{"dataset_kwargs": {"balance_weights": true, "batch_size": 64, "dataset_kwargs_list": [{"name": "fractal20220817_data", "data_dir": "gs://rail-datasets-europe-west4/oxe/resize_256_256", "image_obs_keys": {"primary": "image", "wrist": null}, "action_normalization_mask": [true, true, true, true, true, true, false], "language_key": "language_instruction", "standardize_fn": {"module": "octo.data.oxe.oxe_standardization_transforms", "name": "rt1_dataset_transform", "args": [], "kwargs": {}}}, {"name": "kuka", "data_dir": "gs://rail-datasets-europe-west4/oxe/resize_256_256", "image_obs_keys": {"primary": "image", "wrist": null}, "action_normalization_mask": [true, true, true, true, true, true, false], "language_key": "language_instruction", "standardize_fn": {"module": "octo.data.oxe.oxe_standardization_transforms", "name": "kuka_dataset_transform", "args": [], "kwargs": {}}}, {"name": "bridge_dataset", "data_dir": "gs://rail-datasets-europe-west4/oxe/resize_256_256", "image_obs_keys": {"primary": "image_0", "wrist": null}, "action_normalization_mask": [true, true, true, true, true, true, false], "language_key": "language_instruction", "standardize_fn": {"module": "octo.data.oxe.oxe_standardization_transforms", "name": "bridge_dataset_transform", "args": [], "kwargs": {}}}, {"name": "taco_play", "data_dir": "gs://rail-datasets-europe-west4/oxe/resize_256_256", "image_obs_keys": {"primary": "rgb_static", "wrist": "rgb_gripper"}, "action_normalization_mask": [true, true, true, true, true, true, false], "language_key": "language_instruction", "standardize_fn": {"module": "octo.data.oxe.oxe_standardization_transforms", "name": "taco_dataset_transform", "args": [], "kwargs": {}}}, {"name": "taco_extra", "data_dir": "gs://rail-datasets-europe-west4/oxe/resize_256_256", "image_obs_keys": {"primary": "rgb_static", "wrist": "rgb_gripper"}, "action_normalization_mask": [true, true, true, true, true, true, false], "language_key": "language_instruction", "standardize_fn": {"module": "octo.data.oxe.oxe_standardization_transforms", "name": "taco_dataset_transform", "args": [], "kwargs": {}}}, {"name": "jaco_play", "data_dir": "gs://rail-datasets-europe-west4/oxe/resize_256_256", "image_obs_keys": {"primary": "image", "wrist": "image_wrist"}, "action_normalization_mask": [true, true, true, true, true, true, false], "language_key": "language_instruction", "standardize_fn": {"module": "octo.data.oxe.oxe_standardization_transforms", "name": "jaco_play_dataset_transform", "args": [], "kwargs": {}}}, {"name": "berkeley_cable_routing", "data_dir": "gs://rail-datasets-europe-west4/oxe/resize_256_256", "image_obs_keys": {"primary": "image", "wrist": "wrist45_image"}, "action_normalization_mask": [true, true, true, true, true, true, false], "language_key": "language_instruction", "standardize_fn": {"module": "octo.data.oxe.oxe_standardization_transforms", "name": "berkeley_cable_routing_dataset_transform", "args": [], "kwargs": {}}}, {"name": "roboturk", "data_dir": "gs://rail-datasets-europe-west4/oxe/resize_256_256", "image_obs_keys": {"primary": "front_rgb", "wrist": null}, "action_normalization_mask": [true, true, true, true, true, true, false], "language_key": "language_instruction", "standardize_fn": {"module": "octo.data.oxe.oxe_standardization_transforms", "name": "roboturk_dataset_transform", "args": [], "kwargs": {}}}, {"name": "nyu_door_opening_surprising_effectiveness", "data_dir": "gs://rail-datasets-europe-west4/oxe/resize_256_256", "image_obs_keys": {"primary": null, "wrist": "image"}, "action_normalization_mask": [true, true, true, true, true, true, false], "language_key": "language_instruction", "standardize_fn": {"module": "octo.data.oxe.oxe_standardization_transforms", "name": "nyu_door_opening_dataset_transform", "args": [], "kwargs": {}}}, {"name": "viola", "data_dir": "gs://rail-datasets-europe-west4/oxe/resize_256_256", "image_obs_keys": {"primary": "agentview_rgb", "wrist": "eye_in_hand_rgb"}, "action_normalization_mask": [true, true, true, true, true, true, false], "language_key": "language_instruction", "standardize_fn": {"module": "octo.data.oxe.oxe_standardization_transforms", "name": "viola_dataset_transform", "args": [], "kwargs": {}}}, {"name": "berkeley_autolab_ur5", "data_dir": "gs://rail-datasets-europe-west4/oxe/resize_256_256", "image_obs_keys": {"primary": "image", "wrist": "hand_image"}, "action_normalization_mask": [true, true, true, true, true, true, false], "language_key": "language_instruction", "standardize_fn": {"module": "octo.data.oxe.oxe_standardization_transforms", "name": "berkeley_autolab_ur5_dataset_transform", "args": [], "kwargs": {}}}, {"name": "toto", "data_dir": "gs://rail-datasets-europe-west4/oxe/resize_256_256", "image_obs_keys": {"primary": "image", "wrist": null}, "action_normalization_mask": [true, true, true, true, true, true, false], "language_key": "language_instruction", "standardize_fn": {"module": "octo.data.oxe.oxe_standardization_transforms", "name": "toto_dataset_transform", "args": [], "kwargs": {}}}, {"name": "language_table", "data_dir": "gs://rail-datasets-europe-west4/oxe/resize_256_256", "image_obs_keys": {"primary": "rgb", "wrist": null}, "action_normalization_mask": [true, true, true, true, true, true, false], "language_key": "language_instruction", "standardize_fn": {"module": "octo.data.oxe.oxe_standardization_transforms", "name": "language_table_dataset_transform", "args": [], "kwargs": {}}}, {"name": "stanford_hydra_dataset_converted_externally_to_rlds", "data_dir": "gs://rail-datasets-europe-west4/oxe/resize_256_256", "image_obs_keys": {"primary": "image", "wrist": "wrist_image"}, "action_normalization_mask": [true, true, true, true, true, true, false], "language_key": "language_instruction", "standardize_fn": {"module": "octo.data.oxe.oxe_standardization_transforms", "name": "stanford_hydra_dataset_transform", "args": [], "kwargs": {}}}, {"name": "austin_buds_dataset_converted_externally_to_rlds", "data_dir": "gs://rail-datasets-europe-west4/oxe/resize_256_256", "image_obs_keys": {"primary": "image", "wrist": "wrist_image"}, "action_normalization_mask": [true, true, true, true, true, true, false], "language_key": "language_instruction", "standardize_fn": {"module": "octo.data.oxe.oxe_standardization_transforms", "name": "austin_buds_dataset_transform", "args": [], "kwargs": {}}}, {"name": "nyu_franka_play_dataset_converted_externally_to_rlds", "data_dir": "gs://rail-datasets-europe-west4/oxe/resize_256_256", "image_obs_keys": {"primary": "image", "wrist": null}, "action_normalization_mask": [true, true, true, true, true, true, false], "language_key": "language_instruction", "standardize_fn": {"module": "octo.data.oxe.oxe_standardization_transforms", "name": "nyu_franka_play_dataset_transform", "args": [], "kwargs": {}}}, {"name": "furniture_bench_dataset_converted_externally_to_rlds", "data_dir": "gs://rail-datasets-europe-west4/oxe/resize_256_256", "image_obs_keys": {"primary": "image", "wrist": "wrist_image"}, "action_normalization_mask": [true, true, true, true, true, true, false], "language_key": "language_instruction", "standardize_fn": {"module": "octo.data.oxe.oxe_standardization_transforms", "name": "furniture_bench_dataset_transform", "args": [], "kwargs": {}}}, {"name": "ucsd_kitchen_dataset_converted_externally_to_rlds", "data_dir": "gs://rail-datasets-europe-west4/oxe/resize_256_256", "image_obs_keys": {"primary": "image", "wrist": null}, "action_normalization_mask": [true, true, true, true, true, true, false], "language_key": "language_instruction", "standardize_fn": {"module": "octo.data.oxe.oxe_standardization_transforms", "name": "ucsd_kitchen_dataset_transform", "args": [], "kwargs": {}}}, {"name": "austin_sailor_dataset_converted_externally_to_rlds", "data_dir": "gs://rail-datasets-europe-west4/oxe/resize_256_256", "image_obs_keys": {"primary": "image", "wrist": "wrist_image"}, "action_normalization_mask": [true, true, true, true, true, true, false], "language_key": "language_instruction", "standardize_fn": {"module": "octo.data.oxe.oxe_standardization_transforms", "name": "austin_sailor_dataset_transform", "args": [], "kwargs": {}}}, {"name": "austin_sirius_dataset_converted_externally_to_rlds", "data_dir": "gs://rail-datasets-europe-west4/oxe/resize_256_256", "image_obs_keys": {"primary": "image", "wrist": "wrist_image"}, "action_normalization_mask": [true, true, true, true, true, true, false], "language_key": "language_instruction", "standardize_fn": {"module": "octo.data.oxe.oxe_standardization_transforms", "name": "austin_sirius_dataset_transform", "args": [], "kwargs": {}}}, {"name": "bc_z", "data_dir": "gs://rail-datasets-europe-west4/oxe/resize_256_256", "image_obs_keys": {"primary": "image", "wrist": null}, "action_normalization_mask": [true, true, true, true, true, true, false], "language_key": "language_instruction", "standardize_fn": {"module": "octo.data.oxe.oxe_standardization_transforms", "name": "bc_z_dataset_transform", "args": [], "kwargs": {}}}, {"name": "dlr_edan_shared_control_converted_externally_to_rlds", "data_dir": "gs://rail-datasets-europe-west4/oxe/resize_256_256", "image_obs_keys": {"primary": "image", "wrist": null}, "action_normalization_mask": [true, true, true, true, true, true, false], "language_key": "language_instruction", "standardize_fn": {"module": "octo.data.oxe.oxe_standardization_transforms", "name": "dlr_edan_shared_control_dataset_transform", "args": [], "kwargs": {}}}, {"name": "iamlab_cmu_pickup_insert_converted_externally_to_rlds", "data_dir": "gs://rail-datasets-europe-west4/oxe/resize_256_256", "image_obs_keys": {"primary": "image", "wrist": "wrist_image"}, "action_normalization_mask": [true, true, true, true, true, true, false], "language_key": "language_instruction", "standardize_fn": {"module": "octo.data.oxe.oxe_standardization_transforms", "name": "iamlab_pick_insert_dataset_transform", "args": [], "kwargs": {}}}, {"name": "utaustin_mutex", "data_dir": "gs://rail-datasets-europe-west4/oxe/resize_256_256", "image_obs_keys": {"primary": "image", "wrist": "wrist_image"}, "action_normalization_mask": [true, true, true, true, true, true, false], "language_key": "language_instruction", "standardize_fn": {"module": "octo.data.oxe.oxe_standardization_transforms", "name": "utaustin_mutex_dataset_transform", "args": [], "kwargs": {}}}, {"name": "berkeley_fanuc_manipulation", "data_dir": "gs://rail-datasets-europe-west4/oxe/resize_256_256", "image_obs_keys": {"primary": "image", "wrist": "wrist_image"}, "action_normalization_mask": [true, true, true, true, true, true, false], "language_key": "language_instruction", "standardize_fn": {"module": "octo.data.oxe.oxe_standardization_transforms", "name": "berkeley_fanuc_dataset_transform", "args": [], "kwargs": {}}}, {"name": "cmu_stretch", "data_dir": "gs://rail-datasets-europe-west4/oxe/resize_256_256", "image_obs_keys": {"primary": "image", "wrist": null}, "action_normalization_mask": [true, true, true, true, true, true, false], "language_key": "language_instruction", "standardize_fn": {"module": "octo.data.oxe.oxe_standardization_transforms", "name": "cmu_stretch_dataset_transform", "args": [], "kwargs": {}}}], "frame_transform_kwargs": {"image_augment_kwargs": {"primary": {"augment_order": ["random_resized_crop", "random_brightness", "random_contrast", "random_saturation", "random_hue"], "random_brightness": [0.1], "random_contrast": [0.9, 1.1], "random_hue": [0.05], "random_resized_crop": {"ratio": [0.9, 1.1], "scale": [0.8, 1.0]}, "random_saturation": [0.9, 1.1]}, "wrist": {"augment_order": ["random_brightness", "random_contrast", "random_saturation", "random_hue"], "random_brightness": [0.1], "random_contrast": [0.9, 1.1], "random_hue": [0.05], "random_saturation": [0.9, 1.1]}}, "image_dropout_prob": 0.0, "num_parallel_calls": 200, "resize_size": {"primary": [256, 256], "wrist": [128, 128]}}, "sample_weights": [0.54087122203, 0.8341046294, 1.0, 2.0, 2.0, 1.0, 1.0, 2.0, 1.0, 2.0, 2.0, 1.0, 0.1, 2.0, 1.0, 3.0, 0.1, 2.0, 1.0, 1.0, 0.2, 1.0, 1.0, 1.0, 2.0, 1.0], "shuffle_buffer_size": 250000, "traj_read_threads": 48, "traj_transform_kwargs": {"action_horizon": 4, "goal_relabeling_strategy": "uniform", "max_action_dim": 7, "subsample_length": 100, "task_augment_kwargs": {"keep_image_prob": 0.5, "pickle_file_path": "gs://rail-datasets-europe-west4/oxe/resize_256_256/paraphrases_oxe.pkl", "rephrase_prob": 0.5}, "task_augment_strategy": "delete_and_rephrase", "window_size": 2}, "traj_transform_threads": 48}, "eval_datasets": ["bridge_dataset"], "eval_interval": 5000, "log_interval": 100, "model": {"heads": {"action": {"module": "octo.model.components.action_heads", "name": "DiffusionActionHead", "args": [], "kwargs": {"action_horizon": 1, "action_dim": 7, "readout_key": "readout_action"}}}, "max_horizon": 10, "observation_tokenizers": {"primary": {"args": [], "kwargs": {"encoder": {"args": [], "kwargs": {}, "module": "octo.model.components.vit_encoders", "name": "SmallStem16"}, "obs_stack_keys": ["image_primary"], "task_stack_keys": ["image_primary"]}, "module": "octo.model.components.tokenizers", "name": "ImageTokenizer"}}, "readouts": {"action": 1}, "repeat_task_tokens": true, "task_tokenizers": {"language": {"args": [], "kwargs": {"encoder": "t5-base", "finetune_encoder": false}, "module": "octo.model.components.tokenizers", "name": "LanguageTokenizer"}}, "token_embedding_size": 768, "transformer_kwargs": {"add_position_embedding": false, "attention_dropout_rate": 0.0, "dropout_rate": 0.0, "mlp_dim": 3072, "num_attention_heads": 12, "num_layers": 12}, "use_correct_attention": true}, "num_steps": 300000.0, "optimizer": {"clip_gradient": 1.0, "frozen_keys": ["*hf_model*"], "learning_rate": {"init_value": 0.0, "name": "rsqrt", "peak_value": 0.0003, "timescale": 10000, "warmup_steps": 2000}, "weight_decay": 0.1}, "prefetch_num_batches": 0, "pretrained_loaders": [{"module": "octo.utils.train_utils", "name": "hf_weights_loader", "args": [], "kwargs": {"hf_model": "t5-base"}}], "resume_path": null, "save_dir": "gs://rail-tpus-homer-v5/log", "save_interval": 10000, "seed": 42, "start_step": null, "text_processor": {"args": [], "kwargs": {"encode_with_model": false, "tokenizer_kwargs": {"max_length": 16, "padding": "max_length", "return_tensors": "np", "truncation": true}, "tokenizer_name": "t5-base"}, "module": "octo.data.utils.text_processing", "name": "HFTokenizer"}, "val_kwargs": {"num_val_batches": 16, "val_shuffle_buffer_size": 1000}, "viz_interval": 1000000, "viz_kwargs": {"eval_batch_size": 128, "samples_per_state": 8, "trajs_for_metrics": 100, "trajs_for_viz": 8}, "wandb": {"entity": null, "group": null, "project": "octo"}, "wandb_resume_id": null, "window_size": 2}