Upload folder using huggingface_hub

.gitattributes

@@ -33,3 +33,7 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zip filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
+19999/default/checkpoint filter=lfs diff=lfs merge=lfs -text
+21999/default/checkpoint filter=lfs diff=lfs merge=lfs -text
+999/default/checkpoint filter=lfs diff=lfs merge=lfs -text
+9999/default/checkpoint filter=lfs diff=lfs merge=lfs -text

19999/default/checkpoint

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:70f09cbeba9749136b313e72a53baa0efa43b19367472ebd2ff60e9426934b5f
+size 540692595

21999/default/checkpoint

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bc2491250c1fc7bc49d5290f44d7b0914a34693557739c833074d76499376d8a
+size 540692595

999/default/checkpoint

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8529d02fcfc30e9b95794d5310cab87b5fe53138e17af9e8cd101d748504981d
+size 540692595

9999/default/checkpoint

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8fed8908c5d5d19bb1bef2111e1c880e66c6885d5c80a61bb0ce6fce5ca5ab94
+size 540692595

config.json

@@ -0,0 +1 @@
+{"seed": 42, "num_steps": 300000, "save_dir": null, "model": {"observation_tokenizers": {"primary": {"module": "octo.model.components.tokenizers", "name": "ImageTokenizer", "args": [], "kwargs": {"obs_stack_keys": ["image_primary"], "task_stack_keys": ["image_primary"], "encoder": {"module": "octo.model.components.vit_encoders", "name": "SmallStem16", "args": [], "kwargs": {}}}}}, "task_tokenizers": {"language": {"module": "octo.model.components.tokenizers", "name": "LanguageTokenizer", "args": [], "kwargs": {"encoder": "t5-base", "finetune_encoder": false}}}, "heads": {"action": {"module": "octo.model.components.action_heads", "name": "DiffusionActionHead", "args": [], "kwargs": {"readout_key": "readout_action", "use_map": false, "action_dim": 7, "action_horizon": 4}}}, "readouts": {"action": 1}, "token_embedding_size": 384, "transformer_kwargs": {"attention_dropout_rate": 0.0, "add_position_embedding": false, "num_layers": 12, "mlp_dim": 1536, "num_attention_heads": 6, "dropout_rate": 0.0}, "max_horizon": 10}, "window_size": 2, "dataset_kwargs": {"oxe_kwargs": {"data_mix": "oxe_magic_soup", "data_dir": "gs://rail-octo-central2/resize_256_256", "load_camera_views": ["primary", "wrist"], "load_depth": false}, "traj_transform_kwargs": {"window_size": 2, "future_action_window_size": 3, "goal_relabeling_strategy": "uniform", "subsample_length": 100, "task_augment_strategy": "delete_task_conditioning", "task_augment_kwargs": {"keep_image_prob": 0.5}}, "frame_transform_kwargs": {"num_parallel_calls": 200, "resize_size": {"primary": [256, 256], "wrist": [128, 128]}, "image_augment_kwargs": [{"random_resized_crop": {"scale": [0.8, 1.0], "ratio": [0.9, 1.1]}, "random_brightness": [0.1], "random_contrast": [0.9, 1.1], "random_saturation": [0.9, 1.1], "random_hue": [0.05], "augment_order": ["random_resized_crop", "random_brightness", "random_contrast", "random_saturation", "random_hue"]}, {"random_brightness": [0.1], "random_contrast": [0.9, 1.1], "random_saturation": [0.9, 1.1], "random_hue": [0.05], "augment_order": ["random_brightness", "random_contrast", "random_saturation", "random_hue"]}]}, "traj_transform_threads": 48, "traj_read_threads": 48, "shuffle_buffer_size": 500000, "batch_size": 128, "balance_weights": true}, "optimizer": {"learning_rate": {"name": "rsqrt", "init_value": 0.0, "peak_value": 0.0003, "warmup_steps": 2000, "timescale": 10000}, "weight_decay": 0.1, "clip_gradient": 1.0, "frozen_keys": ["*hf_model*"]}, "prefetch_num_batches": 0, "start_step": null, "log_interval": 100, "eval_interval": 5000, "viz_interval": 20000, "save_interval": 10000, "val_kwargs": {"val_shuffle_buffer_size": 1000, "num_val_batches": 16}, "viz_kwargs": {"eval_batch_size": 128, "trajs_for_metrics": 100, "trajs_for_viz": 8, "samples_per_state": 8}, "resume_path": null, "text_processor": {"module": "octo.data.utils.text_processing", "name": "HFTokenizer", "args": [], "kwargs": {"encode_with_model": false, "tokenizer_kwargs": {"max_length": 16, "padding": "max_length", "truncation": true, "return_tensors": "np"}, "tokenizer_name": "t5-base"}}, "pretrained_loaders": [{"module": "octo.utils.train_utils", "name": "hf_weights_loader", "args": [], "kwargs": {"hf_model": "t5-base"}}], "wandb": {"project": "octo", "group": null, "entity": null}, "wandb_resume_id": null, "eval_datasets": ["bridge_dataset"]}

dataset_statistics.json

@@ -0,0 +1 @@
+{"action": {"max": [0.12335848808288574, 0.14755941927433014, 0.12199220061302185, 3.1415927410125732, 1.5704798698425293, 3.141592502593994, 1.0], "mean": [-0.002537647495046258, 0.000192321022041142, -0.0007950603612698615, -0.08267237991094589, -0.11953054368495941, -0.08652400970458984, 0.6203598976135254], "min": [-0.1378057897090912, -0.14709188044071198, -0.1626778244972229, -3.1415927410125732, -1.5705456733703613, -3.141592264175415, 0.0], "p01": [-0.0655677542090416, -0.054904356598854065, -0.06744533777236938, -3.1395938396453857, -1.292111873626709, -3.07131028175354, 0.0], "p99": [0.06519947946071625, 0.0555696003139019, 0.053435586392879486, 3.1396241188049316, 0.9427980780601501, 3.0678741931915283, 1.0], "std": [0.02172495797276497, 0.017638076096773148, 0.020820973441004753, 3.0158143043518066, 0.4210500121116638, 1.4313457012176514, 0.45515739917755127]}, "num_trajectories": 92233, "num_transitions": 27044326}

example_batch.msgpack

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:618d81b34681d6b0bf9aabaa7adaf1ad8189322e078f18de0445b90d4fcfee19
+size 590772