Upload checkpoint-50000

checkpoint-50000/config.json

@@ -0,0 +1,187 @@
+{
+  "batch_size": 32,
+  "buffer_size": 64,
+  "eval_mix": "qiwu_test",
+  "frozen_keys": [
+    "*hf_model*"
+  ],
+  "lora": false,
+  "lora_config": {
+    "bias": "none",
+    "lora_alpha": 16,
+    "lora_dropout": 0.05,
+    "r": 8
+  },
+  "model": {
+    "heads": {
+      "action": {
+        "args": [],
+        "kwargs": {
+          "action_dim": 7,
+          "action_horizon": 4,
+          "dropout_rate": 0.0,
+          "n_diffusion_samples": 1,
+          "readout_key": "readout_action",
+          "token_embedding_size": 768,
+          "use_map": false
+        },
+        "module": "octo.model.components.action_heads",
+        "name": "DiffusionActionHead"
+      }
+    },
+    "max_horizon": 10,
+    "observation_tokenizers": {
+      "primary": {
+        "args": [],
+        "kwargs": {
+          "encoder": {
+            "args": [],
+            "kwargs": {
+              "in_features": 6
+            },
+            "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"
+      },
+      "state": {
+        "args": [],
+        "kwargs": {
+          "bin_type": "normal",
+          "high": 2.0,
+          "low": -2.0,
+          "n_bins": 256,
+          "obs_keys": [
+            "state"
+          ]
+        },
+        "module": "model.octo.components.tokenizers",
+        "name": "LowdimObsTokenizer"
+      },
+      "wrist": {
+        "args": [],
+        "kwargs": {
+          "encoder": {
+            "args": [],
+            "kwargs": {
+              "in_features": 6
+            },
+            "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": 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
+  },
+  "no_state": false,
+  "obs_token_nums": {
+    "primary": 256,
+    "state": 7,
+    "wrist": 64
+  },
+  "output_dir": "ljp_qiwu_banana_4",
+  "pretrained": {
+    "kwargs": {
+      "subpath": "oxe-g2-checkpoint-300000"
+    },
+    "model": "hf://chuanmew/octo_torch"
+  },
+  "run_name": "qiwu_banana_4",
+  "sampler_num_samples": 2560000,
+  "seed": 42,
+  "subsample_length": 99999,
+  "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.components.text_processing",
+    "name": "HFTokenizer"
+  },
+  "train_mix": "ljp_qiwu_banana",
+  "training_arguments": {
+    "bf16": true,
+    "dataloader_pin_memory": true,
+    "ddp_find_unused_parameters": true,
+    "eval_steps": 50000,
+    "gradient_accumulation_steps": 1,
+    "learning_rate": 0.0001,
+    "logging_nan_inf_filter": false,
+    "logging_steps": 10000,
+    "lr_scheduler_kwargs": {
+      "decay_type": "cosine",
+      "min_lr_ratio": 0.05,
+      "num_decay_steps": 47500
+    },
+    "lr_scheduler_type": "warmup_stable_decay",
+    "max_grad_norm": 20,
+    "max_steps": 50000,
+    "optim": "adamw_torch_fused",
+    "per_device_eval_batch_size": 1,
+    "per_device_train_batch_size": 1,
+    "report_to": "wandb",
+    "save_safetensors": true,
+    "save_steps": 50000,
+    "torch_compile": true,
+    "warmup_steps": 2500,
+    "weight_decay": 0.1
+  },
+  "training_keys": [
+    "*action*",
+    "*observation_tokenizers*",
+    "*transformer.transformer*",
+    "*task_projections*",
+    "*obs_projections*",
+    "*readout_embeddings*",
+    "*task_pos_embeddings*",
+    "*obs_pos_embeddings*"
+  ]
+}

checkpoint-50000/dataset_statistics.json

@@ -0,0 +1,98 @@
+[
+    {
+        "action": {
+            "mean": [
+                0.0035762591287493706,
+                0.0038505252450704575,
+                0.0005051779444329441,
+                0.004462657496333122,
+                0.00569327874109149,
+                0.00843613687902689,
+                0.5597720146179199
+            ],
+            "std": [
+                0.030762936919927597,
+                0.022843312472105026,
+                0.037356432527303696,
+                0.07383003830909729,
+                0.07964984327554703,
+                0.10997529327869415,
+                0.45039603114128113
+            ],
+            "min": [
+                -0.12550033628940582,
+                -0.08252930641174316,
+                -0.08629363775253296,
+                -0.5667345523834229,
+                -0.5142402648925781,
+                -1.2305364608764648,
+                0.0
+            ],
+            "max": [
+                0.15785643458366394,
+                0.1229468286037445,
+                0.15745647251605988,
+                0.9040577411651611,
+                0.43572473526000977,
+                1.0832273960113525,
+                1.0
+            ],
+            "mask": [
+                true,
+                true,
+                true,
+                true,
+                true,
+                true,
+                false
+            ]
+        },
+        "state": {
+            "mean": [
+                -1.147829294204712,
+                -1.018812894821167,
+                -1.342349648475647,
+                -0.603922963142395,
+                -0.9552030563354492,
+                0.9721980690956116,
+                0.5594711899757385
+            ],
+            "std": [
+                0.2672520875930786,
+                0.21161462366580963,
+                0.19813641905784607,
+                0.20742598176002502,
+                0.2395419031381607,
+                0.8849623203277588,
+                0.45082923769950867
+            ],
+            "min": [
+                -1.749814748764038,
+                -1.5758228302001953,
+                -1.8777823448181152,
+                -1.1325790882110596,
+                -1.702690839767456,
+                -0.8983733057975769,
+                0.0
+            ],
+            "max": [
+                -0.5615771412849426,
+                -0.41973423957824707,
+                -0.6838898062705994,
+                0.18788468837738037,
+                -0.38179078698158264,
+                3.117105722427368,
+                1.0
+            ],
+            "mask": [
+                true,
+                true,
+                true,
+                true,
+                true,
+                true,
+                false
+            ]
+        }
+    }
+]

checkpoint-50000/model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:591d0e746a6743eb56fe8058da598ad85c535d6fd5e89abb8c393429c0c223c1
+size 405079824

checkpoint-50000/optimizer.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0f67351955e514f45c5287ad47d595585f54b901494501ce48226b556f1c1529
+size 371730170

checkpoint-50000/rng_state.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f036448e1c518cbfba99f8dc6073173a8945e5e544e3be8b29689efe55c5d33c
+size 14244

checkpoint-50000/scheduler.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4c41ebb9a2c5635768a739e9e1b728bd4c30879d068859a65ed1c92f2838a08b
+size 1064

checkpoint-50000/trainer_state.json

@@ -0,0 +1,77 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 0.625,
+  "eval_steps": 50000,
+  "global_step": 50000,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.125,
+      "grad_norm": 15.0625,
+      "learning_rate": 9.427649832377157e-05,
+      "loss": 0.5293,
+      "step": 10000
+    },
+    {
+      "epoch": 0.25,
+      "grad_norm": 17.375,
+      "learning_rate": 7.158340961603695e-05,
+      "loss": 0.2863,
+      "step": 20000
+    },
+    {
+      "epoch": 0.375,
+      "grad_norm": 11.875,
+      "learning_rate": 4.08424848470673e-05,
+      "loss": 0.2011,
+      "step": 30000
+    },
+    {
+      "epoch": 0.5,
+      "grad_norm": 16.25,
+      "learning_rate": 1.5017755491800084e-05,
+      "loss": 0.1537,
+      "step": 40000
+    },
+    {
+      "epoch": 0.625,
+      "grad_norm": 12.875,
+      "learning_rate": 5.000000103890576e-06,
+      "loss": 0.1406,
+      "step": 50000
+    },
+    {
+      "epoch": 0.625,
+      "eval_loss": 2.7418313026428223,
+      "eval_runtime": 50.5148,
+      "eval_samples_per_second": 400.219,
+      "eval_steps_per_second": 400.219,
+      "step": 50000
+    }
+  ],
+  "logging_steps": 10000,
+  "max_steps": 50000,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 1,
+  "save_steps": 50000,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": true
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 0.0,
+  "train_batch_size": 1,
+  "trial_name": null,
+  "trial_params": null
+}

checkpoint-50000/training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:64ce60c98eb243ba0fd307e9bcc15f6403531e42eace07bf6449d73b31ae1b71
+size 5432