init

150000/commit_success.txt

@@ -0,0 +1 @@
+Checkpoint commit was successful to gs://bridge_nils/checkpoints/octo/bridge_pnp_no_drawer_no_v1_ft_encoder_20240811_013251/150000

150000/default/checkpoint

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f42c436d1b0e90adf2aea233c5c2a8c183abff672ab5ad1cf0415cafffc80a4b
+size 540704883

150000/default/commit_success.txt

@@ -0,0 +1 @@
+Checkpoint commit was successful to gs://bridge_nils/checkpoints/octo/bridge_pnp_no_drawer_no_v1_ft_encoder_20240811_013251/150000/default

200000/commit_success.txt

@@ -0,0 +1 @@
+Checkpoint commit was successful to gs://bridge_nils/checkpoints/octo/bridge_pnp_no_drawer_no_v1_ft_encoder_20240811_013251/200000

200000/default/checkpoint

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:57c74adf504aea03bddad0d6a78edba175fec64db280bfa27c8e6521da4e2c51
+size 540704883

200000/default/commit_success.txt

@@ -0,0 +1 @@
+Checkpoint commit was successful to gs://bridge_nils/checkpoints/octo/bridge_pnp_no_drawer_no_v1_ft_encoder_20240811_013251/200000/default

250000/commit_success.txt

@@ -0,0 +1 @@
+Checkpoint commit was successful to gs://bridge_nils/checkpoints/octo/bridge_pnp_no_drawer_no_v1_ft_encoder_20240811_013251/250000

250000/default/checkpoint

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2b3843e9cfe46105ae5fdd2a44b63e9f9ae6af9e4a50afaa27b47d0f6027dc1e
+size 540704883

250000/default/commit_success.txt

@@ -0,0 +1 @@
+Checkpoint commit was successful to gs://bridge_nils/checkpoints/octo/bridge_pnp_no_drawer_no_v1_ft_encoder_20240811_013251/250000/default

config.json

@@ -0,0 +1 @@
+{"dataset_kwargs": {"balance_weights": true, "batch_size": 128, "dataset_kwargs_list": [{"name": "bridge_dataset_pnp_no_drawer_machine_fold_NILS", "data_dir": "gs://bridge_nils_new", "image_obs_keys": {"primary": "image_0"}, "language_key": "language_instruction", "language_key_NILS": "language_instruction_NILS", "Gt_ann_dirs": [], "NILS_ann_dirs": ["put", "pick", "take", "pnp", "icra", "rss", "many_skills", "lift_bowl", "move_drying", "wipe", "right_pepper", "topple", "upright", "flip"], "filter_functions": [{"module": "octo.data.utils.data_utils", "name": "filter_by_task_and_language_key", "args": [], "kwargs": {"language_key_gt_template": "language_instruction", "language_key_NILS_template": "language_instruction_NILS_0", "gt_task_templates": [], "NILS_task_templates": ["put", "pick", "take", "pnp", "icra", "rss", "many_skills", "lift_bowl", "move_drying", "wipe", "right_pepper", "topple", "upright", "flip"], "negative_task_templates": ["drawer", "machine", "fold", "bridgev1"]}}], "action_normalization_mask": [true, true, true, true, true, true, false], "action_proprio_normalization_type": "normal", "standardize_fn": {"module": "octo.data.oxe.oxe_standardization_transforms", "name": "bridge_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": [1.0], "shuffle_buffer_size": 50000, "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}, "task_augment_strategy": "delete_task_conditioning", "window_size": 2}, "traj_transform_threads": 48}, "eval_datasets": null, "eval_interval": 5000, "log_interval": 100, "model": {"heads": {"action": {"args": [], "kwargs": {"action_dim": 7, "action_horizon": 4, "dropout_rate": 0.0, "n_diffusion_samples": 1, "readout_key": "readout_action", "use_map": false}, "module": "octo.model.components.action_heads", "name": "DiffusionActionHead"}}, "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"}, "wrist": {"args": [], "kwargs": {"encoder": {"args": [], "kwargs": {}, "module": "octo.model.components.vit_encoders", "name": "SmallStem16"}, "obs_stack_keys": ["image_wrist"], "task_stack_keys": ["image_wrist"]}, "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": 384, "transformer_kwargs": {"add_position_embedding": false, "attention_dropout_rate": 0.0, "dropout_rate": 0.0, "mlp_dim": 1536, "num_attention_heads": 6, "num_layers": 12}, "use_correct_attention": true}, "num_steps": 250000.0, "optimizer": {"clip_gradient": 1.0, "frozen_keys": ["*.hf_model.shared.embedding", "*.hf_model.encoder.block.0.*", "*.hf_model.encoder.block.1.*", "*.hf_model.encoder.block.2.*", "*.hf_model.encoder.block.3.*", "*.hf_model.encoder.block.4.*", "*.hf_model.encoder.block.5.*", "*.hf_model.encoder.block.6.*", "*.hf_model.encoder.block.7.*", "*.hf_model.encoder.block.8.*", "*.hf_model.encoder.block.9.*"], "grad_accumulation_steps": null, "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://bridge_nils/checkpoints", "save_interval": 50000, "seed": 42, "start_step": null, "text_processor": {"args": [], "kwargs": {"encode_with_model": false, "tokenizer_kwargs": {"max_length": 24, "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": 20000, "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}

dataset_statistics.json

@@ -0,0 +1 @@
+{"bridge_dataset_pnp_no_drawer_machine_fold_NILS": {"action": {"mask": [true, true, true, true, true, true, false], "max": [0.41691166162490845, 0.11164695024490356, 0.21218234300613403, 3.122201919555664, 1.8618112802505493, 6.280478477478027, 1.0], "mean": [8.211634849430993e-05, 6.0986516473349184e-05, -2.3799542759661563e-05, 5.932206477154978e-06, -1.856556446000468e-05, 0.0001814452843973413, 0.6571118831634521], "min": [-0.3887568712234497, -0.11164695024490356, -0.22553899884223938, -3.1120171546936035, -1.8618112802505493, -6.279050827026367, 0.0], "p01": [-0.031007032841444016, -0.04428255930542946, -0.025063129141926765, -0.07583944499492645, -0.10015244781970978, -0.16370384395122528, 0.0], "p99": [0.029562290757894516, 0.04260377213358879, 0.04137621447443962, 0.07635962963104248, 0.07946572452783585, 0.15993088483810425, 1.0], "std": [0.010372006334364414, 0.014462774619460106, 0.012797058559954166, 0.028114112094044685, 0.03287765011191368, 0.07774192094802856, 0.4763176739215851]}, "num_trajectories": 22042, "num_transitions": 805913}}

example_batch.msgpack

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3a4bed6e044fc034f0065726f6a07d33953d88b7c17cfcf4a6894984593d28e6
+size 590894