "type": "task_description_selector", + "task_key": "task", + "subtask_key": "subtask", + "output_key": "task", + "subtask_probability": 0.98, + "drop_subtask_key": true, + "has_subtask_key": "__task_description_has_subtask__", + "used_subtask_key": "__task_description_used_subtask__", + "input_type": "Dict", + "post_norm": false + } + ] + }, + "language_instruction": null, + "training_mode": "ROBOT_FLOWMATCH", + "action_type": "EE_QUAT_POS_XYZW", + "serialize_stats": false, + "video_decoder_cache_size": null, + "repo_id": "xdof/yam_box_wild/distribute_the_index_tabs_onto_each_document", + "streaming": false, + "tolerance_s": 0.1, + "use_imagenet_stats": false, + "split_name": "train", + "prefetch_factor": 2, + "max_episodes": null + }, + "weight": 0.044634575051 + }, + { + "dataset": { + "batch_size": 16, + "num_workers": 8, + "shuffle_buffer_size": 100, + "normalization_mapping": { + "ACTION": "ACTIONCHUNK_QUANTILE", + "ENV": "QUANTILE", + "STATE": "QUANTILE", + "VISUAL": "IDENTITY" + }, + "features": { + "action": { + "type": "ACTION", + "shape": "(20,)" + }, + "observation.image.0": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.1": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.2": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.state": { + "type": "STATE", + "shape": "(20,)" + } + }, + "stats": null, + "clip_values": { + "ENV": [ + -1.05, + 1.05 + ], + "STATE": [ + -1.05, + 1.05 + ] + }, + "chunk_size": 50, + "observation_mapping": { + "action.ee_quat_pos": "action", + "observations.ee_quat_pos": "observation.state", + "observations.images.cam_left": "observation.image.1", + "observations.images.cam_right": "observation.image.2", + "observations.images.cam_top": "observation.image.0" + }, + "observation_whitelist": [ + "observation.state", + "observation.image.0", + "observation.image.1", + "observation.image.2", + "action", + "task", + "subtask", + "dataset_index", + "training_mode" + ], + "observation_blacklist": null, + "transform_mapping": { + "action": [ + { + "type": "convert_to_6d_actions", + "action_key": "action", + "action_type": "EE_QUAT_POS_XYZW", + "input_type": "Dict", + "post_norm": false + }, + { + "type": "convert_to_6d_actions", + "action_key": "observation.state", + "action_type": "EE_QUAT_POS_XYZW", + "input_type": "Dict", + "post_norm": false + }, + { + "type": "delta_actions", + "state_key": "observation.state", + "action_key": "action", + "action_type": "EE_6D_POS", + "input_type": "Dict", + "post_norm": false, + "relative_to_state": true + } + ], + "observation.image.0": [ + { + "type": "random_resized_crop", + "height": 448, + "width": 448, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "observation.image.1": [ + { + "type": "random_resized_crop", + "height": 448, + "width": 448, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "observation.image.2": [ + { + "type": "random_resized_crop", + "height": 448, + "width": 448, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "task": [ + { + "type": "task_description_selector", + "task_key": "task", + "subtask_key": "subtask", + "output_key": "task", + "subtask_probability": 0.98, + "drop_subtask_key": true, + "has_subtask_key": "__task_description_has_subtask__", + "used_subtask_key": "__task_description_used_subtask__", + "input_type": "Dict", + "post_norm": false + } + ] + }, + "language_instruction": null, + "training_mode": "ROBOT_FLOWMATCH", + "action_type": "EE_QUAT_POS_XYZW", + "serialize_stats": false, + "video_decoder_cache_size": null, + "repo_id": "xdof/yam_box_wild/erase_the_pencil_marks_with_an_eraser", + "streaming": false, + "tolerance_s": 0.1, + "use_imagenet_stats": false, + "split_name": "train", + "prefetch_factor": 2, + "max_episodes": null + }, + "weight": 0.047248690049 + }, + { + "dataset": { + "batch_size": 16, + "num_workers": 8, + "shuffle_buffer_size": 100, + "normalization_mapping": { + "ACTION": "ACTIONCHUNK_QUANTILE", + "ENV": "QUANTILE", + "STATE": "QUANTILE", + "VISUAL": "IDENTITY" + }, + "features": { + "action": { + "type": "ACTION", + "shape": "(20,)" + }, + "observation.image.0": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.1": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.2": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.state": { + "type": "STATE", + "shape": "(20,)" + } + }, + "stats": null, + "clip_values": { + "ENV": [ + -1.05, + 1.05 + ], + "STATE": [ + -1.05, + 1.05 + ] + }, + "chunk_size": 50, + "observation_mapping": { + "action.ee_quat_pos": "action", + "observations.ee_quat_pos": "observation.state", + "observations.images.cam_left": "observation.image.1", + "observations.images.cam_right": "observation.image.2", + "observations.images.cam_top": "observation.image.0" + }, + "observation_whitelist": [ + "observation.state", + "observation.image.0", + "observation.image.1", + "observation.image.2", + "action", + "task", + "subtask", + "dataset_index", + "training_mode" + ], + "observation_blacklist": null, + "transform_mapping": { + "action": [ + { + "type": "convert_to_6d_actions", + "action_key": "action", + "action_type": "EE_QUAT_POS_XYZW", + "input_type": "Dict", + "post_norm": false + }, + { + "type": "convert_to_6d_actions", + "action_key": "observation.state", + "action_type": "EE_QUAT_POS_XYZW", + "input_type": "Dict", + "post_norm": false + }, + { + "type": "delta_actions", + "state_key": "observation.state", + "action_key": "action", + "action_type": "EE_6D_POS", + "input_type": "Dict", + "post_norm": false, + "relative_to_state": true + } + ], + "observation.image.0": [ + { + "type": "random_resized_crop", + "height": 448, + "width": 448, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "observation.image.1": [ + { + "type": "random_resized_crop", + "height": 448, + "width": 448, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "observation.image.2": [ + { + "type": "random_resized_crop", + "height": 448, + "width": 448, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "task": [ + { + "type": "task_description_selector", + "task_key": "task", + "subtask_key": "subtask", + "output_key": "task", + "subtask_probability": 0.98, + "drop_subtask_key": true, + "has_subtask_key": "__task_description_has_subtask__", + "used_subtask_key": "__task_description_used_subtask__", + "input_type": "Dict", + "post_norm": false + } + ] + }, + "language_instruction": null, + "training_mode": "ROBOT_FLOWMATCH", + "action_type": "EE_QUAT_POS_XYZW", + "serialize_stats": false, + "video_decoder_cache_size": null, + "repo_id": "xdof/yam_box_wild/folding_tshirt_pile_and_stacking", + "streaming": false, + "tolerance_s": 0.1, + "use_imagenet_stats": false, + "split_name": "train", + "prefetch_factor": 2, + "max_episodes": null + }, + "weight": 0.038092659404 + }, + { + "dataset": { + "batch_size": 16, + "num_workers": 8, + "shuffle_buffer_size": 100, + "normalization_mapping": { + "ACTION": "ACTIONCHUNK_QUANTILE", + "ENV": "QUANTILE", + "STATE": "QUANTILE", + "VISUAL": "IDENTITY" + }, + "features": { + "action": { + "type": "ACTION", + "shape": "(20,)" + }, + "observation.image.0": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.1": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.2": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.state": { + "type": "STATE", + "shape": "(20,)" + } + }, + "stats": null, + "clip_values": { + "ENV": [ + -1.05, + 1.05 + ], + "STATE": [ + -1.05, + 1.05 + ] + }, + "chunk_size": 50, + "observation_mapping": { + "action.ee_quat_pos": "action", + "observations.ee_quat_pos": "observation.state", + "observations.images.cam_left": "observation.image.1", + "observations.images.cam_right": "observation.image.2", + "observations.images.cam_top": "observation.image.0" + }, + "observation_whitelist": [ + "observation.state", + "observation.image.0", + "observation.image.1", + "observation.image.2", + "action", + "task", + "subtask", + "dataset_index", + "training_mode" + ], + "observation_blacklist": null, + "transform_mapping": { + "action": [ + { + "type": "convert_to_6d_actions", + "action_key": "action", + "action_type": "EE_QUAT_POS_XYZW", + "input_type": "Dict", + "post_norm": false + }, + { + "type": "convert_to_6d_actions", + "action_key": "observation.state", + "action_type": "EE_QUAT_POS_XYZW", + "input_type": "Dict", + "post_norm": false + }, + { + "type": "delta_actions", + "state_key": "observation.state", + "action_key": "action", + "action_type": "EE_6D_POS", + "input_type": "Dict", + "post_norm": false, + "relative_to_state": true + } + ], + "observation.image.0": [ + { + "type": "random_resized_crop", + "height": 448, + "width": 448, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "observation.image.1": [ + { + "type": "random_resized_crop", + "height": 448, + "width": 448, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "observation.image.2": [ + { + "type": "random_resized_crop", + "height": 448, + "width": 448, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "task": [ + { + "type": "task_description_selector", + "task_key": "task", + "subtask_key": "subtask", + "output_key": "task", + "subtask_probability": 0.98, + "drop_subtask_key": true, + "has_subtask_key": "__task_description_has_subtask__", + "used_subtask_key": "__task_description_used_subtask__", + "input_type": "Dict", + "post_norm": false + } + ] + }, + "language_instruction": null, + "training_mode": "ROBOT_FLOWMATCH", + "action_type": "EE_QUAT_POS_XYZW", + "serialize_stats": false, + "video_decoder_cache_size": null, + "repo_id": "xdof/yam_box_wild/insert_refill_pages_into_the_notebook", + "streaming": false, + "tolerance_s": 0.1, + "use_imagenet_stats": false, + "split_name": "train", + "prefetch_factor": 2, + "max_episodes": null + }, + "weight": 0.053919963964 + }, + { + "dataset": { + "batch_size": 16, + "num_workers": 8, + "shuffle_buffer_size": 100, + "normalization_mapping": { + "ACTION": "ACTIONCHUNK_QUANTILE", + "ENV": "QUANTILE", + "STATE": "QUANTILE", + "VISUAL": "IDENTITY" + }, + "features": { + "action": { + "type": "ACTION", + "shape": "(20,)" + }, + "observation.image.0": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.1": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.2": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.state": { + "type": "STATE", + "shape": "(20,)" + } + }, + "stats": null, + "clip_values": { + "ENV": [ + -1.05, + 1.05 + ], + "STATE": [ + -1.05, + 1.05 + ] + }, + "chunk_size": 50, + "observation_mapping": { + "action.ee_quat_pos": "action", + "observations.ee_quat_pos": "observation.state", + "observations.images.cam_left": "observation.image.1", + "observations.images.cam_right": "observation.image.2", + "observations.images.cam_top": "observation.image.0" + }, + "observation_whitelist": [ + "observation.state", + "observation.image.0", + "observation.image.1", + "observation.image.2", + "action", + "task", + "subtask", + "dataset_index", + "training_mode" + ], + "observation_blacklist": null, + "transform_mapping": { + "action": [ + { + "type": "convert_to_6d_actions", + "action_key": "action", + "action_type": "EE_QUAT_POS_XYZW", + "input_type": "Dict", + "post_norm": false + }, + { + "type": "convert_to_6d_actions", + "action_key": "observation.state", + "action_type": "EE_QUAT_POS_XYZW", + "input_type": "Dict", + "post_norm": false + }, + { + "type": "delta_actions", + "state_key": "observation.state", + "action_key": "action", + "action_type": "EE_6D_POS", + "input_type": "Dict", + "post_norm": false, + "relative_to_state": true + } + ], + "observation.image.0": [ + { + "type": "random_resized_crop", + "height": 448, + "width": 448, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "observation.image.1": [ + { + "type": "random_resized_crop", + "height": 448, + "width": 448, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "observation.image.2": [ + { + "type": "random_resized_crop", + "height": 448, + "width": 448, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "task": [ + { + "type": "task_description_selector", + "task_key": "task", + "subtask_key": "subtask", + "output_key": "task", + "subtask_probability": 0.98, + "drop_subtask_key": true, + "has_subtask_key": "__task_description_has_subtask__", + "used_subtask_key": "__task_description_used_subtask__", + "input_type": "Dict", + "post_norm": false + } + ] + }, + "language_instruction": null, + "training_mode": "ROBOT_FLOWMATCH", + "action_type": "EE_QUAT_POS_XYZW", + "serialize_stats": false, + "video_decoder_cache_size": null, + "repo_id": "xdof/yam_box_wild/insert_the_pens_into_the_pen_caps", + "streaming": false, + "tolerance_s": 0.1, + "use_imagenet_stats": false, + "split_name": "train", + "prefetch_factor": 2, + "max_episodes": null + }, + "weight": 0.059065137524 + }, + { + "dataset": { + "batch_size": 16, + "num_workers": 8, + "shuffle_buffer_size": 100, + "normalization_mapping": { + "ACTION": "ACTIONCHUNK_QUANTILE", + "ENV": "QUANTILE", + "STATE": "QUANTILE", + "VISUAL": "IDENTITY" + }, + "features": { + "action": { + "type": "ACTION", + "shape": "(20,)" + }, + "observation.image.0": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.1": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.2": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.state": { + "type": "STATE", + "shape": "(20,)" + } + }, + "stats": null, + "clip_values": { + "ENV": [ + -1.05, + 1.05 + ], + "STATE": [ + -1.05, + 1.05 + ] + }, + "chunk_size": 50, + "observation_mapping": { + "action.ee_quat_pos": "action", + "observations.ee_quat_pos": "observation.state", + "observations.images.cam_left": "observation.image.1", + "observations.images.cam_right": "observation.image.2", + "observations.images.cam_top": "observation.image.0" + }, + "observation_whitelist": [ + "observation.state", + "observation.image.0", + "observation.image.1", + "observation.image.2", + "action", + "task", + "subtask", + "dataset_index", + "training_mode" + ], + "observation_blacklist": null, + "transform_mapping": { + "action": [ + { + "type": "convert_to_6d_actions", + "action_key": "action", + "action_type": "EE_QUAT_POS_XYZW", + "input_type": "Dict", + "post_norm": false + }, + { + "type": "convert_to_6d_actions", + "action_key": "observation.state", + "action_type": "EE_QUAT_POS_XYZW", + "input_type": "Dict", + "post_norm": false + }, + { + "type": "delta_actions", + "state_key": "observation.state", + "action_key": "action", + "action_type": "EE_6D_POS", + "input_type": "Dict", + "post_norm": false, + "relative_to_state": true + } + ], + "observation.image.0": [ + { + "type": "random_resized_crop", + "height": 448, + "width": 448, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "observation.image.1": [ + { + "type": "random_resized_crop", + "height": 448, + "width": 448, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "observation.image.2": [ + { + "type": "random_resized_crop", + "height": 448, + "width": 448, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "task": [ + { + "type": "task_description_selector", + "task_key": "task", + "subtask_key": "subtask", + "output_key": "task", + "subtask_probability": 0.98, + "drop_subtask_key": true, + "has_subtask_key": "__task_description_has_subtask__", + "used_subtask_key": "__task_description_used_subtask__", + "input_type": "Dict", + "post_norm": false + } + ] + }, + "language_instruction": null, + "training_mode": "ROBOT_FLOWMATCH", + "action_type": "EE_QUAT_POS_XYZW", + "serialize_stats": false, + "video_decoder_cache_size": null, + "repo_id": "xdof/yam_box_wild/insert_the_wireless_bluetooth_earbuds_into_the_charging_case", + "streaming": false, + "tolerance_s": 0.1, + "use_imagenet_stats": false, + "split_name": "train", + "prefetch_factor": 2, + "max_episodes": null + }, + "weight": 0.055727920567 + }, + { + "dataset": { + "batch_size": 16, + "num_workers": 8, + "shuffle_buffer_size": 100, + "normalization_mapping": { + "ACTION": "ACTIONCHUNK_QUANTILE", + "ENV": "QUANTILE", + "STATE": "QUANTILE", + "VISUAL": "IDENTITY" + }, + "features": { + "action": { + "type": "ACTION", + "shape": "(20,)" + }, + "observation.image.0": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.1": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.2": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.state": { + "type": "STATE", + "shape": "(20,)" + } + }, + "stats": null, + "clip_values": { + "ENV": [ + -1.05, + 1.05 + ], + "STATE": [ + -1.05, + 1.05 + ] + }, + "chunk_size": 50, + "observation_mapping": { + "action.ee_quat_pos": "action", + "observations.ee_quat_pos": "observation.state", + "observations.images.cam_left": "observation.image.1", + "observations.images.cam_right": "observation.image.2", + "observations.images.cam_top": "observation.image.0" + }, + "observation_whitelist": [ + "observation.state", + "observation.image.0", + "observation.image.1", + "observation.image.2", + "action", + "task", + "subtask", + "dataset_index", + "training_mode" + ], + "observation_blacklist": null, + "transform_mapping": { + "action": [ + { + "type": "convert_to_6d_actions", + "action_key": "action", + "action_type": "EE_QUAT_POS_XYZW", + "input_type": "Dict", + "post_norm": false + }, + { + "type": "convert_to_6d_actions", + "action_key": "observation.state", + "action_type": "EE_QUAT_POS_XYZW", + "input_type": "Dict", + "post_norm": false + }, + { + "type": "delta_actions", + "state_key": "observation.state", + "action_key": "action", + "action_type": "EE_6D_POS", + "input_type": "Dict", + "post_norm": false, + "relative_to_state": true + } + ], + "observation.image.0": [ + { + "type": "random_resized_crop", + "height": 448, + "width": 448, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "observation.image.1": [ + { + "type": "random_resized_crop", + "height": 448, + "width": 448, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "observation.image.2": [ + { + "type": "random_resized_crop", + "height": 448, + "width": 448, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "task": [ + { + "type": "task_description_selector", + "task_key": "task", + "subtask_key": "subtask", + "output_key": "task", + "subtask_probability": 0.98, + "drop_subtask_key": true, + "has_subtask_key": "__task_description_has_subtask__", + "used_subtask_key": "__task_description_used_subtask__", + "input_type": "Dict", + "post_norm": false + } + ] + }, + "language_instruction": null, + "training_mode": "ROBOT_FLOWMATCH", + "action_type": "EE_QUAT_POS_XYZW", + "serialize_stats": false, + "video_decoder_cache_size": null, + "repo_id": "xdof/yam_box_wild/line_up_the_cups_on_the_drying_mat", + "streaming": false, + "tolerance_s": 0.1, + "use_imagenet_stats": false, + "split_name": "train", + "prefetch_factor": 2, + "max_episodes": null + }, + "weight": 0.049462695351 + }, + { + "dataset": { + "batch_size": 16, + "num_workers": 8, + "shuffle_buffer_size": 100, + "normalization_mapping": { + "ACTION": "ACTIONCHUNK_QUANTILE", + "ENV": "QUANTILE", + "STATE": "QUANTILE", + "VISUAL": "IDENTITY" + }, + "features": { + "action": { + "type": "ACTION", + "shape": "(20,)" + }, + "observation.image.0": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.1": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.2": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.state": { + "type": "STATE", + "shape": "(20,)" + } + }, + "stats": null, + "clip_values": { + "ENV": [ + -1.05, + 1.05 + ], + "STATE": [ + -1.05, + 1.05 + ] + }, + "chunk_size": 50, + "observation_mapping": { + "action.ee_quat_pos": "action", + "observations.ee_quat_pos": "observation.state", + "observations.images.cam_left": "observation.image.1", + "observations.images.cam_right": "observation.image.2", + "observations.images.cam_top": "observation.image.0" + }, + "observation_whitelist": [ + "observation.state", + "observation.image.0", + "observation.image.1", + "observation.image.2", + "action", + "task", + "subtask", + "dataset_index", + "training_mode" + ], + "observation_blacklist": null, + "transform_mapping": { + "action": [ + { + "type": "convert_to_6d_actions", + "action_key": "action", + "action_type": "EE_QUAT_POS_XYZW", + "input_type": "Dict", + "post_norm": false + }, + { + "type": "convert_to_6d_actions", + "action_key": "observation.state", + "action_type": "EE_QUAT_POS_XYZW", + "input_type": "Dict", + "post_norm": false + }, + { + "type": "delta_actions", + "state_key": "observation.state", + "action_key": "action", + "action_type": "EE_6D_POS", + "input_type": "Dict", + "post_norm": false, + "relative_to_state": true + } + ], + "observation.image.0": [ + { + "type": "random_resized_crop", + "height": 448, + "width": 448, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "observation.image.1": [ + { + "type": "random_resized_crop", + "height": 448, + "width": 448, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "observation.image.2": [ + { + "type": "random_resized_crop", + "height": 448, + "width": 448, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "task": [ + { + "type": "task_description_selector", + "task_key": "task", + "subtask_key": "subtask", + "output_key": "task", + "subtask_probability": 0.98, + "drop_subtask_key": true, + "has_subtask_key": "__task_description_has_subtask__", + "used_subtask_key": "__task_description_used_subtask__", + "input_type": "Dict", + "post_norm": false + } + ] + }, + "language_instruction": null, + "training_mode": "ROBOT_FLOWMATCH", + "action_type": "EE_QUAT_POS_XYZW", + "serialize_stats": false, + "video_decoder_cache_size": null, + "repo_id": "xdof/yam_box_wild/load_the_batteries_into_the_calculator", + "streaming": false, + "tolerance_s": 0.1, + "use_imagenet_stats": false, + "split_name": "train", + "prefetch_factor": 2, + "max_episodes": null + }, + "weight": 0.046792305473 + }, + { + "dataset": { + "batch_size": 16, + "num_workers": 8, + "shuffle_buffer_size": 100, + "normalization_mapping": { + "ACTION": "ACTIONCHUNK_QUANTILE", + "ENV": "QUANTILE", + "STATE": "QUANTILE", + "VISUAL": "IDENTITY" + }, + "features": { + "action": { + "type": "ACTION", + "shape": "(20,)" + }, + "observation.image.0": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.1": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.2": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.state": { + "type": "STATE", + "shape": "(20,)" + } + }, + "stats": null, + "clip_values": { + "ENV": [ + -1.05, + 1.05 + ], + "STATE": [ + -1.05, + 1.05 + ] + }, + "chunk_size": 50, + "observation_mapping": { + "action.ee_quat_pos": "action", + "observations.ee_quat_pos": "observation.state", + "observations.images.cam_left": "observation.image.1", + "observations.images.cam_right": "observation.image.2", + "observations.images.cam_top": "observation.image.0" + }, + "observation_whitelist": [ + "observation.state", + "observation.image.0", + "observation.image.1", + "observation.image.2", + "action", + "task", + "subtask", + "dataset_index", + "training_mode" + ], + "observation_blacklist": null, + "transform_mapping": { + "action": [ + { + "type": "convert_to_6d_actions", + "action_key": "action", + "action_type": "EE_QUAT_POS_XYZW", + "input_type": "Dict", + "post_norm": false + }, + { + "type": "convert_to_6d_actions", + "action_key": "observation.state", + "action_type": "EE_QUAT_POS_XYZW", + "input_type": "Dict", + "post_norm": false + }, + { + "type": "delta_actions", + "state_key": "observation.state", + "action_key": "action", + "action_type": "EE_6D_POS", + "input_type": "Dict", + "post_norm": false, + "relative_to_state": true + } + ], + "observation.image.0": [ + { + "type": "random_resized_crop", + "height": 448, + "width": 448, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "observation.image.1": [ + { + "type": "random_resized_crop", + "height": 448, + "width": 448, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "observation.image.2": [ + { + "type": "random_resized_crop", + "height": 448, + "width": 448, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "task": [ + { + "type": "task_description_selector", + "task_key": "task", + "subtask_key": "subtask", + "output_key": "task", + "subtask_probability": 0.98, + "drop_subtask_key": true, + "has_subtask_key": "__task_description_has_subtask__", + "used_subtask_key": "__task_description_used_subtask__", + "input_type": "Dict", + "post_norm": false + } + ] + }, + "language_instruction": null, + "training_mode": "ROBOT_FLOWMATCH", + "action_type": "EE_QUAT_POS_XYZW", + "serialize_stats": false, + "video_decoder_cache_size": null, + "repo_id": "xdof/yam_box_wild/neatly_place_the_solid_glue_into_the_storage_box", + "streaming": false, + "tolerance_s": 0.1, + "use_imagenet_stats": false, + "split_name": "train", + "prefetch_factor": 2, + "max_episodes": null + }, + "weight": 0.059223921885 + }, + { + "dataset": { + "batch_size": 16, + "num_workers": 8, + "shuffle_buffer_size": 100, + "normalization_mapping": { + "ACTION": "ACTIONCHUNK_QUANTILE", + "ENV": "QUANTILE", + "STATE": "QUANTILE", + "VISUAL": "IDENTITY" + }, + "features": { + "action": { + "type": "ACTION", + "shape": "(20,)" + }, + "observation.image.0": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.1": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.2": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.state": { + "type": "STATE", + "shape": "(20,)" + } + }, + "stats": null, + "clip_values": { + "ENV": [ + -1.05, + 1.05 + ], + "STATE": [ + -1.05, + 1.05 + ] + }, + "chunk_size": 50, + "observation_mapping": { + "action.ee_quat_pos": "action", + "observations.ee_quat_pos": "observation.state", + "observations.images.cam_left": "observation.image.1", + "observations.images.cam_right": "observation.image.2", + "observations.images.cam_top": "observation.image.0" + }, + "observation_whitelist": [ + "observation.state", + "observation.image.0", + "observation.image.1", + "observation.image.2", + "action", + "task", + "subtask", + "dataset_index", + "training_mode" + ], + "observation_blacklist": null, + "transform_mapping": { + "action": [ + { + "type": "convert_to_6d_actions", + "action_key": "action", + "action_type": "EE_QUAT_POS_XYZW", + "input_type": "Dict", + "post_norm": false + }, + { + "type": "convert_to_6d_actions", + "action_key": "observation.state", + "action_type": "EE_QUAT_POS_XYZW", + "input_type": "Dict", + "post_norm": false + }, + { + "type": "delta_actions", + "state_key": "observation.state", + "action_key": "action", + "action_type": "EE_6D_POS", + "input_type": "Dict", + "post_norm": false, + "relative_to_state": true + } + ], + "observation.image.0": [ + { + "type": "random_resized_crop", + "height": 448, + "width": 448, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "observation.image.1": [ + { + "type": "random_resized_crop", + "height": 448, + "width": 448, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "observation.image.2": [ + { + "type": "random_resized_crop", + "height": 448, + "width": 448, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "task": [ + { + "type": "task_description_selector", + "task_key": "task", + "subtask_key": "subtask", + "output_key": "task", + "subtask_probability": 0.98, + "drop_subtask_key": true, + "has_subtask_key": "__task_description_has_subtask__", + "used_subtask_key": "__task_description_used_subtask__", + "input_type": "Dict", + "post_norm": false + } + ] + }, + "language_instruction": null, + "training_mode": "ROBOT_FLOWMATCH", + "action_type": "EE_QUAT_POS_XYZW", + "serialize_stats": false, + "video_decoder_cache_size": null, + "repo_id": "xdof/yam_box_wild/open_the_pen_caps", + "streaming": false, + "tolerance_s": 0.1, + "use_imagenet_stats": false, + "split_name": "train", + "prefetch_factor": 2, + "max_episodes": null + }, + "weight": 0.056649822339 + } + ], + "split_name": null, + "shuffle_type": "FRAME", + "seed": 12345, + "num_workers": 4, + "prefetch_factor": 2, + "batch_size": 64, + "streaming": false, + "shuffle_buffer_size": 100, + "vl_batch_size": null, + "observation_whitelist": [ + "observation.state", + "action", + "observation.image.0", + "observation.image.1", + "observation.image.2", + "task", + "subtask", + "task_index", + "frame_index" + ], + "transform_mapping": null, + "chunk_size": null, + "flatten_nested": true, + "serialize_stats": true, + "video_decoder_cache_size": 128, + "training_mode_override": null + }, + "weight": 0.317225103375 + } + ], + "split_name": null, + "shuffle_type": "FRAME", + "seed": 12345, + "num_workers": 4, + "prefetch_factor": 2, + "batch_size": 64, + "streaming": false, + "shuffle_buffer_size": 100, + "vl_batch_size": null, + "observation_whitelist": [ + "observation.state", + "action", + "observation.image.0", + "observation.image.1", + "observation.image.2", + "task", + "subtask", + "task_index", + "frame_index" + ], + "transform_mapping": null, + "chunk_size": null, + "flatten_nested": true, + "serialize_stats": true, + "video_decoder_cache_size": null, + "training_mode_override": null + }, + "weight": 1.0 + } + ], + "split_name": null, + "shuffle_type": "FRAME", + "seed": 12345, + "num_workers": 4, + "prefetch_factor": 2, + "batch_size": 64, + "streaming": false, + "shuffle_buffer_size": 100, + "vl_batch_size": null, + "observation_whitelist": [ + "observation.state", + "action", + "observation.image.0", + "observation.image.1", + "observation.image.2", + "task", + "subtask", + "task_index", + "frame_index" + ], + "transform_mapping": null, + "chunk_size": null, + "flatten_nested": true, + "serialize_stats": true, + "video_decoder_cache_size": null, + "training_mode_override": null + }, + "weight": 0.753063449478 + }, + { + "dataset": { + "datasets": [ + { + "dataset": { + "batch_size": 16, + "num_workers": 8, + "shuffle_buffer_size": 100, + "normalization_mapping": { + "ACTION": "ACTIONCHUNK_QUANTILE", + "ENV": "QUANTILE", + "STATE": "QUANTILE", + "VISUAL": "IDENTITY" + }, + "features": { + "action": { + "type": "ACTION", + "shape": "(16,)" + }, + "observation.image.0": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.1": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.2": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.state": { + "type": "STATE", + "shape": "(16,)" + } + }, + "stats": null, + "clip_values": { + "ENV": [ + -1.05, + 1.05 + ], + "STATE": [ + -1.05, + 1.05 + ] + }, + "chunk_size": 50, + "observation_mapping": { + "action.ee_quat_pos": "action", + "observation.ee_quat_pos": "observation.state", + "observation.images.cam_high": "observation.image.0", + "observation.images.cam_left_wrist": "observation.image.1", + "observation.images.cam_right_wrist": "observation.image.2" + }, + "observation_whitelist": [ + "observation.state", + "observation.image.0", + "observation.image.1", + "observation.image.2", + "action", + "task", + "subtask", + "dataset_index", + "training_mode" + ], + "observation_blacklist": [ + "observation.images.cam_low" + ], + "transform_mapping": { + "action": [ + { + "type": "convert_to_6d_actions", + "action_key": "action", + "action_type": "EE_QUAT_POS_XYZW", + "input_type": "Dict", + "post_norm": false + }, + { + "type": "convert_to_6d_actions", + "action_key": "observation.state", + "action_type": "EE_QUAT_POS_XYZW", + "input_type": "Dict", + "post_norm": false + }, + { + "type": "delta_actions", + "state_key": "observation.state", + "action_key": "action", + "action_type": "EE_6D_POS", + "input_type": "Dict", + "post_norm": false, + "relative_to_state": true + } + ], + "observation.image.0": [ + { + "type": "random_resized_crop", + "height": 224, + "width": 224, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "observation.image.1": [ + { + "type": "random_resized_crop", + "height": 224, + "width": 224, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "observation.image.2": [ + { + "type": "random_resized_crop", + "height": 224, + "width": 224, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "task": [ + { + "type": "task_description_selector", + "task_key": "task", + "subtask_key": "subtask", + "output_key": "task", + "subtask_probability": 0.9, + "drop_subtask_key": true, + "has_subtask_key": "__task_description_has_subtask__", + "used_subtask_key": "__task_description_used_subtask__", + "input_type": "Dict", + "post_norm": false + } + ] + }, + "language_instruction": null, + "training_mode": "ROBOT_FLOWMATCH", + "action_type": "EE_QUAT_POS_XYZW", + "serialize_stats": false, + "video_decoder_cache_size": null, + "repo_id": "scale/trossen_stationary_ai_480x640_padded_merged", + "streaming": false, + "tolerance_s": 0.1, + "use_imagenet_stats": false, + "split_name": "train", + "prefetch_factor": 2, + "max_episodes": null + }, + "weight": 0.091591753466 + }, + { + "dataset": { + "batch_size": 16, + "num_workers": 8, + "shuffle_buffer_size": 100, + "normalization_mapping": { + "ACTION": "ACTIONCHUNK_QUANTILE", + "ENV": "QUANTILE", + "STATE": "QUANTILE", + "VISUAL": "IDENTITY" + }, + "features": { + "action": { + "type": "ACTION", + "shape": "(16,)" + }, + "observation.image.0": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.1": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.2": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.state": { + "type": "STATE", + "shape": "(16,)" + } + }, + "stats": null, + "clip_values": { + "ENV": [ + -1.05, + 1.05 + ], + "STATE": [ + -1.05, + 1.05 + ] + }, + "chunk_size": 50, + "observation_mapping": { + "action.ee_quat_pos": "action", + "observations.ee_quat_pos": "observation.state", + "observations.images.cam_left": "observation.image.1", + "observations.images.cam_right": "observation.image.2", + "observations.images.cam_top": "observation.image.0" + }, + "observation_whitelist": [ + "observation.state", + "observation.image.0", + "observation.image.1", + "observation.image.2", + "action", + "task", + "subtask", + "dataset_index", + "training_mode" + ], + "observation_blacklist": [], + "transform_mapping": { + "action": [ + { + "type": "convert_to_6d_actions", + "action_key": "action", + "action_type": "EE_QUAT_POS_XYZW", + "input_type": "Dict", + "post_norm": false + }, + { + "type": "convert_to_6d_actions", + "action_key": "observation.state", + "action_type": "EE_QUAT_POS_XYZW", + "input_type": "Dict", + "post_norm": false + }, + { + "type": "delta_actions", + "state_key": "observation.state", + "action_key": "action", + "action_type": "EE_6D_POS", + "input_type": "Dict", + "post_norm": false, + "relative_to_state": true + } + ], + "observation.image.0": [ + { + "type": "random_resized_crop", + "height": 224, + "width": 224, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "observation.image.1": [ + { + "type": "random_resized_crop", + "height": 224, + "width": 224, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "observation.image.2": [ + { + "type": "random_resized_crop", + "height": 224, + "width": 224, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "task": [ + { + "type": "task_description_selector", + "task_key": "task", + "subtask_key": "subtask", + "output_key": "task", + "subtask_probability": 0.9, + "drop_subtask_key": true, + "has_subtask_key": "__task_description_has_subtask__", + "used_subtask_key": "__task_description_used_subtask__", + "input_type": "Dict", + "post_norm": false + } + ] + }, + "language_instruction": null, + "training_mode": "ROBOT_FLOWMATCH", + "action_type": "EE_QUAT_POS_XYZW", + "serialize_stats": false, + "video_decoder_cache_size": null, + "repo_id": "scale/fr3_duo_merged", + "streaming": false, + "tolerance_s": 0.1, + "use_imagenet_stats": false, + "split_name": "train", + "prefetch_factor": 2, + "max_episodes": null + }, + "weight": 0.352192805883 + }, + { + "dataset": { + "datasets": [ + { + "dataset": { + "batch_size": 16, + "num_workers": 8, + "shuffle_buffer_size": 100, + "normalization_mapping": { + "ACTION": "ACTIONCHUNK_QUANTILE", + "ENV": "QUANTILE", + "STATE": "QUANTILE", + "VISUAL": "IDENTITY" + }, + "features": { + "action": { + "type": "ACTION", + "shape": "(16,)" + }, + "observation.image.0": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.1": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.2": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.state": { + "type": "STATE", + "shape": "(16,)" + } + }, + "stats": null, + "clip_values": { + "ENV": [ + -1.05, + 1.05 + ], + "STATE": [ + -1.05, + 1.05 + ] + }, + "chunk_size": 50, + "observation_mapping": { + "action.ee_quat_pos": "action", + "observations.ee_quat_pos": "observation.state", + "observations.images.cam_left": "observation.image.1", + "observations.images.cam_right": "observation.image.2", + "observations.images.cam_top": "observation.image.0" + }, + "observation_whitelist": [ + "observation.state", + "observation.image.0", + "observation.image.1", + "observation.image.2", + "action", + "task", + "subtask", + "dataset_index", + "training_mode" + ], + "observation_blacklist": [], + "transform_mapping": { + "action": [ + { + "type": "convert_to_6d_actions", + "action_key": "action", + "action_type": "EE_QUAT_POS_XYZW", + "input_type": "Dict", + "post_norm": false + }, + { + "type": "convert_to_6d_actions", + "action_key": "observation.state", + "action_type": "EE_QUAT_POS_XYZW", + "input_type": "Dict", + "post_norm": false + }, + { + "type": "delta_actions", + "state_key": "observation.state", + "action_key": "action", + "action_type": "EE_6D_POS", + "input_type": "Dict", + "post_norm": false, + "relative_to_state": true + } + ], + "observation.image.0": [ + { + "type": "random_resized_crop", + "height": 224, + "width": 224, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "observation.image.1": [ + { + "type": "random_resized_crop", + "height": 224, + "width": 224, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "observation.image.2": [ + { + "type": "random_resized_crop", + "height": 224, + "width": 224, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "task": [ + { + "type": "task_description_selector", + "task_key": "task", + "subtask_key": "subtask", + "output_key": "task", + "subtask_probability": 0.9, + "drop_subtask_key": true, + "has_subtask_key": "__task_description_has_subtask__", + "used_subtask_key": "__task_description_used_subtask__", + "input_type": "Dict", + "post_norm": false + } + ] + }, + "language_instruction": null, + "training_mode": "ROBOT_FLOWMATCH", + "action_type": "EE_QUAT_POS_XYZW", + "serialize_stats": false, + "video_decoder_cache_size": null, + "repo_id": "scale/ur_ai_trainer_merged/split_00", + "streaming": false, + "tolerance_s": 0.1, + "use_imagenet_stats": false, + "split_name": "train", + "prefetch_factor": 2, + "max_episodes": null + }, + "weight": 0.083479603011 + }, + { + "dataset": { + "batch_size": 16, + "num_workers": 8, + "shuffle_buffer_size": 100, + "normalization_mapping": { + "ACTION": "ACTIONCHUNK_QUANTILE", + "ENV": "QUANTILE", + "STATE": "QUANTILE", + "VISUAL": "IDENTITY" + }, + "features": { + "action": { + "type": "ACTION", + "shape": "(16,)" + }, + "observation.image.0": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.1": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.2": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.state": { + "type": "STATE", + "shape": "(16,)" + } + }, + "stats": null, + "clip_values": { + "ENV": [ + -1.05, + 1.05 + ], + "STATE": [ + -1.05, + 1.05 + ] + }, + "chunk_size": 50, + "observation_mapping": { + "action.ee_quat_pos": "action", + "observations.ee_quat_pos": "observation.state", + "observations.images.cam_left": "observation.image.1", + "observations.images.cam_right": "observation.image.2", + "observations.images.cam_top": "observation.image.0" + }, + "observation_whitelist": [ + "observation.state", + "observation.image.0", + "observation.image.1", + "observation.image.2", + "action", + "task", + "subtask", + "dataset_index", + "training_mode" + ], + "observation_blacklist": [], + "transform_mapping": { + "action": [ + { + "type": "convert_to_6d_actions", + "action_key": "action", + "action_type": "EE_QUAT_POS_XYZW", + "input_type": "Dict", + "post_norm": false + }, + { + "type": "convert_to_6d_actions", + "action_key": "observation.state", + "action_type": "EE_QUAT_POS_XYZW", + "input_type": "Dict", + "post_norm": false + }, + { + "type": "delta_actions", + "state_key": "observation.state", + "action_key": "action", + "action_type": "EE_6D_POS", + "input_type": "Dict", + "post_norm": false, + "relative_to_state": true + } + ], + "observation.image.0": [ + { + "type": "random_resized_crop", + "height": 224, + "width": 224, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "observation.image.1": [ + { + "type": "random_resized_crop", + "height": 224, + "width": 224, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "observation.image.2": [ + { + "type": "random_resized_crop", + "height": 224, + "width": 224, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "task": [ + { + "type": "task_description_selector", + "task_key": "task", + "subtask_key": "subtask", + "output_key": "task", + "subtask_probability": 0.9, + "drop_subtask_key": true, + "has_subtask_key": "__task_description_has_subtask__", + "used_subtask_key": "__task_description_used_subtask__", + "input_type": "Dict", + "post_norm": false + } + ] + }, + "language_instruction": null, + "training_mode": "ROBOT_FLOWMATCH", + "action_type": "EE_QUAT_POS_XYZW", + "serialize_stats": false, + "video_decoder_cache_size": null, + "repo_id": "scale/ur_ai_trainer_merged/split_01", + "streaming": false, + "tolerance_s": 0.1, + "use_imagenet_stats": false, + "split_name": "train", + "prefetch_factor": 2, + "max_episodes": null + }, + "weight": 0.08307712956 + }, + { + "dataset": { + "batch_size": 16, + "num_workers": 8, + "shuffle_buffer_size": 100, + "normalization_mapping": { + "ACTION": "ACTIONCHUNK_QUANTILE", + "ENV": "QUANTILE", + "STATE": "QUANTILE", + "VISUAL": "IDENTITY" + }, + "features": { + "action": { + "type": "ACTION", + "shape": "(16,)" + }, + "observation.image.0": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.1": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.2": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.state": { + "type": "STATE", + "shape": "(16,)" + } + }, + "stats": null, + "clip_values": { + "ENV": [ + -1.05, + 1.05 + ], + "STATE": [ + -1.05, + 1.05 + ] + }, + "chunk_size": 50, + "observation_mapping": { + "action.ee_quat_pos": "action", + "observations.ee_quat_pos": "observation.state", + "observations.images.cam_left": "observation.image.1", + "observations.images.cam_right": "observation.image.2", + "observations.images.cam_top": "observation.image.0" + }, + "observation_whitelist": [ + "observation.state", + "observation.image.0", + "observation.image.1", + "observation.image.2", + "action", + "task", + "subtask", + "dataset_index", + "training_mode" + ], + "observation_blacklist": [], + "transform_mapping": { + "action": [ + { + "type": "convert_to_6d_actions", + "action_key": "action", + "action_type": "EE_QUAT_POS_XYZW", + "input_type": "Dict", + "post_norm": false + }, + { + "type": "convert_to_6d_actions", + "action_key": "observation.state", + "action_type": "EE_QUAT_POS_XYZW", + "input_type": "Dict", + "post_norm": false + }, + { + "type": "delta_actions", + "state_key": "observation.state", + "action_key": "action", + "action_type": "EE_6D_POS", + "input_type": "Dict", + "post_norm": false, + "relative_to_state": true + } + ], + "observation.image.0": [ + { + "type": "random_resized_crop", + "height": 224, + "width": 224, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "observation.image.1": [ + { + "type": "random_resized_crop", + "height": 224, + "width": 224, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "observation.image.2": [ + { + "type": "random_resized_crop", + "height": 224, + "width": 224, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "task": [ + { + "type": "task_description_selector", + "task_key": "task", + "subtask_key": "subtask", + "output_key": "task", + "subtask_probability": 0.9, + "drop_subtask_key": true, + "has_subtask_key": "__task_description_has_subtask__", + "used_subtask_key": "__task_description_used_subtask__", + "input_type": "Dict", + "post_norm": false + } + ] + }, + "language_instruction": null, + "training_mode": "ROBOT_FLOWMATCH", + "action_type": "EE_QUAT_POS_XYZW", + "serialize_stats": false, + "video_decoder_cache_size": null, + "repo_id": "scale/ur_ai_trainer_merged/split_02", + "streaming": false, + "tolerance_s": 0.1, + "use_imagenet_stats": false, + "split_name": "train", + "prefetch_factor": 2, + "max_episodes": null + }, + "weight": 0.083667789967 + }, + { + "dataset": { + "batch_size": 16, + "num_workers": 8, + "shuffle_buffer_size": 100, + "normalization_mapping": { + "ACTION": "ACTIONCHUNK_QUANTILE", + "ENV": "QUANTILE", + "STATE": "QUANTILE", + "VISUAL": "IDENTITY" + }, + "features": { + "action": { + "type": "ACTION", + "shape": "(16,)" + }, + "observation.image.0": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.1": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.2": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.state": { + "type": "STATE", + "shape": "(16,)" + } + }, + "stats": null, + "clip_values": { + "ENV": [ + -1.05, + 1.05 + ], + "STATE": [ + -1.05, + 1.05 + ] + }, + "chunk_size": 50, + "observation_mapping": { + "action.ee_quat_pos": "action", + "observations.ee_quat_pos": "observation.state", + "observations.images.cam_left": "observation.image.1", + "observations.images.cam_right": "observation.image.2", + "observations.images.cam_top": "observation.image.0" + }, + "observation_whitelist": [ + "observation.state", + "observation.image.0", + "observation.image.1", + "observation.image.2", + "action", + "task", + "subtask", + "dataset_index", + "training_mode" + ], + "observation_blacklist": [], + "transform_mapping": { + "action": [ + { + "type": "convert_to_6d_actions", + "action_key": "action", + "action_type": "EE_QUAT_POS_XYZW", + "input_type": "Dict", + "post_norm": false + }, + { + "type": "convert_to_6d_actions", + "action_key": "observation.state", + "action_type": "EE_QUAT_POS_XYZW", + "input_type": "Dict", + "post_norm": false + }, + { + "type": "delta_actions", + "state_key": "observation.state", + "action_key": "action", + "action_type": "EE_6D_POS", + "input_type": "Dict", + "post_norm": false, + "relative_to_state": true + } + ], + "observation.image.0": [ + { + "type": "random_resized_crop", + "height": 224, + "width": 224, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "observation.image.1": [ + { + "type": "random_resized_crop", + "height": 224, + "width": 224, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "observation.image.2": [ + { + "type": "random_resized_crop", + "height": 224, + "width": 224, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "task": [ + { + "type": "task_description_selector", + "task_key": "task", + "subtask_key": "subtask", + "output_key": "task", + "subtask_probability": 0.9, + "drop_subtask_key": true, + "has_subtask_key": "__task_description_has_subtask__", + "used_subtask_key": "__task_description_used_subtask__", + "input_type": "Dict", + "post_norm": false + } + ] + }, + "language_instruction": null, + "training_mode": "ROBOT_FLOWMATCH", + "action_type": "EE_QUAT_POS_XYZW", + "serialize_stats": false, + "video_decoder_cache_size": null, + "repo_id": "scale/ur_ai_trainer_merged/split_03", + "streaming": false, + "tolerance_s": 0.1, + "use_imagenet_stats": false, + "split_name": "train", + "prefetch_factor": 2, + "max_episodes": null + }, + "weight": 0.08348986545 + }, + { + "dataset": { + "batch_size": 16, + "num_workers": 8, + "shuffle_buffer_size": 100, + "normalization_mapping": { + "ACTION": "ACTIONCHUNK_QUANTILE", + "ENV": "QUANTILE", + "STATE": "QUANTILE", + "VISUAL": "IDENTITY" + }, + "features": { + "action": { + "type": "ACTION", + "shape": "(16,)" + }, + "observation.image.0": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.1": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.2": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.state": { + "type": "STATE", + "shape": "(16,)" + } + }, + "stats": null, + "clip_values": { + "ENV": [ + -1.05, + 1.05 + ], + "STATE": [ + -1.05, + 1.05 + ] + }, + "chunk_size": 50, + "observation_mapping": { + "action.ee_quat_pos": "action", + "observations.ee_quat_pos": "observation.state", + "observations.images.cam_left": "observation.image.1", + "observations.images.cam_right": "observation.image.2", + "observations.images.cam_top": "observation.image.0" + }, + "observation_whitelist": [ + "observation.state", + "observation.image.0", + "observation.image.1", + "observation.image.2", + "action", + "task", + "subtask", + "dataset_index", + "training_mode" + ], + "observation_blacklist": [], + "transform_mapping": { + "action": [ + { + "type": "convert_to_6d_actions", + "action_key": "action", + "action_type": "EE_QUAT_POS_XYZW", + "input_type": "Dict", + "post_norm": false + }, + { + "type": "convert_to_6d_actions", + "action_key": "observation.state", + "action_type": "EE_QUAT_POS_XYZW", + "input_type": "Dict", + "post_norm": false + }, + { + "type": "delta_actions", + "state_key": "observation.state", + "action_key": "action", + "action_type": "EE_6D_POS", + "input_type": "Dict", + "post_norm": false, + "relative_to_state": true + } + ], + "observation.image.0": [ + { + "type": "random_resized_crop", + "height": 224, + "width": 224, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "observation.image.1": [ + { + "type": "random_resized_crop", + "height": 224, + "width": 224, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "observation.image.2": [ + { + "type": "random_resized_crop", + "height": 224, + "width": 224, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "task": [ + { + "type": "task_description_selector", + "task_key": "task", + "subtask_key": "subtask", + "output_key": "task", + "subtask_probability": 0.9, + "drop_subtask_key": true, + "has_subtask_key": "__task_description_has_subtask__", + "used_subtask_key": "__task_description_used_subtask__", + "input_type": "Dict", + "post_norm": false + } + ] + }, + "language_instruction": null, + "training_mode": "ROBOT_FLOWMATCH", + "action_type": "EE_QUAT_POS_XYZW", + "serialize_stats": false, + "video_decoder_cache_size": null, + "repo_id": "scale/ur_ai_trainer_merged/split_04", + "streaming": false, + "tolerance_s": 0.1, + "use_imagenet_stats": false, + "split_name": "train", + "prefetch_factor": 2, + "max_episodes": null + }, + "weight": 0.082925410161 + }, + { + "dataset": { + "batch_size": 16, + "num_workers": 8, + "shuffle_buffer_size": 100, + "normalization_mapping": { + "ACTION": "ACTIONCHUNK_QUANTILE", + "ENV": "QUANTILE", + "STATE": "QUANTILE", + "VISUAL": "IDENTITY" + }, + "features": { + "action": { + "type": "ACTION", + "shape": "(16,)" + }, + "observation.image.0": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.1": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.2": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.state": { + "type": "STATE", + "shape": "(16,)" + } + }, + "stats": null, + "clip_values": { + "ENV": [ + -1.05, + 1.05 + ], + "STATE": [ + -1.05, + 1.05 + ] + }, + "chunk_size": 50, + "observation_mapping": { + "action.ee_quat_pos": "action", + "observations.ee_quat_pos": "observation.state", + "observations.images.cam_left": "observation.image.1", + "observations.images.cam_right": "observation.image.2", + "observations.images.cam_top": "observation.image.0" + }, + "observation_whitelist": [ + "observation.state", + "observation.image.0", + "observation.image.1", + "observation.image.2", + "action", + "task", + "subtask", + "dataset_index", + "training_mode" + ], + "observation_blacklist": [], + "transform_mapping": { + "action": [ + { + "type": "convert_to_6d_actions", + "action_key": "action", + "action_type": "EE_QUAT_POS_XYZW", + "input_type": "Dict", + "post_norm": false + }, + { + "type": "convert_to_6d_actions", + "action_key": "observation.state", + "action_type": "EE_QUAT_POS_XYZW", + "input_type": "Dict", + "post_norm": false + }, + { + "type": "delta_actions", + "state_key": "observation.state", + "action_key": "action", + "action_type": "EE_6D_POS", + "input_type": "Dict", + "post_norm": false, + "relative_to_state": true + } + ], + "observation.image.0": [ + { + "type": "random_resized_crop", + "height": 224, + "width": 224, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "observation.image.1": [ + { + "type": "random_resized_crop", + "height": 224, + "width": 224, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "observation.image.2": [ + { + "type": "random_resized_crop", + "height": 224, + "width": 224, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "task": [ + { + "type": "task_description_selector", + "task_key": "task", + "subtask_key": "subtask", + "output_key": "task", + "subtask_probability": 0.9, + "drop_subtask_key": true, + "has_subtask_key": "__task_description_has_subtask__", + "used_subtask_key": "__task_description_used_subtask__", + "input_type": "Dict", + "post_norm": false + } + ] + }, + "language_instruction": null, + "training_mode": "ROBOT_FLOWMATCH", + "action_type": "EE_QUAT_POS_XYZW", + "serialize_stats": false, + "video_decoder_cache_size": null, + "repo_id": "scale/ur_ai_trainer_merged/split_05", + "streaming": false, + "tolerance_s": 0.1, + "use_imagenet_stats": false, + "split_name": "train", + "prefetch_factor": 2, + "max_episodes": null + }, + "weight": 0.08303696668 + }, + { + "dataset": { + "batch_size": 16, + "num_workers": 8, + "shuffle_buffer_size": 100, + "normalization_mapping": { + "ACTION": "ACTIONCHUNK_QUANTILE", + "ENV": "QUANTILE", + "STATE": "QUANTILE", + "VISUAL": "IDENTITY" + }, + "features": { + "action": { + "type": "ACTION", + "shape": "(16,)" + }, + "observation.image.0": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.1": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.2": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.state": { + "type": "STATE", + "shape": "(16,)" + } + }, + "stats": null, + "clip_values": { + "ENV": [ + -1.05, + 1.05 + ], + "STATE": [ + -1.05, + 1.05 + ] + }, + "chunk_size": 50, + "observation_mapping": { + "action.ee_quat_pos": "action", + "observations.ee_quat_pos": "observation.state", + "observations.images.cam_left": "observation.image.1", + "observations.images.cam_right": "observation.image.2", + "observations.images.cam_top": "observation.image.0" + }, + "observation_whitelist": [ + "observation.state", + "observation.image.0", + "observation.image.1", + "observation.image.2", + "action", + "task", + "subtask", + "dataset_index", + "training_mode" + ], + "observation_blacklist": [], + "transform_mapping": { + "action": [ + { + "type": "convert_to_6d_actions", + "action_key": "action", + "action_type": "EE_QUAT_POS_XYZW", + "input_type": "Dict", + "post_norm": false + }, + { + "type": "convert_to_6d_actions", + "action_key": "observation.state", + "action_type": "EE_QUAT_POS_XYZW", + "input_type": "Dict", + "post_norm": false + }, + { + "type": "delta_actions", + "state_key": "observation.state", + "action_key": "action", + "action_type": "EE_6D_POS", + "input_type": "Dict", + "post_norm": false, + "relative_to_state": true + } + ], + "observation.image.0": [ + { + "type": "random_resized_crop", + "height": 224, + "width": 224, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "observation.image.1": [ + { + "type": "random_resized_crop", + "height": 224, + "width": 224, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "observation.image.2": [ + { + "type": "random_resized_crop", + "height": 224, + "width": 224, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "task": [ + { + "type": "task_description_selector", + "task_key": "task", + "subtask_key": "subtask", + "output_key": "task", + "subtask_probability": 0.9, + "drop_subtask_key": true, + "has_subtask_key": "__task_description_has_subtask__", + "used_subtask_key": "__task_description_used_subtask__", + "input_type": "Dict", + "post_norm": false + } + ] + }, + "language_instruction": null, + "training_mode": "ROBOT_FLOWMATCH", + "action_type": "EE_QUAT_POS_XYZW", + "serialize_stats": false, + "video_decoder_cache_size": null, + "repo_id": "scale/ur_ai_trainer_merged/split_06", + "streaming": false, + "tolerance_s": 0.1, + "use_imagenet_stats": false, + "split_name": "train", + "prefetch_factor": 2, + "max_episodes": null + }, + "weight": 0.083709283164 + }, + { + "dataset": { + "batch_size": 16, + "num_workers": 8, + "shuffle_buffer_size": 100, + "normalization_mapping": { + "ACTION": "ACTIONCHUNK_QUANTILE", + "ENV": "QUANTILE", + "STATE": "QUANTILE", + "VISUAL": "IDENTITY" + }, + "features": { + "action": { + "type": "ACTION", + "shape": "(16,)" + }, + "observation.image.0": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.1": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.2": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.state": { + "type": "STATE", + "shape": "(16,)" + } + }, + "stats": null, + "clip_values": { + "ENV": [ + -1.05, + 1.05 + ], + "STATE": [ + -1.05, + 1.05 + ] + }, + "chunk_size": 50, + "observation_mapping": { + "action.ee_quat_pos": "action", + "observations.ee_quat_pos": "observation.state", + "observations.images.cam_left": "observation.image.1", + "observations.images.cam_right": "observation.image.2", + "observations.images.cam_top": "observation.image.0" + }, + "observation_whitelist": [ + "observation.state", + "observation.image.0", + "observation.image.1", + "observation.image.2", + "action", + "task", + "subtask", + "dataset_index", + "training_mode" + ], + "observation_blacklist": [], + "transform_mapping": { + "action": [ + { + "type": "convert_to_6d_actions", + "action_key": "action", + "action_type": "EE_QUAT_POS_XYZW", + "input_type": "Dict", + "post_norm": false + }, + { + "type": "convert_to_6d_actions", + "action_key": "observation.state", + "action_type": "EE_QUAT_POS_XYZW", + "input_type": "Dict", + "post_norm": false + }, + { + "type": "delta_actions", + "state_key": "observation.state", + "action_key": "action", + "action_type": "EE_6D_POS", + "input_type": "Dict", + "post_norm": false, + "relative_to_state": true + } + ], + "observation.image.0": [ + { + "type": "random_resized_crop", + "height": 224, + "width": 224, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "observation.image.1": [ + { + "type": "random_resized_crop", + "height": 224, + "width": 224, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "observation.image.2": [ + { + "type": "random_resized_crop", + "height": 224, + "width": 224, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "task": [ + { + "type": "task_description_selector", + "task_key": "task", + "subtask_key": "subtask", + "output_key": "task", + "subtask_probability": 0.9, + "drop_subtask_key": true, + "has_subtask_key": "__task_description_has_subtask__", + "used_subtask_key": "__task_description_used_subtask__", + "input_type": "Dict", + "post_norm": false + } + ] + }, + "language_instruction": null, + "training_mode": "ROBOT_FLOWMATCH", + "action_type": "EE_QUAT_POS_XYZW", + "serialize_stats": false, + "video_decoder_cache_size": null, + "repo_id": "scale/ur_ai_trainer_merged/split_07", + "streaming": false, + "tolerance_s": 0.1, + "use_imagenet_stats": false, + "split_name": "train", + "prefetch_factor": 2, + "max_episodes": null + }, + "weight": 0.084145774701 + }, + { + "dataset": { + "batch_size": 16, + "num_workers": 8, + "shuffle_buffer_size": 100, + "normalization_mapping": { + "ACTION": "ACTIONCHUNK_QUANTILE", + "ENV": "QUANTILE", + "STATE": "QUANTILE", + "VISUAL": "IDENTITY" + }, + "features": { + "action": { + "type": "ACTION", + "shape": "(16,)" + }, + "observation.image.0": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.1": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.2": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.state": { + "type": "STATE", + "shape": "(16,)" + } + }, + "stats": null, + "clip_values": { + "ENV": [ + -1.05, + 1.05 + ], + "STATE": [ + -1.05, + 1.05 + ] + }, + "chunk_size": 50, + "observation_mapping": { + "action.ee_quat_pos": "action", + "observations.ee_quat_pos": "observation.state", + "observations.images.cam_left": "observation.image.1", + "observations.images.cam_right": "observation.image.2", + "observations.images.cam_top": "observation.image.0" + }, + "observation_whitelist": [ + "observation.state", + "observation.image.0", + "observation.image.1", + "observation.image.2", + "action", + "task", + "subtask", + "dataset_index", + "training_mode" + ], + "observation_blacklist": [], + "transform_mapping": { + "action": [ + { + "type": "convert_to_6d_actions", + "action_key": "action", + "action_type": "EE_QUAT_POS_XYZW", + "input_type": "Dict", + "post_norm": false + }, + { + "type": "convert_to_6d_actions", + "action_key": "observation.state", + "action_type": "EE_QUAT_POS_XYZW", + "input_type": "Dict", + "post_norm": false + }, + { + "type": "delta_actions", + "state_key": "observation.state", + "action_key": "action", + "action_type": "EE_6D_POS", + "input_type": "Dict", + "post_norm": false, + "relative_to_state": true + } + ], + "observation.image.0": [ + { + "type": "random_resized_crop", + "height": 224, + "width": 224, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "observation.image.1": [ + { + "type": "random_resized_crop", + "height": 224, + "width": 224, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "observation.image.2": [ + { + "type": "random_resized_crop", + "height": 224, + "width": 224, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "task": [ + { + "type": "task_description_selector", + "task_key": "task", + "subtask_key": "subtask", + "output_key": "task", + "subtask_probability": 0.9, + "drop_subtask_key": true, + "has_subtask_key": "__task_description_has_subtask__", + "used_subtask_key": "__task_description_used_subtask__", + "input_type": "Dict", + "post_norm": false + } + ] + }, + "language_instruction": null, + "training_mode": "ROBOT_FLOWMATCH", + "action_type": "EE_QUAT_POS_XYZW", + "serialize_stats": false, + "video_decoder_cache_size": null, + "repo_id": "scale/ur_ai_trainer_merged/split_08", + "streaming": false, + "tolerance_s": 0.1, + "use_imagenet_stats": false, + "split_name": "train", + "prefetch_factor": 2, + "max_episodes": null + }, + "weight": 0.082744529386 + }, + { + "dataset": { + "batch_size": 16, + "num_workers": 8, + "shuffle_buffer_size": 100, + "normalization_mapping": { + "ACTION": "ACTIONCHUNK_QUANTILE", + "ENV": "QUANTILE", + "STATE": "QUANTILE", + "VISUAL": "IDENTITY" + }, + "features": { + "action": { + "type": "ACTION", + "shape": "(16,)" + }, + "observation.image.0": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.1": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.2": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.state": { + "type": "STATE", + "shape": "(16,)" + } + }, + "stats": null, + "clip_values": { + "ENV": [ + -1.05, + 1.05 + ], + "STATE": [ + -1.05, + 1.05 + ] + }, + "chunk_size": 50, + "observation_mapping": { + "action.ee_quat_pos": "action", + "observations.ee_quat_pos": "observation.state", + "observations.images.cam_left": "observation.image.1", + "observations.images.cam_right": "observation.image.2", + "observations.images.cam_top": "observation.image.0" + }, + "observation_whitelist": [ + "observation.state", + "observation.image.0", + "observation.image.1", + "observation.image.2", + "action", + "task", + "subtask", + "dataset_index", + "training_mode" + ], + "observation_blacklist": [], + "transform_mapping": { + "action": [ + { + "type": "convert_to_6d_actions", + "action_key": "action", + "action_type": "EE_QUAT_POS_XYZW", + "input_type": "Dict", + "post_norm": false + }, + { + "type": "convert_to_6d_actions", + "action_key": "observation.state", + "action_type": "EE_QUAT_POS_XYZW", + "input_type": "Dict", + "post_norm": false + }, + { + "type": "delta_actions", + "state_key": "observation.state", + "action_key": "action", + "action_type": "EE_6D_POS", + "input_type": "Dict", + "post_norm": false, + "relative_to_state": true + } + ], + "observation.image.0": [ + { + "type": "random_resized_crop", + "height": 224, + "width": 224, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "observation.image.1": [ + { + "type": "random_resized_crop", + "height": 224, + "width": 224, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "observation.image.2": [ + { + "type": "random_resized_crop", + "height": 224, + "width": 224, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "task": [ + { + "type": "task_description_selector", + "task_key": "task", + "subtask_key": "subtask", + "output_key": "task", + "subtask_probability": 0.9, + "drop_subtask_key": true, + "has_subtask_key": "__task_description_has_subtask__", + "used_subtask_key": "__task_description_used_subtask__", + "input_type": "Dict", + "post_norm": false + } + ] + }, + "language_instruction": null, + "training_mode": "ROBOT_FLOWMATCH", + "action_type": "EE_QUAT_POS_XYZW", + "serialize_stats": false, + "video_decoder_cache_size": null, + "repo_id": "scale/ur_ai_trainer_merged/split_09", + "streaming": false, + "tolerance_s": 0.1, + "use_imagenet_stats": false, + "split_name": "train", + "prefetch_factor": 2, + "max_episodes": null + }, + "weight": 0.08346716561 + }, + { + "dataset": { + "batch_size": 16, + "num_workers": 8, + "shuffle_buffer_size": 100, + "normalization_mapping": { + "ACTION": "ACTIONCHUNK_QUANTILE", + "ENV": "QUANTILE", + "STATE": "QUANTILE", + "VISUAL": "IDENTITY" + }, + "features": { + "action": { + "type": "ACTION", + "shape": "(16,)" + }, + "observation.image.0": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.1": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.2": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.state": { + "type": "STATE", + "shape": "(16,)" + } + }, + "stats": null, + "clip_values": { + "ENV": [ + -1.05, + 1.05 + ], + "STATE": [ + -1.05, + 1.05 + ] + }, + "chunk_size": 50, + "observation_mapping": { + "action.ee_quat_pos": "action", + "observations.ee_quat_pos": "observation.state", + "observations.images.cam_left": "observation.image.1", + "observations.images.cam_right": "observation.image.2", + "observations.images.cam_top": "observation.image.0" + }, + "observation_whitelist": [ + "observation.state", + "observation.image.0", + "observation.image.1", + "observation.image.2", + "action", + "task", + "subtask", + "dataset_index", + "training_mode" + ], + "observation_blacklist": [], + "transform_mapping": { + "action": [ + { + "type": "convert_to_6d_actions", + "action_key": "action", + "action_type": "EE_QUAT_POS_XYZW", + "input_type": "Dict", + "post_norm": false + }, + { + "type": "convert_to_6d_actions", + "action_key": "observation.state", + "action_type": "EE_QUAT_POS_XYZW", + "input_type": "Dict", + "post_norm": false + }, + { + "type": "delta_actions", + "state_key": "observation.state", + "action_key": "action", + "action_type": "EE_6D_POS", + "input_type": "Dict", + "post_norm": false, + "relative_to_state": true + } + ], + "observation.image.0": [ + { + "type": "random_resized_crop", + "height": 224, + "width": 224, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "observation.image.1": [ + { + "type": "random_resized_crop", + "height": 224, + "width": 224, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "observation.image.2": [ + { + "type": "random_resized_crop", + "height": 224, + "width": 224, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "task": [ + { + "type": "task_description_selector", + "task_key": "task", + "subtask_key": "subtask", + "output_key": "task", + "subtask_probability": 0.9, + "drop_subtask_key": true, + "has_subtask_key": "__task_description_has_subtask__", + "used_subtask_key": "__task_description_used_subtask__", + "input_type": "Dict", + "post_norm": false + } + ] + }, + "language_instruction": null, + "training_mode": "ROBOT_FLOWMATCH", + "action_type": "EE_QUAT_POS_XYZW", + "serialize_stats": false, + "video_decoder_cache_size": null, + "repo_id": "scale/ur_ai_trainer_merged/split_10", + "streaming": false, + "tolerance_s": 0.1, + "use_imagenet_stats": false, + "split_name": "train", + "prefetch_factor": 2, + "max_episodes": null + }, + "weight": 0.082810263901 + }, + { + "dataset": { + "batch_size": 16, + "num_workers": 8, + "shuffle_buffer_size": 100, + "normalization_mapping": { + "ACTION": "ACTIONCHUNK_QUANTILE", + "ENV": "QUANTILE", + "STATE": "QUANTILE", + "VISUAL": "IDENTITY" + }, + "features": { + "action": { + "type": "ACTION", + "shape": "(16,)" + }, + "observation.image.0": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.1": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.2": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.state": { + "type": "STATE", + "shape": "(16,)" + } + }, + "stats": null, + "clip_values": { + "ENV": [ + -1.05, + 1.05 + ], + "STATE": [ + -1.05, + 1.05 + ] + }, + "chunk_size": 50, + "observation_mapping": { + "action.ee_quat_pos": "action", + "observations.ee_quat_pos": "observation.state", + "observations.images.cam_left": "observation.image.1", + "observations.images.cam_right": "observation.image.2", + "observations.images.cam_top": "observation.image.0" + }, + "observation_whitelist": [ + "observation.state", + "observation.image.0", + "observation.image.1", + "observation.image.2", + "action", + "task", + "subtask", + "dataset_index", + "training_mode" + ], + "observation_blacklist": [], + "transform_mapping": { + "action": [ + { + "type": "convert_to_6d_actions", + "action_key": "action", + "action_type": "EE_QUAT_POS_XYZW", + "input_type": "Dict", + "post_norm": false + }, + { + "type": "convert_to_6d_actions", + "action_key": "observation.state", + "action_type": "EE_QUAT_POS_XYZW", + "input_type": "Dict", + "post_norm": false + }, + { + "type": "delta_actions", + "state_key": "observation.state", + "action_key": "action", + "action_type": "EE_6D_POS", + "input_type": "Dict", + "post_norm": false, + "relative_to_state": true + } + ], + "observation.image.0": [ + { + "type": "random_resized_crop", + "height": 224, + "width": 224, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "observation.image.1": [ + { + "type": "random_resized_crop", + "height": 224, + "width": 224, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "observation.image.2": [ + { + "type": "random_resized_crop", + "height": 224, + "width": 224, + "scale": [ + 0.9, + 0.9 + ], + "ratio": [ + 1.0, + 1.0 + ], + "input_type": "Tensor" + }, + { + "type": "color_jitter", + "brightness": 0.2, + "contrast": [ + 0.8, + 1.2 + ], + "saturation": [ + 0.8, + 1.2 + ], + "hue": 0.05, + "input_type": "Tensor" + } + ], + "task": [ + { + "type": "task_description_selector", + "task_key": "task", + "subtask_key": "subtask", + "output_key": "task", + "subtask_probability": 0.9, + "drop_subtask_key": true, + "has_subtask_key": "__task_description_has_subtask__", + "used_subtask_key": "__task_description_used_subtask__", + "input_type": "Dict", + "post_norm": false + } + ] + }, + "language_instruction": null, + "training_mode": "ROBOT_FLOWMATCH", + "action_type": "EE_QUAT_POS_XYZW", + "serialize_stats": false, + "video_decoder_cache_size": null, + "repo_id": "scale/ur_ai_trainer_merged/split_11", + "streaming": false, + "tolerance_s": 0.1, + "use_imagenet_stats": false, + "split_name": "train", + "prefetch_factor": 2, + "max_episodes": null + }, + "weight": 0.083446218408 + } + ], + "split_name": null, + "shuffle_type": "FRAME", + "seed": 12345, + "num_workers": 4, + "prefetch_factor": 2, + "batch_size": 64, + "streaming": false, + "shuffle_buffer_size": null, + "vl_batch_size": null, + "observation_whitelist": null, + "transform_mapping": null, + "chunk_size": null, + "flatten_nested": true, + "serialize_stats": true, + "video_decoder_cache_size": null, + "training_mode_override": null + }, + "weight": 0.556215440651 + } + ], + "split_name": null, + "shuffle_type": "FRAME", + "seed": 12345, + "num_workers": 4, + "prefetch_factor": 2, + "batch_size": 64, + "streaming": false, + "shuffle_buffer_size": 100, + "vl_batch_size": null, + "observation_whitelist": [ + "observation.state", + "action", + "observation.image.0", + "observation.image.1", + "observation.image.2", + "task", + "subtask", + "task_index", + "frame_index" + ], + "transform_mapping": null, + "chunk_size": null, + "flatten_nested": true, + "serialize_stats": true, + "video_decoder_cache_size": null, + "training_mode_override": null + }, + "weight": 0.17430815331 + } + ], + "split_name": null, + "shuffle_type": "FRAME", + "seed": 12345, + "num_workers": 8, + "prefetch_factor": 2, + "batch_size": 64, + "streaming": false, + "shuffle_buffer_size": null, + "vl_batch_size": null, + "observation_whitelist": [ + "observation.state", + "observation.image.0", + "observation.image.1", + "observation.image.2", + "action", + "task", + "subtask", + "task_index", + "frame_index" + ], + "transform_mapping": null, + "chunk_size": 50, + "flatten_nested": true, + "serialize_stats": false, + "video_decoder_cache_size": 128, + "training_mode_override": null + }, + "weight": 9.0 + }, + { + "dataset": { + "datasets": [ + { + "dataset": { + "batch_size": 64, + "num_workers": 8, + "shuffle_buffer_size": 1000, + "normalization_mapping": { + "observation.image": "IDENTITY", + "observation.environment_state": "MEAN_STD", + "observation.state": "MEAN_STD", + "action": "MEAN_STD" + }, + "features": { + "action": { + "type": "ACTION", + "shape": "(20,)" + }, + "observation.image.0": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.1": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.2": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.state": { + "type": "STATE", + "shape": "(20,)" + } + }, + "stats": null, + "clip_values": null, + "chunk_size": 50, + "observation_mapping": null, + "observation_whitelist": [ + "observation.state", + "observation.image.0", + "observation.image.1", + "observation.image.2", + "action", + "task", + "subtask", + "dataset_index", + "training_mode" + ], + "observation_blacklist": null, + "transform_mapping": null, + "language_instruction": null, + "training_mode": "VQA", + "action_type": "POSITION", + "serialize_stats": false, + "video_decoder_cache_size": null, + "manifest_path": "${COTRAINING_DATA_ROOT}/robopoint/robopoint_1432k.json", + "images_dir": "${COTRAINING_DATA_ROOT}/robopoint/images", + "image_size": null, + "max_images": 1, + "preserve_image_tokens": false, + "max_answer_bytes": 2048, + "_resolved_manifest": "/data/msrx-eai/datasets/cotraining/robopoint/robopoint_1432k.json", + "_resolved_images": "/data/msrx-eai/datasets/cotraining/robopoint/images", + "_cached_length": null + }, + "weight": 1.0 + }, + { + "dataset": { + "datasets": [ + { + "dataset": { + "batch_size": 64, + "num_workers": 8, + "shuffle_buffer_size": 1000, + "normalization_mapping": { + "observation.image": "IDENTITY", + "observation.environment_state": "MEAN_STD", + "observation.state": "MEAN_STD", + "action": "MEAN_STD" + }, + "features": { + "action": { + "type": "ACTION", + "shape": "(20,)" + }, + "observation.image.0": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.1": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.2": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.state": { + "type": "STATE", + "shape": "(20,)" + } + }, + "stats": null, + "clip_values": null, + "chunk_size": 50, + "observation_mapping": null, + "observation_whitelist": [ + "observation.state", + "observation.image.0", + "observation.image.1", + "observation.image.2", + "action", + "task", + "subtask", + "dataset_index", + "training_mode" + ], + "observation_blacklist": null, + "transform_mapping": null, + "language_instruction": null, + "training_mode": "VQA", + "action_type": "POSITION", + "serialize_stats": false, + "video_decoder_cache_size": null, + "manifest_path": "${COTRAINING_DATA_ROOT}/refspatial/2D/choice_qa.json", + "images_dir": "${COTRAINING_DATA_ROOT}/refspatial/2D/image", + "image_size": null, + "max_images": 1, + "preserve_image_tokens": false, + "max_answer_bytes": 2048, + "_resolved_manifest": "/data/msrx-eai/datasets/cotraining/refspatial/2D/choice_qa.json", + "_resolved_images": "/data/msrx-eai/datasets/cotraining/refspatial/2D/image", + "_cached_length": null + }, + "weight": 1.0 + }, + { + "dataset": { + "batch_size": 64, + "num_workers": 8, + "shuffle_buffer_size": 1000, + "normalization_mapping": { + "observation.image": "IDENTITY", + "observation.environment_state": "MEAN_STD", + "observation.state": "MEAN_STD", + "action": "MEAN_STD" + }, + "features": { + "action": { + "type": "ACTION", + "shape": "(20,)" + }, + "observation.image.0": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.1": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.2": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.state": { + "type": "STATE", + "shape": "(20,)" + } + }, + "stats": null, + "clip_values": null, + "chunk_size": 50, + "observation_mapping": null, + "observation_whitelist": [ + "observation.state", + "observation.image.0", + "observation.image.1", + "observation.image.2", + "action", + "task", + "subtask", + "dataset_index", + "training_mode" + ], + "observation_blacklist": null, + "transform_mapping": null, + "language_instruction": null, + "training_mode": "VQA", + "action_type": "POSITION", + "serialize_stats": false, + "video_decoder_cache_size": null, + "manifest_path": "${COTRAINING_DATA_ROOT}/refspatial/2D/reasoning_template_qa.json", + "images_dir": "${COTRAINING_DATA_ROOT}/refspatial/2D/image", + "image_size": null, + "max_images": 1, + "preserve_image_tokens": false, + "max_answer_bytes": 2048, + "_resolved_manifest": "/data/msrx-eai/datasets/cotraining/refspatial/2D/reasoning_template_qa.json", + "_resolved_images": "/data/msrx-eai/datasets/cotraining/refspatial/2D/image", + "_cached_length": null + }, + "weight": 1.0 + }, + { + "dataset": { + "batch_size": 64, + "num_workers": 8, + "shuffle_buffer_size": 1000, + "normalization_mapping": { + "observation.image": "IDENTITY", + "observation.environment_state": "MEAN_STD", + "observation.state": "MEAN_STD", + "action": "MEAN_STD" + }, + "features": { + "action": { + "type": "ACTION", + "shape": "(20,)" + }, + "observation.image.0": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.1": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.2": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.state": { + "type": "STATE", + "shape": "(20,)" + } + }, + "stats": null, + "clip_values": null, + "chunk_size": 50, + "observation_mapping": null, + "observation_whitelist": [ + "observation.state", + "observation.image.0", + "observation.image.1", + "observation.image.2", + "action", + "task", + "subtask", + "dataset_index", + "training_mode" + ], + "observation_blacklist": null, + "transform_mapping": null, + "language_instruction": null, + "training_mode": "VQA", + "action_type": "POSITION", + "serialize_stats": false, + "video_decoder_cache_size": null, + "manifest_path": "${COTRAINING_DATA_ROOT}/refspatial/3D/choice_qa.json", + "images_dir": "${COTRAINING_DATA_ROOT}/refspatial/3D/image", + "image_size": null, + "max_images": 1, + "preserve_image_tokens": false, + "max_answer_bytes": 2048, + "_resolved_manifest": "/data/msrx-eai/datasets/cotraining/refspatial/3D/choice_qa.json", + "_resolved_images": "/data/msrx-eai/datasets/cotraining/refspatial/3D/image", + "_cached_length": null + }, + "weight": 1.0 + }, + { + "dataset": { + "batch_size": 64, + "num_workers": 8, + "shuffle_buffer_size": 1000, + "normalization_mapping": { + "observation.image": "IDENTITY", + "observation.environment_state": "MEAN_STD", + "observation.state": "MEAN_STD", + "action": "MEAN_STD" + }, + "features": { + "action": { + "type": "ACTION", + "shape": "(20,)" + }, + "observation.image.0": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.1": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.2": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.state": { + "type": "STATE", + "shape": "(20,)" + } + }, + "stats": null, + "clip_values": null, + "chunk_size": 50, + "observation_mapping": null, + "observation_whitelist": [ + "observation.state", + "observation.image.0", + "observation.image.1", + "observation.image.2", + "action", + "task", + "subtask", + "dataset_index", + "training_mode" + ], + "observation_blacklist": null, + "transform_mapping": null, + "language_instruction": null, + "training_mode": "VQA", + "action_type": "POSITION", + "serialize_stats": false, + "video_decoder_cache_size": null, + "manifest_path": "${COTRAINING_DATA_ROOT}/refspatial/3D/reasoning_template_qa.json", + "images_dir": "${COTRAINING_DATA_ROOT}/refspatial/3D/image", + "image_size": null, + "max_images": 1, + "preserve_image_tokens": false, + "max_answer_bytes": 2048, + "_resolved_manifest": "/data/msrx-eai/datasets/cotraining/refspatial/3D/reasoning_template_qa.json", + "_resolved_images": "/data/msrx-eai/datasets/cotraining/refspatial/3D/image", + "_cached_length": null + }, + "weight": 1.0 + }, + { + "dataset": { + "batch_size": 64, + "num_workers": 8, + "shuffle_buffer_size": 1000, + "normalization_mapping": { + "observation.image": "IDENTITY", + "observation.environment_state": "MEAN_STD", + "observation.state": "MEAN_STD", + "action": "MEAN_STD" + }, + "features": { + "action": { + "type": "ACTION", + "shape": "(20,)" + }, + "observation.image.0": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.1": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.2": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.state": { + "type": "STATE", + "shape": "(20,)" + } + }, + "stats": null, + "clip_values": null, + "chunk_size": 50, + "observation_mapping": null, + "observation_whitelist": [ + "observation.state", + "observation.image.0", + "observation.image.1", + "observation.image.2", + "action", + "task", + "subtask", + "dataset_index", + "training_mode" + ], + "observation_blacklist": null, + "transform_mapping": null, + "language_instruction": null, + "training_mode": "POINTING", + "action_type": "POSITION", + "serialize_stats": false, + "video_decoder_cache_size": null, + "manifest_path": "${COTRAINING_DATA_ROOT}/refspatial/3D/vacant_qa.json", + "images_dir": "${COTRAINING_DATA_ROOT}/refspatial/3D/image", + "image_size": null, + "max_images": 1, + "preserve_image_tokens": false, + "max_answer_bytes": 2048, + "_resolved_manifest": "/data/msrx-eai/datasets/cotraining/refspatial/3D/vacant_qa.json", + "_resolved_images": "/data/msrx-eai/datasets/cotraining/refspatial/3D/image", + "_cached_length": null + }, + "weight": 1.0 + }, + { + "dataset": { + "batch_size": 64, + "num_workers": 8, + "shuffle_buffer_size": 1000, + "normalization_mapping": { + "observation.image": "IDENTITY", + "observation.environment_state": "MEAN_STD", + "observation.state": "MEAN_STD", + "action": "MEAN_STD" + }, + "features": { + "action": { + "type": "ACTION", + "shape": "(20,)" + }, + "observation.image.0": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.1": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.image.2": { + "type": "VISUAL", + "shape": "(3, 224, 224)" + }, + "observation.state": { + "type": "STATE", + "shape": "(20,)" + } + }, + "stats": null, + "clip_values": null, + "chunk_size": 50, + "observation_mapping": null, + "observation_whitelist": [ + "observation.state", + "observation.image.0", + "observation.image.1", + "observation.image.2", + "action", + "task", + "subtask", + "dataset_index", + "training_mode" + ], + "observation_blacklist": null, + "transform_mapping": null, + "language_instruction": null, + "training_mode": "VQA", + "action_type": "POSITION", + "serialize_stats": false, + "video_decoder_cache_size": null, + "manifest_path": "${COTRAINING_DATA_ROOT}/refspatial/3D/visual_choice_qa.json", + "images_dir": "${COTRAINING_DATA_ROOT}/refspatial/3D/image_visual_choice", + "image_size": null, + "max_images": 1, + "preserve_image_tokens": false, + "max_answer_bytes": 2048, + "_resolved_manifest": "/data/msrx-eai/datasets/cotraining/refspatial/3D/visual_choice_qa.json", + "_resolved_images": "/data/msrx-eai/datasets/cotraining/refspatial/3D/image_visual_choice", + "_cached_length": null + }, + "weight": 1.0 + } + ], + "split_name": null, + "shuffle_type": "FRAME", + "seed": 12345, + "num_workers": 4, + "prefetch_factor": 2, + "batch_size": 64, + "streaming": false, + "shuffle_buffer_size": null, + "vl_batch_size": null, + "observation_whitelist": null, + "transform_mapping": null, + "chunk_size": null, + "flatten_nested": true, + "serialize_stats": true, + "video_decoder_cache_size": null, + "training_mode_override": null + }, + "weight": 1.0 + } + ], + "split_name": null, + "shuffle_type": "FRAME", + "seed": 12345, + "num_workers": 4, + "prefetch_factor": 2, + "batch_size": 64, + "streaming": false, + "shuffle_buffer_size": null, + "vl_batch_size": null, + "observation_whitelist": null, + "transform_mapping": null, + "chunk_size": null, + "flatten_nested": true, + "serialize_stats": true, + "video_decoder_cache_size": null, + "training_mode_override": null + }, + "weight": 1.0 + } + ], + "split_name": null, + "shuffle_type": "FRAME", + "seed": 12345, + "num_workers": 8, + "prefetch_factor": 2, + "batch_size": 192, + "streaming": false, + "shuffle_buffer_size": null, + "vl_batch_size": 8, + "observation_whitelist": [ + "observation.state", + "observation.image.0", + "observation.image.1", + "observation.image.2", + "action", + "task", + "subtask", + "dataset_index", + "training_mode" + ], + "transform_mapping": null, + "chunk_size": 50, + "flatten_nested": true, + "serialize_stats": false, + "video_decoder_cache_size": 128, + "training_mode_override": null +}