Upload checkpoint-50000

checkpoint-50000/config.json

@@ -0,0 +1,187 @@
+{
+  "batch_size": 64,
+  "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_merged_20250617",
+  "pretrained": {
+    "kwargs": {
+      "subpath": "base"
+    },
+    "model": "hf://chuanmew/octo_torch"
+  },
+  "run_name": "ljp_qiwu_merged_20250617",
+  "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_merged_20250519to20250529_20250604",
+  "training_arguments": {
+    "bf16": true,
+    "dataloader_pin_memory": true,
+    "ddp_find_unused_parameters": false,
+    "eval_steps": 5000,
+    "gradient_accumulation_steps": 1,
+    "learning_rate": 0.0003,
+    "logging_nan_inf_filter": false,
+    "logging_steps": 5000,
+    "lr_scheduler_kwargs": {
+      "decay_type": "cosine",
+      "min_lr_ratio": 0.1,
+      "num_decay_steps": 50000
+    },
+    "lr_scheduler_type": "warmup_stable_decay",
+    "max_grad_norm": 1,
+    "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": 2000,
+    "weight_decay": 0.01
+  },
+  "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": [
+                -0.001587849808856845,
+                0.0010738142300397158,
+                -0.000697988027241081,
+                -0.006544447969645262,
+                0.009556379169225693,
+                0.020673424005508423,
+                0.45607081055641174
+            ],
+            "std": [
+                0.03518082574009895,
+                0.02438654750585556,
+                0.0410635769367218,
+                0.11703108996152878,
+                0.10797987133264542,
+                0.17201891541481018,
+                0.46286532282829285
+            ],
+            "min": [
+                -0.2540467381477356,
+                -0.16329476237297058,
+                -0.19035658240318298,
+                -1.4179747104644775,
+                -0.8620505332946777,
+                -2.1886229515075684,
+                0.0
+            ],
+            "max": [
+                0.18435201048851013,
+                0.15960073471069336,
+                0.2124226987361908,
+                1.345801591873169,
+                1.2207748889923096,
+                2.1311380863189697,
+                1.0
+            ],
+            "mask": [
+                true,
+                true,
+                true,
+                true,
+                true,
+                true,
+                false
+            ]
+        },
+        "state": {
+            "mean": [
+                -1.0741499662399292,
+                -1.0714620351791382,
+                -1.357372760772705,
+                -0.6174096465110779,
+                -0.880132257938385,
+                0.6550877094268799,
+                0.4558027982711792
+            ],
+            "std": [
+                0.2520765960216522,
+                0.2545098066329956,
+                0.1924552172422409,
+                0.559248685836792,
+                0.41956013441085815,
+                1.6006327867507935,
+                0.46324989199638367
+            ],
+            "min": [
+                -2.1934425830841064,
+                -1.8970158100128174,
+                -2.2100930213928223,
+                -3.1213817596435547,
+                -2.0163440704345703,
+                -6.283377170562744,
+                0.0
+            ],
+            "max": [
+                -0.30637508630752563,
+                -0.13648474216461182,
+                -0.47022661566734314,
+                2.9531495571136475,
+                1.5494160652160645,
+                3.9177753925323486,
+                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 sha256:766b98569fcfd684c0c784c74a997d8d821e49dd378b31027ec717593b5f8f1d
+size 405079824

checkpoint-50000/optimizer.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:84708e6e31c0f7b2f3625075b8e64466ab7415d46182fe238b4e3f63fb9ce6d5
+size 371730170

checkpoint-50000/rng_state_0.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3a650642393a3a381a8a2823234abb6a563a46c3dcc0e697ec52ae03202e97fe
+size 15984

checkpoint-50000/rng_state_1.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:937a6237622e2c1cab9836e08409d6dca738bbfaac7903ec67708ff3eb967473
+size 15984

checkpoint-50000/rng_state_2.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b1ff7dc9506f1716fd24bcfa48cccfe1f49e6205c22585de7b5eaa0a211a7eb7
+size 15984

checkpoint-50000/rng_state_3.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0ddfd0e87c3201ae1c793a61be552947fc95ef008a4c1c4ebf7ec7afdd54414a
+size 15984

checkpoint-50000/scheduler.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3f9d92856c22d5b435e814cf3cb979c1f63908e6d47c831c5870cff56f147496
+size 1064

checkpoint-50000/trainer_state.json

@@ -0,0 +1,184 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 5.0,
+  "eval_steps": 5000,
+  "global_step": 50000,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.5,
+      "grad_norm": 16.625,
+      "learning_rate": 0.00029339525062130936,
+      "loss": 0.6821,
+      "step": 5000
+    },
+    {
+      "epoch": 0.5,
+      "eval_loss": 0.8639410734176636,
+      "eval_runtime": 14.9757,
+      "eval_samples_per_second": 1349.986,
+      "eval_steps_per_second": 337.547,
+      "step": 5000
+    },
+    {
+      "epoch": 1.0,
+      "grad_norm": 11.0,
+      "learning_rate": 0.00027422227979597064,
+      "loss": 0.6693,
+      "step": 10000
+    },
+    {
+      "epoch": 1.0,
+      "eval_loss": 1.0540602207183838,
+      "eval_runtime": 13.9502,
+      "eval_samples_per_second": 1449.223,
+      "eval_steps_per_second": 362.36,
+      "step": 10000
+    },
+    {
+      "epoch": 1.5,
+      "grad_norm": 10.25,
+      "learning_rate": 0.0002443578712278313,
+      "loss": 0.4299,
+      "step": 15000
+    },
+    {
+      "epoch": 1.5,
+      "eval_loss": 1.183585286140442,
+      "eval_runtime": 14.3742,
+      "eval_samples_per_second": 1406.479,
+      "eval_steps_per_second": 351.672,
+      "step": 15000
+    },
+    {
+      "epoch": 2.0,
+      "grad_norm": 7.40625,
+      "learning_rate": 0.00020672536130511077,
+      "loss": 0.3463,
+      "step": 20000
+    },
+    {
+      "epoch": 2.0,
+      "eval_loss": 1.3365648984909058,
+      "eval_runtime": 13.9804,
+      "eval_samples_per_second": 1446.098,
+      "eval_steps_per_second": 361.578,
+      "step": 20000
+    },
+    {
+      "epoch": 2.5,
+      "grad_norm": 5.4375,
+      "learning_rate": 0.00016500848230015908,
+      "loss": 0.2826,
+      "step": 25000
+    },
+    {
+      "epoch": 2.5,
+      "eval_loss": 1.6001883745193481,
+      "eval_runtime": 14.1078,
+      "eval_samples_per_second": 1433.035,
+      "eval_steps_per_second": 358.312,
+      "step": 25000
+    },
+    {
+      "epoch": 3.0,
+      "grad_norm": 10.25,
+      "learning_rate": 0.00012329077298856825,
+      "loss": 0.229,
+      "step": 30000
+    },
+    {
+      "epoch": 3.0,
+      "eval_loss": 1.7588920593261719,
+      "eval_runtime": 13.9516,
+      "eval_samples_per_second": 1449.085,
+      "eval_steps_per_second": 362.325,
+      "step": 30000
+    },
+    {
+      "epoch": 3.5,
+      "grad_norm": 5.4375,
+      "learning_rate": 8.565585342212883e-05,
+      "loss": 0.1854,
+      "step": 35000
+    },
+    {
+      "epoch": 3.5,
+      "eval_loss": 2.074185371398926,
+      "eval_runtime": 14.3285,
+      "eval_samples_per_second": 1410.965,
+      "eval_steps_per_second": 352.794,
+      "step": 35000
+    },
+    {
+      "epoch": 4.0,
+      "grad_norm": 5.21875,
+      "learning_rate": 5.5787691745907415e-05,
+      "loss": 0.1527,
+      "step": 40000
+    },
+    {
+      "epoch": 4.0,
+      "eval_loss": 2.3301103115081787,
+      "eval_runtime": 14.184,
+      "eval_samples_per_second": 1425.335,
+      "eval_steps_per_second": 356.387,
+      "step": 40000
+    },
+    {
+      "epoch": 4.5,
+      "grad_norm": 7.15625,
+      "learning_rate": 3.6609991728491735e-05,
+      "loss": 0.1346,
+      "step": 45000
+    },
+    {
+      "epoch": 4.5,
+      "eval_loss": 2.443498373031616,
+      "eval_runtime": 14.4995,
+      "eval_samples_per_second": 1394.32,
+      "eval_steps_per_second": 348.632,
+      "step": 45000
+    },
+    {
+      "epoch": 5.0,
+      "grad_norm": 4.875,
+      "learning_rate": 3.000000026647932e-05,
+      "loss": 0.1258,
+      "step": 50000
+    },
+    {
+      "epoch": 5.0,
+      "eval_loss": 2.586019277572632,
+      "eval_runtime": 14.4123,
+      "eval_samples_per_second": 1402.757,
+      "eval_steps_per_second": 350.741,
+      "step": 50000
+    }
+  ],
+  "logging_steps": 5000,
+  "max_steps": 50000,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 5,
+  "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 sha256:2f76da97d105f4a011c8c8ceea81db6ba502c8ad553055fff92ef2afca7eda75
+size 5432