test-2-checkpoint-2000

test-2-checkpoint-2000/config.json

@@ -0,0 +1,165 @@
+{
+  "batch_size": 8,
+  "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"
+      }
+    },
+    "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
+  },
+  "output_dir": "qiwu_checkpoints3",
+  "pretrained": {
+    "kwargs": {
+      "subpath": "oxe-g2-checkpoint-300000"
+    },
+    "model": "hf://chuanmew/octo_torch"
+  },
+  "run_name": "qiwu",
+  "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": "qiwu_train",
+  "training_arguments": {
+    "bf16": true,
+    "dataloader_pin_memory": true,
+    "ddp_find_unused_parameters": true,
+    "eval_steps": 100,
+    "gradient_accumulation_steps": 1,
+    "learning_rate": 3e-05,
+    "logging_nan_inf_filter": false,
+    "logging_steps": 100,
+    "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": 300000,
+    "optim": "adamw_torch_fused",
+    "per_device_eval_batch_size": 1,
+    "per_device_train_batch_size": 1,
+    "report_to": "wandb",
+    "save_safetensors": true,
+    "save_steps": 500,
+    "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*"
+  ]
+}

test-2-checkpoint-2000/dataset_statistics.json

@@ -0,0 +1,98 @@
+[
+    {
+        "action": {
+            "mean": [
+                0.005713334307074547,
+                -8.683370651851874e-06,
+                -0.0027405666187405586,
+                -0.01428219024091959,
+                -0.00569964898750186,
+                0.03178633004426956,
+                0.31847861409187317
+            ],
+            "std": [
+                0.017133653163909912,
+                0.012163525447249413,
+                0.028012460097670555,
+                0.056618135422468185,
+                0.059768617153167725,
+                0.08718060702085495,
+                0.45297929644584656
+            ],
+            "min": [
+                -0.045306507498025894,
+                -0.058269500732421875,
+                -0.0945921391248703,
+                -0.39226841926574707,
+                -0.31778502464294434,
+                -0.46861958503723145,
+                0.0
+            ],
+            "max": [
+                0.10087767243385315,
+                0.06844352185726166,
+                0.13764026761054993,
+                0.39099645614624023,
+                0.4911177158355713,
+                0.7427346706390381,
+                1.0
+            ],
+            "mask": [
+                true,
+                true,
+                true,
+                true,
+                true,
+                true,
+                false
+            ]
+        },
+        "state": {
+            "mean": [
+                -1.2245632410049438,
+                -1.050716519355774,
+                -1.258064866065979,
+                -0.5181024670600891,
+                -0.9791191816329956,
+                0.11057784408330917,
+                0.26232674717903137
+            ],
+            "std": [
+                0.23113861680030823,
+                0.32922741770744324,
+                0.2693513333797455,
+                0.32120275497436523,
+                0.2501857280731201,
+                0.4673791527748108,
+                0.42628398537635803
+            ],
+            "min": [
+                -1.8634008169174194,
+                -1.9834270477294922,
+                -1.991769790649414,
+                -1.7809165716171265,
+                -1.5987040996551514,
+                -1.2469679117202759,
+                0.0
+            ],
+            "max": [
+                -0.6335545182228088,
+                -0.20619319379329681,
+                -0.40194931626319885,
+                0.5385388135910034,
+                0.03621558099985123,
+                1.5524179935455322,
+                1.0
+            ],
+            "mask": [
+                true,
+                true,
+                true,
+                true,
+                true,
+                true,
+                false
+            ]
+        }
+    }
+]

test-2-checkpoint-2000/model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7e81d2759d04c5eea1db4761579d4ea8d5800da063e61d0c9163cba0d4f99211
+size 401189544

test-2-checkpoint-2000/optimizer.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fae7b36a068eb17c87bd508cde288ffa61390dcc97eee726889a1ba37a8b9a1f
+size 363938106

test-2-checkpoint-2000/rng_state_0.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bff0878a4eccca3b9b1de510ca09955fd298dd8736b88c64ef1cb591012362bb
+size 15984

test-2-checkpoint-2000/rng_state_1.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6e91e14970a95a3ee119899e921e9a25f28fe2b10aedfcdfc6e53c3eb5676ea5
+size 15984

test-2-checkpoint-2000/rng_state_2.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:adff690642974293f57c4e52fc8b88bc2c078610025d855770838fa01d4c6754
+size 15984

test-2-checkpoint-2000/rng_state_3.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:341b46179aa9c8e706b9952c1d8c72351cf0f5b8b157b2a61269b6ff9a3ff11c
+size 15984

test-2-checkpoint-2000/rng_state_4.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6c45f63512c7c687da50d86a1ea8adec7a873da19263066ffbdb71c12b40577f
+size 15984

test-2-checkpoint-2000/rng_state_5.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1f436d86dc256fd381c267b9a0b9b45b2e128e415ef17631b54d11405f88661b
+size 15984

test-2-checkpoint-2000/rng_state_6.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4e6ffe78f3dea063345a309ca0ed415f2e804a9b91d2b859a5ecfa86d6838915
+size 15984

test-2-checkpoint-2000/rng_state_7.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3863022327731121b0562549e522fd7c7ef9819958d81ea1a7d666c33da4f919
+size 15984

test-2-checkpoint-2000/scheduler.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7f8706422a303b331fd63bcf4a251c7596a2eb9e3c2b1bf05f8de25d89a479e0
+size 1064

test-2-checkpoint-2000/trainer_state.json

@@ -0,0 +1,334 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 0.05,
+  "eval_steps": 100,
+  "global_step": 2000,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.0025,
+      "grad_norm": 18.5,
+      "learning_rate": 2.9715000000000003e-05,
+      "loss": 1.316,
+      "step": 100
+    },
+    {
+      "epoch": 0.0025,
+      "eval_loss": 1.1122723817825317,
+      "eval_runtime": 14.3147,
+      "eval_samples_per_second": 1412.327,
+      "eval_steps_per_second": 176.602,
+      "step": 100
+    },
+    {
+      "epoch": 0.005,
+      "grad_norm": 19.75,
+      "learning_rate": 3e-05,
+      "loss": 0.9953,
+      "step": 200
+    },
+    {
+      "epoch": 0.005,
+      "eval_loss": 1.021384596824646,
+      "eval_runtime": 13.8082,
+      "eval_samples_per_second": 1464.134,
+      "eval_steps_per_second": 183.08,
+      "step": 200
+    },
+    {
+      "epoch": 0.0075,
+      "grad_norm": 9.1875,
+      "learning_rate": 3e-05,
+      "loss": 0.954,
+      "step": 300
+    },
+    {
+      "epoch": 0.0075,
+      "eval_loss": 0.9818989038467407,
+      "eval_runtime": 13.8032,
+      "eval_samples_per_second": 1464.658,
+      "eval_steps_per_second": 183.146,
+      "step": 300
+    },
+    {
+      "epoch": 0.01,
+      "grad_norm": 11.375,
+      "learning_rate": 3e-05,
+      "loss": 0.9414,
+      "step": 400
+    },
+    {
+      "epoch": 0.01,
+      "eval_loss": 0.9455794095993042,
+      "eval_runtime": 13.7724,
+      "eval_samples_per_second": 1467.937,
+      "eval_steps_per_second": 183.556,
+      "step": 400
+    },
+    {
+      "epoch": 0.0125,
+      "grad_norm": 13.125,
+      "learning_rate": 3e-05,
+      "loss": 0.8945,
+      "step": 500
+    },
+    {
+      "epoch": 0.0125,
+      "eval_loss": 0.9558092951774597,
+      "eval_runtime": 13.7473,
+      "eval_samples_per_second": 1470.617,
+      "eval_steps_per_second": 183.891,
+      "step": 500
+    },
+    {
+      "epoch": 0.015,
+      "grad_norm": 12.3125,
+      "learning_rate": 3e-05,
+      "loss": 0.9172,
+      "step": 600
+    },
+    {
+      "epoch": 0.015,
+      "eval_loss": 0.937251627445221,
+      "eval_runtime": 13.6805,
+      "eval_samples_per_second": 1477.795,
+      "eval_steps_per_second": 184.788,
+      "step": 600
+    },
+    {
+      "epoch": 0.0175,
+      "grad_norm": 15.875,
+      "learning_rate": 3e-05,
+      "loss": 0.8892,
+      "step": 700
+    },
+    {
+      "epoch": 0.0175,
+      "eval_loss": 0.917012631893158,
+      "eval_runtime": 14.0694,
+      "eval_samples_per_second": 1436.945,
+      "eval_steps_per_second": 179.68,
+      "step": 700
+    },
+    {
+      "epoch": 0.02,
+      "grad_norm": 13.875,
+      "learning_rate": 3e-05,
+      "loss": 0.8603,
+      "step": 800
+    },
+    {
+      "epoch": 0.02,
+      "eval_loss": 0.923341691493988,
+      "eval_runtime": 13.5109,
+      "eval_samples_per_second": 1496.347,
+      "eval_steps_per_second": 187.108,
+      "step": 800
+    },
+    {
+      "epoch": 0.0225,
+      "grad_norm": 15.875,
+      "learning_rate": 3e-05,
+      "loss": 0.8581,
+      "step": 900
+    },
+    {
+      "epoch": 0.0225,
+      "eval_loss": 0.9151796698570251,
+      "eval_runtime": 13.6119,
+      "eval_samples_per_second": 1485.246,
+      "eval_steps_per_second": 185.72,
+      "step": 900
+    },
+    {
+      "epoch": 0.025,
+      "grad_norm": 13.3125,
+      "learning_rate": 3e-05,
+      "loss": 0.8519,
+      "step": 1000
+    },
+    {
+      "epoch": 0.025,
+      "eval_loss": 0.8869242668151855,
+      "eval_runtime": 13.6931,
+      "eval_samples_per_second": 1476.44,
+      "eval_steps_per_second": 184.619,
+      "step": 1000
+    },
+    {
+      "epoch": 0.0275,
+      "grad_norm": 10.5625,
+      "learning_rate": 3e-05,
+      "loss": 0.8299,
+      "step": 1100
+    },
+    {
+      "epoch": 0.0275,
+      "eval_loss": 0.8959755301475525,
+      "eval_runtime": 13.5195,
+      "eval_samples_per_second": 1495.393,
+      "eval_steps_per_second": 186.989,
+      "step": 1100
+    },
+    {
+      "epoch": 0.03,
+      "grad_norm": 10.125,
+      "learning_rate": 3e-05,
+      "loss": 0.8479,
+      "step": 1200
+    },
+    {
+      "epoch": 0.03,
+      "eval_loss": 0.9126322269439697,
+      "eval_runtime": 13.5597,
+      "eval_samples_per_second": 1490.959,
+      "eval_steps_per_second": 186.434,
+      "step": 1200
+    },
+    {
+      "epoch": 0.0325,
+      "grad_norm": 13.6875,
+      "learning_rate": 3e-05,
+      "loss": 0.7961,
+      "step": 1300
+    },
+    {
+      "epoch": 0.0325,
+      "eval_loss": 0.8997920751571655,
+      "eval_runtime": 13.6483,
+      "eval_samples_per_second": 1481.282,
+      "eval_steps_per_second": 185.224,
+      "step": 1300
+    },
+    {
+      "epoch": 0.035,
+      "grad_norm": 6.90625,
+      "learning_rate": 3e-05,
+      "loss": 0.7685,
+      "step": 1400
+    },
+    {
+      "epoch": 0.035,
+      "eval_loss": 0.8956555724143982,
+      "eval_runtime": 14.1397,
+      "eval_samples_per_second": 1429.807,
+      "eval_steps_per_second": 178.788,
+      "step": 1400
+    },
+    {
+      "epoch": 0.0375,
+      "grad_norm": 21.375,
+      "learning_rate": 3e-05,
+      "loss": 0.7655,
+      "step": 1500
+    },
+    {
+      "epoch": 0.0375,
+      "eval_loss": 0.9247519373893738,
+      "eval_runtime": 13.8703,
+      "eval_samples_per_second": 1457.57,
+      "eval_steps_per_second": 182.259,
+      "step": 1500
+    },
+    {
+      "epoch": 0.04,
+      "grad_norm": 17.5,
+      "learning_rate": 3e-05,
+      "loss": 0.7969,
+      "step": 1600
+    },
+    {
+      "epoch": 0.04,
+      "eval_loss": 0.9344497919082642,
+      "eval_runtime": 13.6098,
+      "eval_samples_per_second": 1485.47,
+      "eval_steps_per_second": 185.748,
+      "step": 1600
+    },
+    {
+      "epoch": 0.0425,
+      "grad_norm": 26.125,
+      "learning_rate": 3e-05,
+      "loss": 0.7804,
+      "step": 1700
+    },
+    {
+      "epoch": 0.0425,
+      "eval_loss": 0.9101055264472961,
+      "eval_runtime": 14.9224,
+      "eval_samples_per_second": 1354.811,
+      "eval_steps_per_second": 169.41,
+      "step": 1700
+    },
+    {
+      "epoch": 0.045,
+      "grad_norm": 15.6875,
+      "learning_rate": 3e-05,
+      "loss": 0.7683,
+      "step": 1800
+    },
+    {
+      "epoch": 0.045,
+      "eval_loss": 0.8898786902427673,
+      "eval_runtime": 13.7497,
+      "eval_samples_per_second": 1470.365,
+      "eval_steps_per_second": 183.859,
+      "step": 1800
+    },
+    {
+      "epoch": 0.0475,
+      "grad_norm": 14.5,
+      "learning_rate": 3e-05,
+      "loss": 0.7706,
+      "step": 1900
+    },
+    {
+      "epoch": 0.0475,
+      "eval_loss": 0.9011427760124207,
+      "eval_runtime": 13.7318,
+      "eval_samples_per_second": 1472.275,
+      "eval_steps_per_second": 184.098,
+      "step": 1900
+    },
+    {
+      "epoch": 0.05,
+      "grad_norm": 14.6875,
+      "learning_rate": 3e-05,
+      "loss": 0.7418,
+      "step": 2000
+    },
+    {
+      "epoch": 0.05,
+      "eval_loss": 0.9137729406356812,
+      "eval_runtime": 13.7874,
+      "eval_samples_per_second": 1466.341,
+      "eval_steps_per_second": 183.356,
+      "step": 2000
+    }
+  ],
+  "logging_steps": 100,
+  "max_steps": 300000,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 8,
+  "save_steps": 500,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 0.0,
+  "train_batch_size": 1,
+  "trial_name": null,
+  "trial_params": null
+}

test-2-checkpoint-2000/training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:66d16934169c2b3d68a8743262f6d1383f095f508eb4fe41714e5c7b908a3689
+size 5432