Upload checkpoint-5200

checkpoint-5200/config.json

@@ -0,0 +1,150 @@
+{
+  "batch_size": 32,
+  "buffer_size": 64,
+  "eval_mix": "chris_aubo",
+  "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"
+      }
+    },
+    "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
+  },
+  "normalize_method": "sign",
+  "obs_token_nums": {
+    "primary": 256
+  },
+  "output_dir": "ljp_aubo_20250610_test",
+  "pretrained": {
+    "kwargs": {
+      "subpath": "oxe-g2-checkpoint-300000"
+    },
+    "model": "hf://chuanmew/octo_torch"
+  },
+  "run_name": "aubo",
+  "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_aubo_merged_20250519to20250529_20250604",
+  "training_arguments": {
+    "bf16": true,
+    "dataloader_pin_memory": true,
+    "ddp_find_unused_parameters": true,
+    "eval_steps": 1000,
+    "gradient_accumulation_steps": 1,
+    "learning_rate": 3e-05,
+    "logging_nan_inf_filter": false,
+    "logging_steps": 1000,
+    "lr_scheduler_kwargs": {
+      "decay_type": "cosine",
+      "min_lr_ratio": 0.05,
+      "num_decay_steps": 0
+    },
+    "lr_scheduler_type": "warmup_stable_decay",
+    "max_grad_norm": 1.0,
+    "max_steps": 5200,
+    "optim": "adamw_torch_fused",
+    "per_device_eval_batch_size": 1,
+    "per_device_train_batch_size": 1,
+    "report_to": "wandb",
+    "save_safetensors": true,
+    "save_steps": 5200,
+    "torch_compile": true,
+    "warmup_steps": 100,
+    "weight_decay": 0.1
+  },
+  "training_keys": [
+    "*action*",
+    "*observation_tokenizers*",
+    "*transformer.transformer*",
+    "*task_projections*",
+    "*obs_projections*",
+    "*readout_embeddings*",
+    "*task_pos_embeddings*",
+    "*obs_pos_embeddings*"
+  ]
+}

checkpoint-5200/dataset_statistics.json

@@ -0,0 +1,98 @@
+[
+    {
+        "action": {
+            "mean": [
+                -0.00010803448094520718,
+                -0.00029065669514238834,
+                -0.0002709113177843392,
+                -2.0056504581589252e-05,
+                2.6392210656922543e-06,
+                -0.0002713081194087863,
+                0.532845139503479
+            ],
+            "std": [
+                0.006096565164625645,
+                0.006544822361320257,
+                0.007622068747878075,
+                0.0045709386467933655,
+                0.0024091359227895737,
+                0.011421299539506435,
+                0.4991016685962677
+            ],
+            "min": [
+                -0.009999999776482582,
+                -0.009999999776482582,
+                -0.009999999776482582,
+                -0.06283185631036758,
+                -0.06283185631036758,
+                -0.06283185631036758,
+                0.0
+            ],
+            "max": [
+                0.009999999776482582,
+                0.009999999776482582,
+                0.009999999776482582,
+                0.06283185631036758,
+                0.06283185631036758,
+                0.06283185631036758,
+                1.0
+            ],
+            "mask": [
+                true,
+                true,
+                true,
+                true,
+                true,
+                true,
+                false
+            ]
+        },
+        "state": {
+            "mean": [
+                -0.5312747955322266,
+                -0.18303827941417694,
+                0.00013026571832597256,
+                -1.1791322231292725,
+                0.008879062719643116,
+                -0.04497119411826134,
+                0.5303144454956055
+            ],
+            "std": [
+                0.0928654745221138,
+                0.11658772081136703,
+                0.04497113078832626,
+                2.8499042987823486,
+                0.053659822791814804,
+                0.9866101145744324,
+                0.49907076358795166
+            ],
+            "min": [
+                -0.7517302632331848,
+                -0.5310887694358826,
+                -0.11154260486364365,
+                -3.1415765285491943,
+                -0.6673082113265991,
+                -3.099342107772827,
+                0.0
+            ],
+            "max": [
+                -0.25762036442756653,
+                0.1041782796382904,
+                0.1655297726392746,
+                3.141587018966675,
+                0.3089899718761444,
+                3.1375341415405273,
+                1.0
+            ],
+            "mask": [
+                true,
+                true,
+                true,
+                true,
+                true,
+                true,
+                false
+            ]
+        }
+    }
+]

checkpoint-5200/model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0888b1128263528119416f410a46000f6ad96cf426ac7315d40d9cb356da0012
+size 401078544

checkpoint-5200/optimizer.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fdf8a8e17e7e057b7258aca5f96c6e56a1e45950b48d435e5c8c1d14ca182ebb
+size 363714362

checkpoint-5200/rng_state.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d4086c8bca4da6f79892494c895b36c82784fd90616a8ff70d0acb0be57d6988
+size 15958

checkpoint-5200/scheduler.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cfcbd8326c1d981dae3ac84fb07d1afe2482644aa7866ceef8b55cd8846798d5
+size 1064

checkpoint-5200/trainer_state.json

@@ -0,0 +1,109 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 0.065,
+  "eval_steps": 1000,
+  "global_step": 5200,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.0125,
+      "grad_norm": 19.25,
+      "learning_rate": 3e-05,
+      "loss": 0.6663,
+      "step": 1000
+    },
+    {
+      "epoch": 0.0125,
+      "eval_loss": 0.5592955350875854,
+      "eval_runtime": 10.3881,
+      "eval_samples_per_second": 364.457,
+      "eval_steps_per_second": 364.457,
+      "step": 1000
+    },
+    {
+      "epoch": 0.025,
+      "grad_norm": 14.4375,
+      "learning_rate": 3e-05,
+      "loss": 0.521,
+      "step": 2000
+    },
+    {
+      "epoch": 0.025,
+      "eval_loss": 0.5978671908378601,
+      "eval_runtime": 9.3435,
+      "eval_samples_per_second": 405.202,
+      "eval_steps_per_second": 405.202,
+      "step": 2000
+    },
+    {
+      "epoch": 0.0375,
+      "grad_norm": 14.3125,
+      "learning_rate": 3e-05,
+      "loss": 0.48,
+      "step": 3000
+    },
+    {
+      "epoch": 0.0375,
+      "eval_loss": 0.5614296793937683,
+      "eval_runtime": 9.222,
+      "eval_samples_per_second": 410.542,
+      "eval_steps_per_second": 410.542,
+      "step": 3000
+    },
+    {
+      "epoch": 0.05,
+      "grad_norm": 8.75,
+      "learning_rate": 3e-05,
+      "loss": 0.4596,
+      "step": 4000
+    },
+    {
+      "epoch": 0.05,
+      "eval_loss": 0.5450547337532043,
+      "eval_runtime": 9.3144,
+      "eval_samples_per_second": 406.469,
+      "eval_steps_per_second": 406.469,
+      "step": 4000
+    },
+    {
+      "epoch": 0.0625,
+      "grad_norm": 8.625,
+      "learning_rate": 3e-05,
+      "loss": 0.4418,
+      "step": 5000
+    },
+    {
+      "epoch": 0.0625,
+      "eval_loss": 0.5600486993789673,
+      "eval_runtime": 9.2291,
+      "eval_samples_per_second": 410.225,
+      "eval_steps_per_second": 410.225,
+      "step": 5000
+    }
+  ],
+  "logging_steps": 1000,
+  "max_steps": 5200,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 1,
+  "save_steps": 5200,
+  "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-5200/training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:094ed2b645a2be2ced00774718dbfb2861c458e427f38f3a2f7f1b5120e2274b
+size 5432