Kuavo original action: gr00t_n17 checkpoint 30K
Browse files- autostep_v2_bs128_30k/gr00t_n17/checkpoints/checkpoint-30K/README.md +9 -0
- autostep_v2_bs128_30k/gr00t_n17/checkpoints/checkpoint-30K/checkpoint_manifest.json +265 -0
- autostep_v2_bs128_30k/gr00t_n17/checkpoints/checkpoint-30K/config.json +88 -0
- autostep_v2_bs128_30k/gr00t_n17/checkpoints/checkpoint-30K/embodiment_id.json +55 -0
- autostep_v2_bs128_30k/gr00t_n17/checkpoints/checkpoint-30K/experiment_cfg/conf.yaml +232 -0
- autostep_v2_bs128_30k/gr00t_n17/checkpoints/checkpoint-30K/experiment_cfg/config.yaml +254 -0
- autostep_v2_bs128_30k/gr00t_n17/checkpoints/checkpoint-30K/experiment_cfg/dataset_statistics.json +294 -0
- autostep_v2_bs128_30k/gr00t_n17/checkpoints/checkpoint-30K/experiment_cfg/final_model_config.json +55 -0
- autostep_v2_bs128_30k/gr00t_n17/checkpoints/checkpoint-30K/experiment_cfg/final_processor_config.json +0 -0
- autostep_v2_bs128_30k/gr00t_n17/checkpoints/checkpoint-30K/model-00001-of-00002.safetensors +3 -0
- autostep_v2_bs128_30k/gr00t_n17/checkpoints/checkpoint-30K/model-00002-of-00002.safetensors +3 -0
- autostep_v2_bs128_30k/gr00t_n17/checkpoints/checkpoint-30K/model.safetensors.index.json +0 -0
- autostep_v2_bs128_30k/gr00t_n17/checkpoints/checkpoint-30K/processor_config.json +1140 -0
- autostep_v2_bs128_30k/gr00t_n17/checkpoints/checkpoint-30K/provenance/configs/Bs128_30k/data/dataset_autostep_v2.yaml +31 -0
- autostep_v2_bs128_30k/gr00t_n17/checkpoints/checkpoint-30K/provenance/configs/Bs128_30k/train/gr00t_n17_autostep_v2.yaml +57 -0
- autostep_v2_bs128_30k/gr00t_n17/checkpoints/checkpoint-30K/provenance/configs/Bs128_30k/train/launch.sh +18 -0
- autostep_v2_bs128_30k/gr00t_n17/checkpoints/checkpoint-30K/provenance/data_audit/data_manifest.json +56 -0
- autostep_v2_bs128_30k/gr00t_n17/checkpoints/checkpoint-30K/provenance/data_audit/statistics.json +335 -0
- autostep_v2_bs128_30k/gr00t_n17/checkpoints/checkpoint-30K/provenance/launch_contract.json +28 -0
- autostep_v2_bs128_30k/gr00t_n17/checkpoints/checkpoint-30K/provenance/scripts/__init__.py +1 -0
- autostep_v2_bs128_30k/gr00t_n17/checkpoints/checkpoint-30K/provenance/scripts/adapters/__init__.py +1 -0
- autostep_v2_bs128_30k/gr00t_n17/checkpoints/checkpoint-30K/provenance/scripts/adapters/groot_v3.py +86 -0
- autostep_v2_bs128_30k/gr00t_n17/checkpoints/checkpoint-30K/provenance/scripts/data/__init__.py +1 -0
- autostep_v2_bs128_30k/gr00t_n17/checkpoints/checkpoint-30K/provenance/scripts/data/prepare.py +18 -0
- autostep_v2_bs128_30k/gr00t_n17/checkpoints/checkpoint-30K/provenance/scripts/data/validate.py +67 -0
- autostep_v2_bs128_30k/gr00t_n17/checkpoints/checkpoint-30K/provenance/scripts/data/view.py +166 -0
- autostep_v2_bs128_30k/gr00t_n17/checkpoints/checkpoint-30K/provenance/scripts/train/__init__.py +1 -0
- autostep_v2_bs128_30k/gr00t_n17/checkpoints/checkpoint-30K/provenance/scripts/train/checkpoints.py +35 -0
- autostep_v2_bs128_30k/gr00t_n17/checkpoints/checkpoint-30K/provenance/scripts/train/hf_watcher.py +371 -0
- autostep_v2_bs128_30k/gr00t_n17/checkpoints/checkpoint-30K/provenance/scripts/train/native_assets.py +48 -0
- autostep_v2_bs128_30k/gr00t_n17/checkpoints/checkpoint-30K/provenance/scripts/train/run_n17.py +173 -0
- autostep_v2_bs128_30k/gr00t_n17/checkpoints/checkpoint-30K/rwh_save_complete.json +4 -0
- autostep_v2_bs128_30k/gr00t_n17/checkpoints/checkpoint-30K/statistics.json +0 -0
- autostep_v2_bs128_30k/gr00t_n17/checkpoints/checkpoint-30K/trainer_state.json +1834 -0
- autostep_v2_bs128_30k/gr00t_n17/checkpoints/checkpoint-30K/training_args.bin +3 -0
- autostep_v2_bs128_30k/gr00t_n17/checkpoints/checkpoint-30K/wandb_config.json +1 -0
autostep_v2_bs128_30k/gr00t_n17/checkpoints/checkpoint-30K/README.md
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
# gr00t_n17: original-action Kuavo, checkpoint 30000 of 30000 updates
|
| 2 |
+
|
| 3 |
+
Original v3 data; 30Hz to 10Hz logical view; state20/action23; seed42; actual batch128; accumulation1.
|
| 4 |
+
Torso targets are the original action commands. Invalid action elements are masked.
|
| 5 |
+
N1.5/DiT use LeTools-pinned LeRobot; N1.7 uses bundled native model, processor, sharded loader and Trainer.
|
| 6 |
+
LeTools revision: 6d2002613241ae2a32c5d8fd753f80caa1383fba. See provenance for adapter and YAML.
|
| 7 |
+
N1.7 uses native checkpoint layout. Optimizer/RNG states stay on the training server.
|
| 8 |
+
|
| 9 |
+
Dataset: `RwH-Lab/rwh-auto-step-v2` at revision `7ccc8d62be1d6ff997267dcf55b831cbe865eedd`.
|
autostep_v2_bs128_30k/gr00t_n17/checkpoints/checkpoint-30K/checkpoint_manifest.json
ADDED
|
@@ -0,0 +1,265 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"schema": 2,
|
| 3 |
+
"policy": "gr00t_n17",
|
| 4 |
+
"step": 30000,
|
| 5 |
+
"repo_id": "RwH-Lab/Kuavo_ckpt",
|
| 6 |
+
"remote_prefix": "autostep_v2_bs128_30k/gr00t_n17/checkpoints/checkpoint-30K",
|
| 7 |
+
"source_checkpoint": "/home/work/mntvol/RwH-Kuavo_ckpt/autostep_v2_bs128_30k/gr00t_n17/checkpoint-30000",
|
| 8 |
+
"source_signature": {
|
| 9 |
+
"config.json": [
|
| 10 |
+
2146,
|
| 11 |
+
1793210529475846999
|
| 12 |
+
],
|
| 13 |
+
"embodiment_id.json": [
|
| 14 |
+
2169,
|
| 15 |
+
1793172955178082267
|
| 16 |
+
],
|
| 17 |
+
"experiment_cfg/conf.yaml": [
|
| 18 |
+
5309,
|
| 19 |
+
1793172941696201016
|
| 20 |
+
],
|
| 21 |
+
"experiment_cfg/config.yaml": [
|
| 22 |
+
6424,
|
| 23 |
+
1793172946949716724
|
| 24 |
+
],
|
| 25 |
+
"experiment_cfg/dataset_statistics.json": [
|
| 26 |
+
8173,
|
| 27 |
+
1793172989723336625
|
| 28 |
+
],
|
| 29 |
+
"experiment_cfg/final_model_config.json": [
|
| 30 |
+
1478,
|
| 31 |
+
1793172989894993824
|
| 32 |
+
],
|
| 33 |
+
"experiment_cfg/final_processor_config.json": [
|
| 34 |
+
2547727,
|
| 35 |
+
1793172954878175539
|
| 36 |
+
],
|
| 37 |
+
"model-00001-of-00002.safetensors": [
|
| 38 |
+
4993709288,
|
| 39 |
+
1793210573342253324
|
| 40 |
+
],
|
| 41 |
+
"model-00002-of-00002.safetensors": [
|
| 42 |
+
4535492016,
|
| 43 |
+
1793210541512887896
|
| 44 |
+
],
|
| 45 |
+
"model.safetensors.index.json": [
|
| 46 |
+
104912,
|
| 47 |
+
1793210577305905350
|
| 48 |
+
],
|
| 49 |
+
"optimizer.pt": [
|
| 50 |
+
12964594710,
|
| 51 |
+
1793210596139068160
|
| 52 |
+
],
|
| 53 |
+
"processor_config.json": [
|
| 54 |
+
26765,
|
| 55 |
+
1793172949638902680
|
| 56 |
+
],
|
| 57 |
+
"rng_state.pth": [
|
| 58 |
+
14645,
|
| 59 |
+
1793210558436406903
|
| 60 |
+
],
|
| 61 |
+
"rwh_save_complete.json": [
|
| 62 |
+
105,
|
| 63 |
+
1793210594638572328
|
| 64 |
+
],
|
| 65 |
+
"scheduler.pt": [
|
| 66 |
+
1465,
|
| 67 |
+
1793210553029346010
|
| 68 |
+
],
|
| 69 |
+
"statistics.json": [
|
| 70 |
+
4301266,
|
| 71 |
+
1793172992080098505
|
| 72 |
+
],
|
| 73 |
+
"trainer_state.json": [
|
| 74 |
+
42822,
|
| 75 |
+
1793210594003434734
|
| 76 |
+
],
|
| 77 |
+
"training_args.bin": [
|
| 78 |
+
5905,
|
| 79 |
+
1793210541751090776
|
| 80 |
+
],
|
| 81 |
+
"wandb_config.json": [
|
| 82 |
+
43,
|
| 83 |
+
1793172941697917785
|
| 84 |
+
]
|
| 85 |
+
},
|
| 86 |
+
"tensor_count": 1030,
|
| 87 |
+
"training_state_included": false,
|
| 88 |
+
"files": {
|
| 89 |
+
"README.md": {
|
| 90 |
+
"size": 642,
|
| 91 |
+
"sha256": "36979bde02d3f01328c1d273a606105d28f814b7045a671f65e85b69fdd11f37",
|
| 92 |
+
"git_blob_sha1": "b9590a09f0d306bc520694d25f0525834f80546e"
|
| 93 |
+
},
|
| 94 |
+
"config.json": {
|
| 95 |
+
"size": 2146,
|
| 96 |
+
"sha256": "1b018c4df929422667e849befdf94b07dbee42cc119055d03f5b57faa87b7d08",
|
| 97 |
+
"git_blob_sha1": "89955b375812296d45a984b89b46cb7e090034d5"
|
| 98 |
+
},
|
| 99 |
+
"embodiment_id.json": {
|
| 100 |
+
"size": 2169,
|
| 101 |
+
"sha256": "6f01100d50c52602f0c7c68304eea48f79a1f8c2426a960c04824c0ed4557869",
|
| 102 |
+
"git_blob_sha1": "292403c3caa9fbdd10d2d09693248bc275b9417d"
|
| 103 |
+
},
|
| 104 |
+
"experiment_cfg/conf.yaml": {
|
| 105 |
+
"size": 5309,
|
| 106 |
+
"sha256": "1b7d90bad4a406f521f1f3346a3f5a51eec25ee032d052b566a71f73387403bb",
|
| 107 |
+
"git_blob_sha1": "8f914e9d83dbd7f0ea00b07d88ae12124efb474a"
|
| 108 |
+
},
|
| 109 |
+
"experiment_cfg/config.yaml": {
|
| 110 |
+
"size": 6424,
|
| 111 |
+
"sha256": "086ff46ce4c63946c2c808ca39a8716aec73572ea706da0327e1d4174ace2155",
|
| 112 |
+
"git_blob_sha1": "f6b83d543b3085d98a656ef7b2c884ecafa90031"
|
| 113 |
+
},
|
| 114 |
+
"experiment_cfg/dataset_statistics.json": {
|
| 115 |
+
"size": 8173,
|
| 116 |
+
"sha256": "66c7efb4c2a93e3d00d5ef650eff603d3e15a6c37bcf791ce22884cd1383b533",
|
| 117 |
+
"git_blob_sha1": "110992d62bec43c968195a8db1442663d92950fe"
|
| 118 |
+
},
|
| 119 |
+
"experiment_cfg/final_model_config.json": {
|
| 120 |
+
"size": 1478,
|
| 121 |
+
"sha256": "5514de8e630417735739240c730dd3a39cdcd76e503cbf1dc09b62d237b4bfcd",
|
| 122 |
+
"git_blob_sha1": "3d4f7f1472efa5c985409e4e9bed41454dfefe31"
|
| 123 |
+
},
|
| 124 |
+
"experiment_cfg/final_processor_config.json": {
|
| 125 |
+
"size": 2547727,
|
| 126 |
+
"sha256": "51f323f19f88c2cac2df709f1dc7705fe915f9702035b80f66b768d4dc54c621",
|
| 127 |
+
"git_blob_sha1": "ef569b7b1de504142312637ceb3581f7542684e3"
|
| 128 |
+
},
|
| 129 |
+
"model-00001-of-00002.safetensors": {
|
| 130 |
+
"size": 4993709288,
|
| 131 |
+
"sha256": "93d2b4a2715711b604606e57d6ab80693b02921b43f58b55bc20704945cd9b55",
|
| 132 |
+
"git_blob_sha1": "d7994621fea7376dcf1bb13eae3e112f648bbdf8"
|
| 133 |
+
},
|
| 134 |
+
"model-00002-of-00002.safetensors": {
|
| 135 |
+
"size": 4535492016,
|
| 136 |
+
"sha256": "eba5cda753e5d7619997d69b33fc4296998defbb4ffaf089ce94e419bbe44502",
|
| 137 |
+
"git_blob_sha1": "684f108dad044275b5bb7c97d1ebb626f2aa965b"
|
| 138 |
+
},
|
| 139 |
+
"model.safetensors.index.json": {
|
| 140 |
+
"size": 104912,
|
| 141 |
+
"sha256": "ff5a9ac57951658b775ef7e958a3469900f767cec60cbf487b3c8d0a3d787bc7",
|
| 142 |
+
"git_blob_sha1": "944b4c8bb57c0b58ec85a014de519abd90a96fa3"
|
| 143 |
+
},
|
| 144 |
+
"processor_config.json": {
|
| 145 |
+
"size": 26765,
|
| 146 |
+
"sha256": "98bea5a6dff3fbcc144923f72161a3222d49d19067ae9e2e2671a945fd92bfec",
|
| 147 |
+
"git_blob_sha1": "86bc2ee1e57911f488eeacb45c10d8c8fc15579a"
|
| 148 |
+
},
|
| 149 |
+
"provenance/configs/Bs128_30k/data/dataset_autostep_v2.yaml": {
|
| 150 |
+
"size": 598,
|
| 151 |
+
"sha256": "71778a003f0eb52cf5154879a831c63f1391d534792217649e588f3c376ec3c7",
|
| 152 |
+
"git_blob_sha1": "382e39c41e999f3eea7f7b2cb15ddf08fa28f5d9"
|
| 153 |
+
},
|
| 154 |
+
"provenance/configs/Bs128_30k/train/gr00t_n17_autostep_v2.yaml": {
|
| 155 |
+
"size": 1405,
|
| 156 |
+
"sha256": "0d9729c76c36792e060368c79a8c53d8b0657d3fb55394acecefca250a56f4c8",
|
| 157 |
+
"git_blob_sha1": "060425a26f73d66394380b3ad1c3f6c22cc5fec5"
|
| 158 |
+
},
|
| 159 |
+
"provenance/configs/Bs128_30k/train/launch.sh": {
|
| 160 |
+
"size": 811,
|
| 161 |
+
"sha256": "a267aadd48e1ac00c564561e89715406a2d653a7ea4fe3dfa8c23f0c56210e10",
|
| 162 |
+
"git_blob_sha1": "e5ce544f5645597ca960084933aad619f24cba28"
|
| 163 |
+
},
|
| 164 |
+
"provenance/data_audit/data_manifest.json": {
|
| 165 |
+
"size": 1253,
|
| 166 |
+
"sha256": "45547a93d1934d7608f9c366185e187f2a262a9a2d9e1baed3ae878210fca7ae",
|
| 167 |
+
"git_blob_sha1": "f6df953492724ef6191ae65156735375f0521cc7"
|
| 168 |
+
},
|
| 169 |
+
"provenance/data_audit/statistics.json": {
|
| 170 |
+
"size": 7581,
|
| 171 |
+
"sha256": "6f96bf46bdd5b4cd82e92e418af49b5a64dc9fb8b41f2db40242741faf293c9b",
|
| 172 |
+
"git_blob_sha1": "e834bab7aa8eba2d0ace86aae57b0915685a9af6"
|
| 173 |
+
},
|
| 174 |
+
"provenance/launch_contract.json": {
|
| 175 |
+
"size": 1291,
|
| 176 |
+
"sha256": "cae2cc20af025718b6ac5600d2cd43c4f9e75f9a95ff2a70e4cb57e34bc525fb",
|
| 177 |
+
"git_blob_sha1": "35421c60841e97acde50c92798df837bb23fa952"
|
| 178 |
+
},
|
| 179 |
+
"provenance/scripts/__init__.py": {
|
| 180 |
+
"size": 65,
|
| 181 |
+
"sha256": "d5cb3b6372ce5367b0e1c6a3aabcacf5ae0c69c1f123c3fdc768dbdab86615fd",
|
| 182 |
+
"git_blob_sha1": "b60a71c83c559cc4515886759eb03bff532ef809"
|
| 183 |
+
},
|
| 184 |
+
"provenance/scripts/adapters/__init__.py": {
|
| 185 |
+
"size": 85,
|
| 186 |
+
"sha256": "d66d70788ab53bdfd8e0235ba7150604476d3654dece06d0871aa512bbf02db0",
|
| 187 |
+
"git_blob_sha1": "ed130013a85392ebd51771dbb14d18e0f3aa164a"
|
| 188 |
+
},
|
| 189 |
+
"provenance/scripts/adapters/groot_v3.py": {
|
| 190 |
+
"size": 4834,
|
| 191 |
+
"sha256": "f23c095efaefcaa3bbeb20487ba70141520783ffd820cf39354b9ee1a93463b1",
|
| 192 |
+
"git_blob_sha1": "577fef6bfdb8f6bd13c4b8ee48316fabc5fd1809"
|
| 193 |
+
},
|
| 194 |
+
"provenance/scripts/data/__init__.py": {
|
| 195 |
+
"size": 38,
|
| 196 |
+
"sha256": "d3bc18a39403772a7ab3803cb57fd46fa6b47c10140162291029a1800b62cfc0",
|
| 197 |
+
"git_blob_sha1": "263a31b0589ca2481bfddad6515eb3a86a645c12"
|
| 198 |
+
},
|
| 199 |
+
"provenance/scripts/data/prepare.py": {
|
| 200 |
+
"size": 492,
|
| 201 |
+
"sha256": "f37d9e1c00c0abc08c32851ea50931c5a810a161fb4d0b602a916ac120bd6bb6",
|
| 202 |
+
"git_blob_sha1": "343931e66cea374ec8beb6d1e94830a67c167bb3"
|
| 203 |
+
},
|
| 204 |
+
"provenance/scripts/data/validate.py": {
|
| 205 |
+
"size": 3761,
|
| 206 |
+
"sha256": "ea0a55b480136c519baaae589ad527a4953338ed8894a89fce80c33ffdd1ad93",
|
| 207 |
+
"git_blob_sha1": "564aeb727b01544a83e48f40f48dbe4026e4f75c"
|
| 208 |
+
},
|
| 209 |
+
"provenance/scripts/data/view.py": {
|
| 210 |
+
"size": 10237,
|
| 211 |
+
"sha256": "ca7f03c7fca94575b9e089bed2727981406791b8ddbf382b61059bc3bbf937e6",
|
| 212 |
+
"git_blob_sha1": "7c86c48b6815e79f303edd54a4bf4db45b89f339"
|
| 213 |
+
},
|
| 214 |
+
"provenance/scripts/train/__init__.py": {
|
| 215 |
+
"size": 34,
|
| 216 |
+
"sha256": "5f51caf72525efcf81ce8bdfc06321746fd185d4c932f734a3b0ffa141af74ce",
|
| 217 |
+
"git_blob_sha1": "61ef7af826bb64fd194a8cabe53818989f3cbcdb"
|
| 218 |
+
},
|
| 219 |
+
"provenance/scripts/train/checkpoints.py": {
|
| 220 |
+
"size": 1766,
|
| 221 |
+
"sha256": "c5feb6d74497a28945f574fad39846c6d8e16f9798bb7fbe60f23af65c5dcb78",
|
| 222 |
+
"git_blob_sha1": "869f01e29a3cf6fd343df5589718dd1d2f82b2cf"
|
| 223 |
+
},
|
| 224 |
+
"provenance/scripts/train/hf_watcher.py": {
|
| 225 |
+
"size": 19261,
|
| 226 |
+
"sha256": "7d2a57f6876a313a995d2d57c778f32571ffde71b51b6983bb1826cd71f5d56f",
|
| 227 |
+
"git_blob_sha1": "495c7885a6323206c62f8b7152ff7f47c8f9531b"
|
| 228 |
+
},
|
| 229 |
+
"provenance/scripts/train/native_assets.py": {
|
| 230 |
+
"size": 2450,
|
| 231 |
+
"sha256": "01ddf2cbef9fe60131a90b0b4866014c9ad7e617f98bad4cd548065658845b3d",
|
| 232 |
+
"git_blob_sha1": "30b2c0504938f8d9e32bdde5b387687a2aba6e70"
|
| 233 |
+
},
|
| 234 |
+
"provenance/scripts/train/run_n17.py": {
|
| 235 |
+
"size": 10066,
|
| 236 |
+
"sha256": "f1a726e523a4a71e59de2d1d69cda3dd7f341dd257b98586c8673d166b6598ec",
|
| 237 |
+
"git_blob_sha1": "1dc14b7aed133e084a729659eeb11766709198a7"
|
| 238 |
+
},
|
| 239 |
+
"rwh_save_complete.json": {
|
| 240 |
+
"size": 105,
|
| 241 |
+
"sha256": "8fe5baf8ec5aefebbcac962201c5fa978bb150d0d39f615c92afe4562458a605",
|
| 242 |
+
"git_blob_sha1": "e199ff0a2a09612cf65b90a516e26a8d61639e73"
|
| 243 |
+
},
|
| 244 |
+
"statistics.json": {
|
| 245 |
+
"size": 4301266,
|
| 246 |
+
"sha256": "ba89e5391e09c236a204cf4a08d71cc13f030ed89661440034fa8f9b21fef30d",
|
| 247 |
+
"git_blob_sha1": "2b624597f3d1acd036176d51cca880faeeddbd38"
|
| 248 |
+
},
|
| 249 |
+
"trainer_state.json": {
|
| 250 |
+
"size": 42822,
|
| 251 |
+
"sha256": "77e6e65e5a00a593ec66c81157b4e85f8136e98fa66971797f3b6bb839be4132",
|
| 252 |
+
"git_blob_sha1": "d6f71159a8845918b17374c3f0dcd9ace7b3a253"
|
| 253 |
+
},
|
| 254 |
+
"training_args.bin": {
|
| 255 |
+
"size": 5905,
|
| 256 |
+
"sha256": "124274f0bf033204ac3ba81ccf3ee9dcbadadbb56a17aa12548b2e6bcdbb37b8",
|
| 257 |
+
"git_blob_sha1": "2934d41791f7b5171bf3fa8ddd20a86802558bc9"
|
| 258 |
+
},
|
| 259 |
+
"wandb_config.json": {
|
| 260 |
+
"size": 43,
|
| 261 |
+
"sha256": "d3aa17c4d9a25dac73c251f4a11a4225a383e57682e21de82eaad8f14d86a3e1",
|
| 262 |
+
"git_blob_sha1": "eccaf5c7c5065b83bc33ea6fd7cffbd71d5a2d10"
|
| 263 |
+
}
|
| 264 |
+
}
|
| 265 |
+
}
|
autostep_v2_bs128_30k/gr00t_n17/checkpoints/checkpoint-30K/config.json
ADDED
|
@@ -0,0 +1,88 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"action_horizon": 40,
|
| 3 |
+
"add_pos_embed": true,
|
| 4 |
+
"apply_sincos_state_encoding": false,
|
| 5 |
+
"architectures": [
|
| 6 |
+
"Gr00tN1d7"
|
| 7 |
+
],
|
| 8 |
+
"attn_dropout": 0.2,
|
| 9 |
+
"attn_implementation": null,
|
| 10 |
+
"backbone_embedding_dim": 2048,
|
| 11 |
+
"backbone_trainable_params_fp32": true,
|
| 12 |
+
"color_jitter_params": {
|
| 13 |
+
"brightness": 0.3,
|
| 14 |
+
"contrast": 0.4,
|
| 15 |
+
"hue": 0.08,
|
| 16 |
+
"saturation": 0.5
|
| 17 |
+
},
|
| 18 |
+
"crop_fraction": 0.95,
|
| 19 |
+
"diffusion_model_cfg": {
|
| 20 |
+
"attention_head_dim": 48,
|
| 21 |
+
"dropout": 0.2,
|
| 22 |
+
"final_dropout": true,
|
| 23 |
+
"interleave_self_attention": true,
|
| 24 |
+
"norm_type": "ada_norm",
|
| 25 |
+
"num_attention_heads": 32,
|
| 26 |
+
"num_layers": 32,
|
| 27 |
+
"output_dim": 1024,
|
| 28 |
+
"positional_embeddings": null
|
| 29 |
+
},
|
| 30 |
+
"dtype": "bfloat16",
|
| 31 |
+
"exclude_state": false,
|
| 32 |
+
"formalize_language": true,
|
| 33 |
+
"hidden_size": 1024,
|
| 34 |
+
"image_crop_size": [
|
| 35 |
+
230,
|
| 36 |
+
230
|
| 37 |
+
],
|
| 38 |
+
"image_target_size": [
|
| 39 |
+
256,
|
| 40 |
+
256
|
| 41 |
+
],
|
| 42 |
+
"letter_box_transform": false,
|
| 43 |
+
"load_bf16": true,
|
| 44 |
+
"max_action_dim": 132,
|
| 45 |
+
"max_num_embodiments": 32,
|
| 46 |
+
"max_seq_len": 1024,
|
| 47 |
+
"max_state_dim": 132,
|
| 48 |
+
"model_dtype": "bfloat16",
|
| 49 |
+
"model_name": "nvidia/Cosmos-Reason2-2B",
|
| 50 |
+
"model_type": "Gr00tN1d7",
|
| 51 |
+
"noise_beta_alpha": 1.5,
|
| 52 |
+
"noise_beta_beta": 1.0,
|
| 53 |
+
"noise_s": 0.999,
|
| 54 |
+
"num_inference_timesteps": 4,
|
| 55 |
+
"num_timestep_buckets": 1000,
|
| 56 |
+
"random_history_crop": true,
|
| 57 |
+
"random_rotation_angle": 0,
|
| 58 |
+
"reproject_vision": false,
|
| 59 |
+
"rtc_ramp_rate": 6.0,
|
| 60 |
+
"select_layer": 16,
|
| 61 |
+
"shortest_image_edge": 256,
|
| 62 |
+
"state_dropout_prob": 0.2,
|
| 63 |
+
"state_gaussian_noise_std": 0.0,
|
| 64 |
+
"transformers_version": "4.57.3",
|
| 65 |
+
"tune_diffusion_model": true,
|
| 66 |
+
"tune_linear": true,
|
| 67 |
+
"tune_llm": false,
|
| 68 |
+
"tune_projector": true,
|
| 69 |
+
"tune_top_llm_layers": 0,
|
| 70 |
+
"tune_visual": false,
|
| 71 |
+
"tune_vlln": true,
|
| 72 |
+
"use_albumentations": true,
|
| 73 |
+
"use_alternate_vl_dit": true,
|
| 74 |
+
"use_flash_attention": true,
|
| 75 |
+
"use_future_tokens": false,
|
| 76 |
+
"use_mean_std": false,
|
| 77 |
+
"use_percentiles": true,
|
| 78 |
+
"use_vl_self_attention": true,
|
| 79 |
+
"use_vlln": true,
|
| 80 |
+
"vl_self_attention_cfg": {
|
| 81 |
+
"attention_head_dim": 64,
|
| 82 |
+
"dropout": 0.2,
|
| 83 |
+
"final_dropout": true,
|
| 84 |
+
"num_attention_heads": 32,
|
| 85 |
+
"num_layers": 4,
|
| 86 |
+
"positional_embeddings": null
|
| 87 |
+
}
|
| 88 |
+
}
|
autostep_v2_bs128_30k/gr00t_n17/checkpoints/checkpoint-30K/embodiment_id.json
ADDED
|
@@ -0,0 +1,55 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"robocasa_panda_omron": 13,
|
| 3 |
+
"oxe_droid": 17,
|
| 4 |
+
"oxe_fractal": 18,
|
| 5 |
+
"oxe_language_table": 19,
|
| 6 |
+
"oxe_bridge": 20,
|
| 7 |
+
"unknown": 22,
|
| 8 |
+
"gr1_unified": 20,
|
| 9 |
+
"agibot": 26,
|
| 10 |
+
"sim_behavior_r1_pro": 23,
|
| 11 |
+
"xdof": 24,
|
| 12 |
+
"xdof_oss_data": 25,
|
| 13 |
+
"unitree_g1_full_body_with_waist_height_nav_cmd": 25,
|
| 14 |
+
"real_r1_pro_sharpa": 27,
|
| 15 |
+
"real_r1_pro_sharpa_add_view": 27,
|
| 16 |
+
"real_r1_pro_sharpa_relative_arm_joint": 26,
|
| 17 |
+
"real_r1_pro_sharpa_delta_eef": 26,
|
| 18 |
+
"real_r1_pro_sharpa_absolute_eef": 26,
|
| 19 |
+
"real_r1_pro_sharpa_meanstd": 26,
|
| 20 |
+
"real_r1_pro_sharpa_relative_eef": 26,
|
| 21 |
+
"real_r1_pro_sharpa_relative_eef_add_view": 26,
|
| 22 |
+
"real_r1_pro_sharpa_relative_eef_relative_hand": 26,
|
| 23 |
+
"real_r1_pro_sharpa_relative_eef_human": 26,
|
| 24 |
+
"real_r1_pro_sharpa_relative_eef_human_add_view": 26,
|
| 25 |
+
"real_r1_pro_sharpa_relative_eef_human_relative_hand": 26,
|
| 26 |
+
"real_r1_pro_sharpa_relative_eef_egodex": 26,
|
| 27 |
+
"real_r1_pro_sharpa_relative_eef_egodex_relative_hand": 26,
|
| 28 |
+
"real_r1_pro_sharpa_relative_eef_egodex_wrist_only": 26,
|
| 29 |
+
"real_r1_pro_sharpa_relative_eef_maxinsights": 26,
|
| 30 |
+
"real_r1_pro_sharpa_relative_eef_maxinsights_relative_hand": 26,
|
| 31 |
+
"real_r1_pro_sharpa_relative_eef_mecka": 26,
|
| 32 |
+
"real_r1_pro_sharpa_relative_eef_mecka_relative_hand": 26,
|
| 33 |
+
"real_g1_relative_eef_absolute_joints": 25,
|
| 34 |
+
"real_g1_relative_eef_absolute_joints_wrist_cam": 25,
|
| 35 |
+
"real_g1_relative_eef_relative_joints": 25,
|
| 36 |
+
"real_r1_pro_sharpa_relative_eef_relative_hand_relative_joint": 26,
|
| 37 |
+
"real_r1_pro_sharpa_relative_joint": 29,
|
| 38 |
+
"oxe_droid_relative_eef_relative_joint": 24,
|
| 39 |
+
"oxe_droid_relative_eef_relative_joint_swapped": 24,
|
| 40 |
+
"oxe_droid_relative_eef_relative_joint_upweight_z": 24,
|
| 41 |
+
"oxe_droid_relative_eef_relative_joint_upweight_z_swapped": 24,
|
| 42 |
+
"oxe_droid_relative_eef_relative_joint_3view": 24,
|
| 43 |
+
"oxe_droid_relative_eef_relative_joint_3view_swapped": 24,
|
| 44 |
+
"oxe_droid_relative_eef": 24,
|
| 45 |
+
"oxe_droid_joint_position_relative": 24,
|
| 46 |
+
"xdof_relative_eef_relative_joint": 27,
|
| 47 |
+
"xdof_relative_eef_relative_joint_subtask": 27,
|
| 48 |
+
"xdof_relative_eef": 27,
|
| 49 |
+
"xdof_relative_joint": 28,
|
| 50 |
+
"simpler_env_google": 0,
|
| 51 |
+
"simpler_env_widowx": 1,
|
| 52 |
+
"libero_sim": 2,
|
| 53 |
+
"droid_sim": 3,
|
| 54 |
+
"new_embodiment": 10
|
| 55 |
+
}
|
autostep_v2_bs128_30k/gr00t_n17/checkpoints/checkpoint-30K/experiment_cfg/conf.yaml
ADDED
|
@@ -0,0 +1,232 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
load_config_path: null
|
| 2 |
+
model:
|
| 3 |
+
model_type: Gr00tN1d7
|
| 4 |
+
model_dtype: bfloat16
|
| 5 |
+
model_name: nvidia/Cosmos-Reason2-2B
|
| 6 |
+
backbone_model_type: qwen
|
| 7 |
+
model_revision: null
|
| 8 |
+
tune_top_llm_layers: 0
|
| 9 |
+
backbone_embedding_dim: 2048
|
| 10 |
+
tune_llm: false
|
| 11 |
+
tune_visual: false
|
| 12 |
+
select_layer: 12
|
| 13 |
+
reproject_vision: false
|
| 14 |
+
use_flash_attention: true
|
| 15 |
+
load_bf16: true
|
| 16 |
+
backbone_trainable_params_fp32: true
|
| 17 |
+
image_crop_size:
|
| 18 |
+
- 230
|
| 19 |
+
- 230
|
| 20 |
+
image_target_size:
|
| 21 |
+
- 256
|
| 22 |
+
- 256
|
| 23 |
+
shortest_image_edge: null
|
| 24 |
+
crop_fraction: null
|
| 25 |
+
random_rotation_angle: null
|
| 26 |
+
color_jitter_params: null
|
| 27 |
+
use_albumentations_transforms: true
|
| 28 |
+
extra_augmentation_config: null
|
| 29 |
+
formalize_language: true
|
| 30 |
+
apply_sincos_state_encoding: false
|
| 31 |
+
use_percentiles: true
|
| 32 |
+
use_relative_action: false
|
| 33 |
+
max_state_dim: 132
|
| 34 |
+
max_action_dim: 132
|
| 35 |
+
action_horizon: 40
|
| 36 |
+
hidden_size: 1024
|
| 37 |
+
input_embedding_dim: 1536
|
| 38 |
+
state_history_length: 1
|
| 39 |
+
add_pos_embed: true
|
| 40 |
+
attn_dropout: 0.2
|
| 41 |
+
use_vlln: true
|
| 42 |
+
max_seq_len: 1024
|
| 43 |
+
use_alternate_vl_dit: true
|
| 44 |
+
attend_text_every_n_blocks: 2
|
| 45 |
+
diffusion_model_cfg:
|
| 46 |
+
positional_embeddings: null
|
| 47 |
+
num_layers: 16
|
| 48 |
+
num_attention_heads: 32
|
| 49 |
+
attention_head_dim: 48
|
| 50 |
+
norm_type: ada_norm
|
| 51 |
+
dropout: 0.2
|
| 52 |
+
final_dropout: true
|
| 53 |
+
output_dim: 1024
|
| 54 |
+
interleave_self_attention: true
|
| 55 |
+
num_inference_timesteps: 4
|
| 56 |
+
noise_beta_alpha: 1.5
|
| 57 |
+
noise_beta_beta: 1.0
|
| 58 |
+
noise_s: 0.999
|
| 59 |
+
num_timestep_buckets: 1000
|
| 60 |
+
tune_projector: true
|
| 61 |
+
tune_diffusion_model: true
|
| 62 |
+
tune_vlln: true
|
| 63 |
+
state_dropout_prob: 0.2
|
| 64 |
+
exclude_state: false
|
| 65 |
+
use_mean_std: false
|
| 66 |
+
max_num_embodiments: 32
|
| 67 |
+
data:
|
| 68 |
+
datasets:
|
| 69 |
+
- dataset_paths:
|
| 70 |
+
- /home/work/mntvol/data/RwH-Kuavo-v2/rwh-auto-step-v2
|
| 71 |
+
embodiment_tag: new_embodiment
|
| 72 |
+
mix_ratio: 1.0
|
| 73 |
+
dataset_type: physical_embodiment
|
| 74 |
+
val_dataset_path: null
|
| 75 |
+
modality_configs:
|
| 76 |
+
new_embodiment:
|
| 77 |
+
video:
|
| 78 |
+
delta_indices:
|
| 79 |
+
- 0
|
| 80 |
+
modality_keys:
|
| 81 |
+
- observation.images.head
|
| 82 |
+
- observation.images.left_wrist
|
| 83 |
+
- observation.images.right_wrist
|
| 84 |
+
sin_cos_embedding_keys: null
|
| 85 |
+
mean_std_embedding_keys: null
|
| 86 |
+
action_configs: null
|
| 87 |
+
state:
|
| 88 |
+
delta_indices:
|
| 89 |
+
- 0
|
| 90 |
+
modality_keys:
|
| 91 |
+
- body
|
| 92 |
+
sin_cos_embedding_keys: null
|
| 93 |
+
mean_std_embedding_keys: null
|
| 94 |
+
action_configs: null
|
| 95 |
+
action:
|
| 96 |
+
delta_indices:
|
| 97 |
+
- 0
|
| 98 |
+
- 1
|
| 99 |
+
- 2
|
| 100 |
+
- 3
|
| 101 |
+
- 4
|
| 102 |
+
- 5
|
| 103 |
+
- 6
|
| 104 |
+
- 7
|
| 105 |
+
- 8
|
| 106 |
+
- 9
|
| 107 |
+
- 10
|
| 108 |
+
- 11
|
| 109 |
+
- 12
|
| 110 |
+
- 13
|
| 111 |
+
- 14
|
| 112 |
+
- 15
|
| 113 |
+
- 16
|
| 114 |
+
- 17
|
| 115 |
+
- 18
|
| 116 |
+
- 19
|
| 117 |
+
- 20
|
| 118 |
+
- 21
|
| 119 |
+
- 22
|
| 120 |
+
- 23
|
| 121 |
+
- 24
|
| 122 |
+
- 25
|
| 123 |
+
- 26
|
| 124 |
+
- 27
|
| 125 |
+
- 28
|
| 126 |
+
- 29
|
| 127 |
+
- 30
|
| 128 |
+
- 31
|
| 129 |
+
- 32
|
| 130 |
+
- 33
|
| 131 |
+
- 34
|
| 132 |
+
- 35
|
| 133 |
+
- 36
|
| 134 |
+
- 37
|
| 135 |
+
- 38
|
| 136 |
+
- 39
|
| 137 |
+
modality_keys:
|
| 138 |
+
- body
|
| 139 |
+
sin_cos_embedding_keys: null
|
| 140 |
+
mean_std_embedding_keys: null
|
| 141 |
+
action_configs:
|
| 142 |
+
- rep: ABSOLUTE
|
| 143 |
+
type: NON_EEF
|
| 144 |
+
format: DEFAULT
|
| 145 |
+
state_key: null
|
| 146 |
+
language:
|
| 147 |
+
delta_indices:
|
| 148 |
+
- 0
|
| 149 |
+
modality_keys:
|
| 150 |
+
- task
|
| 151 |
+
sin_cos_embedding_keys: null
|
| 152 |
+
mean_std_embedding_keys: null
|
| 153 |
+
action_configs: null
|
| 154 |
+
download_cache: false
|
| 155 |
+
shard_size: 1024
|
| 156 |
+
episode_sampling_rate: 0.1
|
| 157 |
+
num_shards_per_epoch: 100000
|
| 158 |
+
override_pretraining_statistics: true
|
| 159 |
+
mode: single_turn
|
| 160 |
+
random_chop: 0.0
|
| 161 |
+
mock_dataset_mode: false
|
| 162 |
+
shuffle: true
|
| 163 |
+
seed: 42
|
| 164 |
+
multiprocessing_context: fork
|
| 165 |
+
allow_padding: false
|
| 166 |
+
subsample_ratio: 1.0
|
| 167 |
+
image_crop_size:
|
| 168 |
+
- 244
|
| 169 |
+
- 244
|
| 170 |
+
image_target_size:
|
| 171 |
+
- 224
|
| 172 |
+
- 224
|
| 173 |
+
video_backend: torchcodec
|
| 174 |
+
training:
|
| 175 |
+
output_dir: /home/work/mntvol/RwH-Kuavo_ckpt/autostep_v2_bs128_30k/gr00t_n17
|
| 176 |
+
experiment_name: null
|
| 177 |
+
max_steps: 30000
|
| 178 |
+
global_batch_size: 128
|
| 179 |
+
batch_size: null
|
| 180 |
+
gradient_accumulation_steps: 1
|
| 181 |
+
learning_rate: 0.0001
|
| 182 |
+
lr_scheduler_type: cosine
|
| 183 |
+
weight_decay: 1.0e-05
|
| 184 |
+
warmup_ratio: 0.05
|
| 185 |
+
warmup_steps: 0
|
| 186 |
+
max_grad_norm: 1.0
|
| 187 |
+
optim: adamw_torch
|
| 188 |
+
start_from_checkpoint: /home/work/.cache/huggingface/hub/models--nvidia--GR00T-N1.7-3B/snapshots/2fc962b973bccdd5d8ce4f67cc63b264d6886495
|
| 189 |
+
skip_weight_loading: false
|
| 190 |
+
tf32: true
|
| 191 |
+
fp16: false
|
| 192 |
+
bf16: true
|
| 193 |
+
eval_bf16: true
|
| 194 |
+
logging_steps: 100
|
| 195 |
+
save_steps: 10000
|
| 196 |
+
save_total_limit: 3
|
| 197 |
+
save_vl_model: false
|
| 198 |
+
save_only_model: false
|
| 199 |
+
upload_checkpoints: false
|
| 200 |
+
upload_every: 1000
|
| 201 |
+
upload_last_n_checkpoints: 5
|
| 202 |
+
max_concurrent_uploads: 2
|
| 203 |
+
eval_strategy: 'no'
|
| 204 |
+
eval_steps: 500
|
| 205 |
+
eval_set_split_ratio: 0.1
|
| 206 |
+
eval_batch_size: 2
|
| 207 |
+
save_best_eval_metric_name: ''
|
| 208 |
+
save_best_eval_metric_greater_is_better: true
|
| 209 |
+
deepspeed_stage: 2
|
| 210 |
+
gradient_checkpointing: false
|
| 211 |
+
transformers_trust_remote_code: true
|
| 212 |
+
transformers_local_files_only: true
|
| 213 |
+
transformers_cache_dir: null
|
| 214 |
+
transformers_access_token: null
|
| 215 |
+
use_ddp: false
|
| 216 |
+
ddp_bucket_cap_mb: 100
|
| 217 |
+
num_gpus: 1
|
| 218 |
+
dataloader_num_workers: 6
|
| 219 |
+
remove_unused_columns: false
|
| 220 |
+
use_wandb: true
|
| 221 |
+
wandb_project: kuavo
|
| 222 |
+
enable_profiling: false
|
| 223 |
+
max_retries: 3
|
| 224 |
+
assert_loss_less_than: null
|
| 225 |
+
add_rl_callback: false
|
| 226 |
+
enable_open_loop_eval: false
|
| 227 |
+
open_loop_eval_traj_ids:
|
| 228 |
+
- 0
|
| 229 |
+
open_loop_eval_steps_per_traj: 100
|
| 230 |
+
open_loop_eval_plot_indices: null
|
| 231 |
+
max_steps: 30000
|
| 232 |
+
save_steps: 10000
|
autostep_v2_bs128_30k/gr00t_n17/checkpoints/checkpoint-30K/experiment_cfg/config.yaml
ADDED
|
@@ -0,0 +1,254 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
!!python/object:gr00t.configs.base_config.Config
|
| 2 |
+
data: !!python/object:gr00t.configs.data.data_config.DataConfig
|
| 3 |
+
allow_padding: false
|
| 4 |
+
datasets:
|
| 5 |
+
- !!python/object:gr00t.configs.data.data_config.SingleDatasetConfig
|
| 6 |
+
dataset_paths:
|
| 7 |
+
- /home/work/mntvol/data/RwH-Kuavo-v2/rwh-auto-step-v2
|
| 8 |
+
dataset_type: physical_embodiment
|
| 9 |
+
embodiment_tag: new_embodiment
|
| 10 |
+
mix_ratio: 1.0
|
| 11 |
+
val_dataset_path: null
|
| 12 |
+
download_cache: false
|
| 13 |
+
episode_sampling_rate: 0.1
|
| 14 |
+
image_crop_size:
|
| 15 |
+
- 244
|
| 16 |
+
- 244
|
| 17 |
+
image_target_size:
|
| 18 |
+
- 224
|
| 19 |
+
- 224
|
| 20 |
+
mock_dataset_mode: false
|
| 21 |
+
modality_configs:
|
| 22 |
+
new_embodiment:
|
| 23 |
+
action: !!python/object:gr00t.data.types.ModalityConfig
|
| 24 |
+
action_configs:
|
| 25 |
+
- !!python/object:gr00t.data.types.ActionConfig
|
| 26 |
+
format: !!python/object/apply:gr00t.data.types.ActionFormat
|
| 27 |
+
- default
|
| 28 |
+
rep: !!python/object/apply:gr00t.data.types.ActionRepresentation
|
| 29 |
+
- absolute
|
| 30 |
+
state_key: null
|
| 31 |
+
type: !!python/object/apply:gr00t.data.types.ActionType
|
| 32 |
+
- non_eef
|
| 33 |
+
delta_indices:
|
| 34 |
+
- 0
|
| 35 |
+
- 1
|
| 36 |
+
- 2
|
| 37 |
+
- 3
|
| 38 |
+
- 4
|
| 39 |
+
- 5
|
| 40 |
+
- 6
|
| 41 |
+
- 7
|
| 42 |
+
- 8
|
| 43 |
+
- 9
|
| 44 |
+
- 10
|
| 45 |
+
- 11
|
| 46 |
+
- 12
|
| 47 |
+
- 13
|
| 48 |
+
- 14
|
| 49 |
+
- 15
|
| 50 |
+
- 16
|
| 51 |
+
- 17
|
| 52 |
+
- 18
|
| 53 |
+
- 19
|
| 54 |
+
- 20
|
| 55 |
+
- 21
|
| 56 |
+
- 22
|
| 57 |
+
- 23
|
| 58 |
+
- 24
|
| 59 |
+
- 25
|
| 60 |
+
- 26
|
| 61 |
+
- 27
|
| 62 |
+
- 28
|
| 63 |
+
- 29
|
| 64 |
+
- 30
|
| 65 |
+
- 31
|
| 66 |
+
- 32
|
| 67 |
+
- 33
|
| 68 |
+
- 34
|
| 69 |
+
- 35
|
| 70 |
+
- 36
|
| 71 |
+
- 37
|
| 72 |
+
- 38
|
| 73 |
+
- 39
|
| 74 |
+
mean_std_embedding_keys: null
|
| 75 |
+
modality_keys:
|
| 76 |
+
- body
|
| 77 |
+
sin_cos_embedding_keys: null
|
| 78 |
+
language: !!python/object:gr00t.data.types.ModalityConfig
|
| 79 |
+
action_configs: null
|
| 80 |
+
delta_indices:
|
| 81 |
+
- 0
|
| 82 |
+
mean_std_embedding_keys: null
|
| 83 |
+
modality_keys:
|
| 84 |
+
- task
|
| 85 |
+
sin_cos_embedding_keys: null
|
| 86 |
+
state: !!python/object:gr00t.data.types.ModalityConfig
|
| 87 |
+
action_configs: null
|
| 88 |
+
delta_indices:
|
| 89 |
+
- 0
|
| 90 |
+
mean_std_embedding_keys: null
|
| 91 |
+
modality_keys:
|
| 92 |
+
- body
|
| 93 |
+
sin_cos_embedding_keys: null
|
| 94 |
+
video: !!python/object:gr00t.data.types.ModalityConfig
|
| 95 |
+
action_configs: null
|
| 96 |
+
delta_indices:
|
| 97 |
+
- 0
|
| 98 |
+
mean_std_embedding_keys: null
|
| 99 |
+
modality_keys:
|
| 100 |
+
- observation.images.head
|
| 101 |
+
- observation.images.left_wrist
|
| 102 |
+
- observation.images.right_wrist
|
| 103 |
+
sin_cos_embedding_keys: null
|
| 104 |
+
mode: single_turn
|
| 105 |
+
multiprocessing_context: fork
|
| 106 |
+
num_shards_per_epoch: 100000
|
| 107 |
+
override_pretraining_statistics: true
|
| 108 |
+
random_chop: 0.0
|
| 109 |
+
seed: 42
|
| 110 |
+
shard_size: 1024
|
| 111 |
+
shuffle: true
|
| 112 |
+
subsample_ratio: 1.0
|
| 113 |
+
video_backend: torchcodec
|
| 114 |
+
load_config_path: null
|
| 115 |
+
model: !!python/object:gr00t.configs.model.gr00t_n1d7.Gr00tN1d7Config
|
| 116 |
+
_attn_implementation_internal: null
|
| 117 |
+
_commit_hash: null
|
| 118 |
+
_name_or_path: ''
|
| 119 |
+
_output_attentions: false
|
| 120 |
+
action_horizon: 40
|
| 121 |
+
add_cross_attention: false
|
| 122 |
+
architectures: null
|
| 123 |
+
backbone_trainable_params_fp32: true
|
| 124 |
+
bad_words_ids: null
|
| 125 |
+
begin_suppress_tokens: null
|
| 126 |
+
bos_token_id: null
|
| 127 |
+
chunk_size_feed_forward: 0
|
| 128 |
+
cross_attention_hidden_size: null
|
| 129 |
+
decoder_start_token_id: null
|
| 130 |
+
diffusion_model_cfg:
|
| 131 |
+
attention_head_dim: 48
|
| 132 |
+
dropout: 0.2
|
| 133 |
+
final_dropout: true
|
| 134 |
+
interleave_self_attention: true
|
| 135 |
+
norm_type: ada_norm
|
| 136 |
+
num_attention_heads: 32
|
| 137 |
+
num_layers: 16
|
| 138 |
+
output_dim: 1024
|
| 139 |
+
positional_embeddings: null
|
| 140 |
+
diversity_penalty: 0.0
|
| 141 |
+
do_sample: false
|
| 142 |
+
dtype: null
|
| 143 |
+
early_stopping: false
|
| 144 |
+
encoder_no_repeat_ngram_size: 0
|
| 145 |
+
eos_token_id: null
|
| 146 |
+
exponential_decay_length_penalty: null
|
| 147 |
+
finetuning_task: null
|
| 148 |
+
forced_bos_token_id: null
|
| 149 |
+
forced_eos_token_id: null
|
| 150 |
+
id2label:
|
| 151 |
+
0: LABEL_0
|
| 152 |
+
1: LABEL_1
|
| 153 |
+
is_decoder: false
|
| 154 |
+
is_encoder_decoder: false
|
| 155 |
+
label2id:
|
| 156 |
+
LABEL_0: 0
|
| 157 |
+
LABEL_1: 1
|
| 158 |
+
length_penalty: 1.0
|
| 159 |
+
load_bf16: true
|
| 160 |
+
max_length: 20
|
| 161 |
+
min_length: 0
|
| 162 |
+
model_name: nvidia/Cosmos-Reason2-2B
|
| 163 |
+
no_repeat_ngram_size: 0
|
| 164 |
+
num_beam_groups: 1
|
| 165 |
+
num_beams: 1
|
| 166 |
+
num_return_sequences: 1
|
| 167 |
+
output_hidden_states: false
|
| 168 |
+
output_scores: false
|
| 169 |
+
pad_token_id: null
|
| 170 |
+
prefix: null
|
| 171 |
+
problem_type: null
|
| 172 |
+
pruned_heads: {}
|
| 173 |
+
remove_invalid_values: false
|
| 174 |
+
repetition_penalty: 1.0
|
| 175 |
+
return_dict: true
|
| 176 |
+
return_dict_in_generate: false
|
| 177 |
+
sep_token_id: null
|
| 178 |
+
state_dropout_prob: 0.2
|
| 179 |
+
suppress_tokens: null
|
| 180 |
+
task_specific_params: null
|
| 181 |
+
temperature: 1.0
|
| 182 |
+
tf_legacy_loss: false
|
| 183 |
+
tie_encoder_decoder: false
|
| 184 |
+
tie_word_embeddings: true
|
| 185 |
+
tokenizer_class: null
|
| 186 |
+
top_k: 50
|
| 187 |
+
top_p: 1.0
|
| 188 |
+
torchscript: false
|
| 189 |
+
transformers_version: null
|
| 190 |
+
tune_diffusion_model: true
|
| 191 |
+
tune_llm: false
|
| 192 |
+
tune_projector: true
|
| 193 |
+
tune_visual: false
|
| 194 |
+
tune_vlln: true
|
| 195 |
+
typical_p: 1.0
|
| 196 |
+
use_bfloat16: false
|
| 197 |
+
use_relative_action: false
|
| 198 |
+
training: !!python/object:gr00t.configs.training.training_config.TrainingConfig
|
| 199 |
+
add_rl_callback: false
|
| 200 |
+
assert_loss_less_than: null
|
| 201 |
+
batch_size: null
|
| 202 |
+
bf16: true
|
| 203 |
+
dataloader_num_workers: 6
|
| 204 |
+
ddp_bucket_cap_mb: 100
|
| 205 |
+
deepspeed_stage: 2
|
| 206 |
+
enable_open_loop_eval: false
|
| 207 |
+
enable_profiling: false
|
| 208 |
+
eval_batch_size: 2
|
| 209 |
+
eval_bf16: true
|
| 210 |
+
eval_set_split_ratio: 0.1
|
| 211 |
+
eval_steps: 500
|
| 212 |
+
eval_strategy: 'no'
|
| 213 |
+
experiment_name: null
|
| 214 |
+
fp16: false
|
| 215 |
+
global_batch_size: 128
|
| 216 |
+
gradient_accumulation_steps: 1
|
| 217 |
+
gradient_checkpointing: false
|
| 218 |
+
learning_rate: 0.0001
|
| 219 |
+
logging_steps: 100
|
| 220 |
+
lr_scheduler_type: cosine
|
| 221 |
+
max_concurrent_uploads: 2
|
| 222 |
+
max_grad_norm: 1.0
|
| 223 |
+
max_retries: 3
|
| 224 |
+
max_steps: 30000
|
| 225 |
+
num_gpus: 1
|
| 226 |
+
open_loop_eval_plot_indices: null
|
| 227 |
+
open_loop_eval_steps_per_traj: 100
|
| 228 |
+
open_loop_eval_traj_ids:
|
| 229 |
+
- 0
|
| 230 |
+
optim: adamw_torch
|
| 231 |
+
output_dir: /home/work/mntvol/RwH-Kuavo_ckpt/autostep_v2_bs128_30k/gr00t_n17
|
| 232 |
+
remove_unused_columns: false
|
| 233 |
+
save_best_eval_metric_greater_is_better: true
|
| 234 |
+
save_best_eval_metric_name: ''
|
| 235 |
+
save_only_model: false
|
| 236 |
+
save_steps: 10000
|
| 237 |
+
save_total_limit: 3
|
| 238 |
+
save_vl_model: false
|
| 239 |
+
skip_weight_loading: false
|
| 240 |
+
start_from_checkpoint: /home/work/.cache/huggingface/hub/models--nvidia--GR00T-N1.7-3B/snapshots/2fc962b973bccdd5d8ce4f67cc63b264d6886495
|
| 241 |
+
tf32: true
|
| 242 |
+
transformers_access_token: null
|
| 243 |
+
transformers_cache_dir: null
|
| 244 |
+
transformers_local_files_only: true
|
| 245 |
+
transformers_trust_remote_code: true
|
| 246 |
+
upload_checkpoints: false
|
| 247 |
+
upload_every: 1000
|
| 248 |
+
upload_last_n_checkpoints: 5
|
| 249 |
+
use_ddp: false
|
| 250 |
+
use_wandb: true
|
| 251 |
+
wandb_project: kuavo
|
| 252 |
+
warmup_ratio: 0.05
|
| 253 |
+
warmup_steps: 0
|
| 254 |
+
weight_decay: 1.0e-05
|
autostep_v2_bs128_30k/gr00t_n17/checkpoints/checkpoint-30K/experiment_cfg/dataset_statistics.json
ADDED
|
@@ -0,0 +1,294 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"new_embodiment": {
|
| 3 |
+
"state": {
|
| 4 |
+
"body": {
|
| 5 |
+
"min": [
|
| 6 |
+
0.03855503723025322,
|
| 7 |
+
-1.976850986480713,
|
| 8 |
+
-0.10788518190383911,
|
| 9 |
+
0.00013844996283296496,
|
| 10 |
+
-1.930004596710205,
|
| 11 |
+
-0.33190345764160156,
|
| 12 |
+
-0.32552143931388855,
|
| 13 |
+
-2.442972183227539,
|
| 14 |
+
-1.4680404663085938,
|
| 15 |
+
-1.2988954782485962,
|
| 16 |
+
-0.00035752542316913605,
|
| 17 |
+
-1.8960524797439575,
|
| 18 |
+
-0.6130184531211853,
|
| 19 |
+
-0.12195385247468948,
|
| 20 |
+
-2.442214012145996,
|
| 21 |
+
0.7672629952430725,
|
| 22 |
+
0.00378138548694551,
|
| 23 |
+
0.00040852295933291316,
|
| 24 |
+
0.0,
|
| 25 |
+
0.0
|
| 26 |
+
],
|
| 27 |
+
"max": [
|
| 28 |
+
0.8783406019210815,
|
| 29 |
+
-0.021845076233148575,
|
| 30 |
+
1.0994032621383667,
|
| 31 |
+
0.00029990068287588656,
|
| 32 |
+
0.7006514668464661,
|
| 33 |
+
0.17508120834827423,
|
| 34 |
+
0.18718431890010834,
|
| 35 |
+
-0.30937686562538147,
|
| 36 |
+
-0.9190890192985535,
|
| 37 |
+
-0.001868014456704259,
|
| 38 |
+
0.6851634383201599,
|
| 39 |
+
0.7017965316772461,
|
| 40 |
+
0.331140398979187,
|
| 41 |
+
0.4411070644855499,
|
| 42 |
+
-0.13656805455684662,
|
| 43 |
+
1.5035191774368286,
|
| 44 |
+
1.3030930757522583,
|
| 45 |
+
0.6927947402000427,
|
| 46 |
+
0.9329638481140137,
|
| 47 |
+
0.9411078691482544
|
| 48 |
+
],
|
| 49 |
+
"mean": [
|
| 50 |
+
0.21196168186950348,
|
| 51 |
+
-0.7514868040925324,
|
| 52 |
+
0.5390654384994545,
|
| 53 |
+
0.00019716837657020593,
|
| 54 |
+
-0.7646202199218703,
|
| 55 |
+
-0.15427511241404404,
|
| 56 |
+
-0.09935649553034477,
|
| 57 |
+
-1.10718588663773,
|
| 58 |
+
-1.2239546234753425,
|
| 59 |
+
-0.7037268158015755,
|
| 60 |
+
0.4223936951295517,
|
| 61 |
+
-0.8214044226929247,
|
| 62 |
+
0.13009621515421746,
|
| 63 |
+
0.1399642855361991,
|
| 64 |
+
-0.9799681805261878,
|
| 65 |
+
1.2945560363497624,
|
| 66 |
+
0.6619886975048248,
|
| 67 |
+
0.24624616878157482,
|
| 68 |
+
0.45670749518567855,
|
| 69 |
+
0.45710977127333363
|
| 70 |
+
],
|
| 71 |
+
"std": [
|
| 72 |
+
0.16394003010843358,
|
| 73 |
+
0.3611505305466768,
|
| 74 |
+
0.23261554074762888,
|
| 75 |
+
2.1892267621294567e-05,
|
| 76 |
+
0.9256915089791073,
|
| 77 |
+
0.2060212315583477,
|
| 78 |
+
0.10228258351389044,
|
| 79 |
+
0.8544703996723307,
|
| 80 |
+
0.1634101201682666,
|
| 81 |
+
0.4491162549819308,
|
| 82 |
+
0.2717423453912273,
|
| 83 |
+
0.9605204688205453,
|
| 84 |
+
0.19753839451148422,
|
| 85 |
+
0.09507557900175964,
|
| 86 |
+
0.942503897282871,
|
| 87 |
+
0.1406902834138149,
|
| 88 |
+
0.422464227895359,
|
| 89 |
+
0.18565328903744022,
|
| 90 |
+
0.4182900296106432,
|
| 91 |
+
0.43909710239351607
|
| 92 |
+
],
|
| 93 |
+
"q01": [
|
| 94 |
+
0.039413707107305525,
|
| 95 |
+
-1.9746425795555114,
|
| 96 |
+
-0.10192423313856125,
|
| 97 |
+
0.00015722897660452873,
|
| 98 |
+
-1.8041181564331055,
|
| 99 |
+
-0.3307555043697357,
|
| 100 |
+
-0.24922363519668578,
|
| 101 |
+
-2.4425907135009766,
|
| 102 |
+
-1.4638391757011413,
|
| 103 |
+
-1.294311285018921,
|
| 104 |
+
-0.0003532645624363795,
|
| 105 |
+
-1.8342540287971496,
|
| 106 |
+
-0.176608667075634,
|
| 107 |
+
-0.11966221049427986,
|
| 108 |
+
-2.441826105117798,
|
| 109 |
+
0.7733715760707855,
|
| 110 |
+
0.004162844484671951,
|
| 111 |
+
0.0007931166712660343,
|
| 112 |
+
0.05745922774076462,
|
| 113 |
+
0.026526950299739838
|
| 114 |
+
],
|
| 115 |
+
"q99": [
|
| 116 |
+
0.8756298422813416,
|
| 117 |
+
-0.02880830265581608,
|
| 118 |
+
1.0977506637573242,
|
| 119 |
+
0.00026508700102567673,
|
| 120 |
+
0.6976009607315063,
|
| 121 |
+
0.17469997704029083,
|
| 122 |
+
0.1841490572690967,
|
| 123 |
+
-0.3097602128982544,
|
| 124 |
+
-0.9221421444416046,
|
| 125 |
+
-0.001873763487674296,
|
| 126 |
+
0.6847780346870422,
|
| 127 |
+
0.6991260051727295,
|
| 128 |
+
0.32122418999671953,
|
| 129 |
+
0.29881534099578855,
|
| 130 |
+
-0.13694968819618225,
|
| 131 |
+
1.501986985206604,
|
| 132 |
+
1.30270672082901,
|
| 133 |
+
0.6878299129009247,
|
| 134 |
+
0.9186827540397644,
|
| 135 |
+
0.9404404759407043
|
| 136 |
+
]
|
| 137 |
+
}
|
| 138 |
+
},
|
| 139 |
+
"action": {
|
| 140 |
+
"body": {
|
| 141 |
+
"min": [
|
| 142 |
+
0.03928852453827858,
|
| 143 |
+
-1.974665880203247,
|
| 144 |
+
-0.1092473641037941,
|
| 145 |
+
-3.523909386515811e-09,
|
| 146 |
+
-1.928551435470581,
|
| 147 |
+
-0.3490658402442932,
|
| 148 |
+
-0.3255215287208557,
|
| 149 |
+
-2.4434609413146973,
|
| 150 |
+
-1.4736061096191406,
|
| 151 |
+
-1.3089969158172607,
|
| 152 |
+
-0.00035663045127876103,
|
| 153 |
+
-1.8954039812088013,
|
| 154 |
+
-0.612984836101532,
|
| 155 |
+
-0.1222471222281456,
|
| 156 |
+
-2.4434609413146973,
|
| 157 |
+
0.7676474452018738,
|
| 158 |
+
0.0,
|
| 159 |
+
0.0,
|
| 160 |
+
0.0,
|
| 161 |
+
0.0,
|
| 162 |
+
-0.37572357058525085,
|
| 163 |
+
-0.3798280656337738,
|
| 164 |
+
-0.7853981852531433
|
| 165 |
+
],
|
| 166 |
+
"max": [
|
| 167 |
+
0.8758663535118103,
|
| 168 |
+
-0.024403074756264687,
|
| 169 |
+
1.0989824533462524,
|
| 170 |
+
2.085197126078242e-09,
|
| 171 |
+
0.6981316804885864,
|
| 172 |
+
0.1747012734413147,
|
| 173 |
+
0.18565650284290314,
|
| 174 |
+
-0.30905622243881226,
|
| 175 |
+
-0.917204737663269,
|
| 176 |
+
0.0,
|
| 177 |
+
0.6981316804885864,
|
| 178 |
+
0.7017980217933655,
|
| 179 |
+
0.3490658402442932,
|
| 180 |
+
0.4399641156196594,
|
| 181 |
+
-0.136188805103302,
|
| 182 |
+
1.5064144134521484,
|
| 183 |
+
1.3089969158172607,
|
| 184 |
+
0.6982267498970032,
|
| 185 |
+
1.0,
|
| 186 |
+
1.0,
|
| 187 |
+
0.37999626994132996,
|
| 188 |
+
0.29002830386161804,
|
| 189 |
+
0.6108652353286743
|
| 190 |
+
],
|
| 191 |
+
"mean": [
|
| 192 |
+
0.21199114325154228,
|
| 193 |
+
-0.7515651376243128,
|
| 194 |
+
0.5394359501007526,
|
| 195 |
+
-4.724065802401234e-11,
|
| 196 |
+
-0.7640375334495146,
|
| 197 |
+
-0.15636381779229616,
|
| 198 |
+
-0.09991505538804454,
|
| 199 |
+
-1.1060487120382727,
|
| 200 |
+
-1.2256422669537617,
|
| 201 |
+
-0.703843592485072,
|
| 202 |
+
0.4241832441010222,
|
| 203 |
+
-0.821803362483175,
|
| 204 |
+
0.1328330199614216,
|
| 205 |
+
0.1392625335852382,
|
| 206 |
+
-0.9792303322460836,
|
| 207 |
+
1.2943495028869478,
|
| 208 |
+
0.6577485896044258,
|
| 209 |
+
0.2449609850843616,
|
| 210 |
+
0.4816522577702455,
|
| 211 |
+
0.47454964992457654,
|
| 212 |
+
0.0022111699295641937,
|
| 213 |
+
-0.0006924764198898383,
|
| 214 |
+
0.007395968592941293
|
| 215 |
+
],
|
| 216 |
+
"std": [
|
| 217 |
+
0.16390777791086045,
|
| 218 |
+
0.36107212940601785,
|
| 219 |
+
0.23259818549440509,
|
| 220 |
+
3.217892825913958e-10,
|
| 221 |
+
0.9245446140013067,
|
| 222 |
+
0.20716480111407662,
|
| 223 |
+
0.10180602005088589,
|
| 224 |
+
0.8544051181680116,
|
| 225 |
+
0.16472923903734876,
|
| 226 |
+
0.4495591082747891,
|
| 227 |
+
0.27244934047338093,
|
| 228 |
+
0.9592625162743048,
|
| 229 |
+
0.19810646003690371,
|
| 230 |
+
0.09492076203273464,
|
| 231 |
+
0.9425144906293679,
|
| 232 |
+
0.14126048864876875,
|
| 233 |
+
0.42263954785045627,
|
| 234 |
+
0.18648967656964802,
|
| 235 |
+
0.47961121197640094,
|
| 236 |
+
0.4858309124429861,
|
| 237 |
+
0.07022875316058003,
|
| 238 |
+
0.049274146506525694,
|
| 239 |
+
0.15042563237977474
|
| 240 |
+
],
|
| 241 |
+
"q01": [
|
| 242 |
+
0.03963063284754753,
|
| 243 |
+
-1.9746354818344116,
|
| 244 |
+
-0.10297044366598129,
|
| 245 |
+
-1.3552186906018448e-09,
|
| 246 |
+
-1.8033545017242432,
|
| 247 |
+
-0.34313402712345126,
|
| 248 |
+
-0.2492240071296692,
|
| 249 |
+
-2.4434609413146973,
|
| 250 |
+
-1.4668941497802734,
|
| 251 |
+
-1.2981292009353638,
|
| 252 |
+
-0.00035145752597600223,
|
| 253 |
+
-1.8342512845993042,
|
| 254 |
+
-0.17660836935043334,
|
| 255 |
+
-0.11851905286312103,
|
| 256 |
+
-2.4434609413146973,
|
| 257 |
+
0.7733729481697083,
|
| 258 |
+
0.0,
|
| 259 |
+
0.0,
|
| 260 |
+
0.0,
|
| 261 |
+
0.0,
|
| 262 |
+
-0.2992435097694397,
|
| 263 |
+
-0.23223118484020233,
|
| 264 |
+
-0.7853588461875916
|
| 265 |
+
],
|
| 266 |
+
"q99": [
|
| 267 |
+
0.8756532669067383,
|
| 268 |
+
-0.028718173503875732,
|
| 269 |
+
1.0989824533462524,
|
| 270 |
+
9.750755758375362e-10,
|
| 271 |
+
0.6981316804885864,
|
| 272 |
+
0.1745329201221466,
|
| 273 |
+
0.18336854875087738,
|
| 274 |
+
-0.3101411461830139,
|
| 275 |
+
-0.920646071434021,
|
| 276 |
+
0.0,
|
| 277 |
+
0.6981316804885864,
|
| 278 |
+
0.6991260051727295,
|
| 279 |
+
0.327315012216568,
|
| 280 |
+
0.2969474792480469,
|
| 281 |
+
-0.13694968819618225,
|
| 282 |
+
1.5035138130187988,
|
| 283 |
+
1.302707314491272,
|
| 284 |
+
0.692794144153595,
|
| 285 |
+
1.0,
|
| 286 |
+
1.0,
|
| 287 |
+
0.28374183177948,
|
| 288 |
+
0.22523708641529083,
|
| 289 |
+
0.6108652353286743
|
| 290 |
+
]
|
| 291 |
+
}
|
| 292 |
+
}
|
| 293 |
+
}
|
| 294 |
+
}
|
autostep_v2_bs128_30k/gr00t_n17/checkpoints/checkpoint-30K/experiment_cfg/final_model_config.json
ADDED
|
@@ -0,0 +1,55 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"model_type": "Gr00tN1d7",
|
| 3 |
+
"model_dtype": "bfloat16",
|
| 4 |
+
"model_name": "nvidia/Cosmos-Reason2-2B",
|
| 5 |
+
"backbone_model_type": "qwen",
|
| 6 |
+
"model_revision": null,
|
| 7 |
+
"tune_top_llm_layers": 0,
|
| 8 |
+
"backbone_embedding_dim": 2048,
|
| 9 |
+
"tune_llm": false,
|
| 10 |
+
"tune_visual": false,
|
| 11 |
+
"select_layer": 16,
|
| 12 |
+
"reproject_vision": false,
|
| 13 |
+
"use_flash_attention": true,
|
| 14 |
+
"load_bf16": true,
|
| 15 |
+
"backbone_trainable_params_fp32": true,
|
| 16 |
+
"extra_augmentation_config": null,
|
| 17 |
+
"apply_sincos_state_encoding": false,
|
| 18 |
+
"use_percentiles": true,
|
| 19 |
+
"use_relative_action": false,
|
| 20 |
+
"max_state_dim": 132,
|
| 21 |
+
"max_action_dim": 132,
|
| 22 |
+
"action_horizon": 40,
|
| 23 |
+
"hidden_size": 1024,
|
| 24 |
+
"input_embedding_dim": 1536,
|
| 25 |
+
"state_history_length": 1,
|
| 26 |
+
"add_pos_embed": true,
|
| 27 |
+
"attn_dropout": 0.2,
|
| 28 |
+
"use_vlln": true,
|
| 29 |
+
"max_seq_len": 1024,
|
| 30 |
+
"use_alternate_vl_dit": true,
|
| 31 |
+
"attend_text_every_n_blocks": 2,
|
| 32 |
+
"diffusion_model_cfg": {
|
| 33 |
+
"attention_head_dim": 48,
|
| 34 |
+
"dropout": 0.2,
|
| 35 |
+
"final_dropout": true,
|
| 36 |
+
"interleave_self_attention": true,
|
| 37 |
+
"norm_type": "ada_norm",
|
| 38 |
+
"num_attention_heads": 32,
|
| 39 |
+
"num_layers": 32,
|
| 40 |
+
"output_dim": 1024,
|
| 41 |
+
"positional_embeddings": null
|
| 42 |
+
},
|
| 43 |
+
"num_inference_timesteps": 4,
|
| 44 |
+
"noise_beta_alpha": 1.5,
|
| 45 |
+
"noise_beta_beta": 1.0,
|
| 46 |
+
"noise_s": 0.999,
|
| 47 |
+
"num_timestep_buckets": 1000,
|
| 48 |
+
"tune_projector": true,
|
| 49 |
+
"tune_diffusion_model": true,
|
| 50 |
+
"tune_vlln": true,
|
| 51 |
+
"state_dropout_prob": 0.2,
|
| 52 |
+
"exclude_state": false,
|
| 53 |
+
"use_mean_std": false,
|
| 54 |
+
"max_num_embodiments": 32
|
| 55 |
+
}
|
autostep_v2_bs128_30k/gr00t_n17/checkpoints/checkpoint-30K/experiment_cfg/final_processor_config.json
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
autostep_v2_bs128_30k/gr00t_n17/checkpoints/checkpoint-30K/model-00001-of-00002.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:93d2b4a2715711b604606e57d6ab80693b02921b43f58b55bc20704945cd9b55
|
| 3 |
+
size 4993709288
|
autostep_v2_bs128_30k/gr00t_n17/checkpoints/checkpoint-30K/model-00002-of-00002.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:eba5cda753e5d7619997d69b33fc4296998defbb4ffaf089ce94e419bbe44502
|
| 3 |
+
size 4535492016
|
autostep_v2_bs128_30k/gr00t_n17/checkpoints/checkpoint-30K/model.safetensors.index.json
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
autostep_v2_bs128_30k/gr00t_n17/checkpoints/checkpoint-30K/processor_config.json
ADDED
|
@@ -0,0 +1,1140 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"processor_class": "Gr00tN1d7Processor",
|
| 3 |
+
"processor_kwargs": {
|
| 4 |
+
"modality_configs": {
|
| 5 |
+
"real_g1_relative_eef_relative_joints": {
|
| 6 |
+
"video": {
|
| 7 |
+
"delta_indices": [
|
| 8 |
+
-20,
|
| 9 |
+
0
|
| 10 |
+
],
|
| 11 |
+
"modality_keys": [
|
| 12 |
+
"ego_view"
|
| 13 |
+
],
|
| 14 |
+
"sin_cos_embedding_keys": null,
|
| 15 |
+
"mean_std_embedding_keys": null,
|
| 16 |
+
"action_configs": null
|
| 17 |
+
},
|
| 18 |
+
"state": {
|
| 19 |
+
"delta_indices": [
|
| 20 |
+
0
|
| 21 |
+
],
|
| 22 |
+
"modality_keys": [
|
| 23 |
+
"left_wrist_eef_9d",
|
| 24 |
+
"right_wrist_eef_9d",
|
| 25 |
+
"left_hand",
|
| 26 |
+
"right_hand",
|
| 27 |
+
"left_arm",
|
| 28 |
+
"right_arm",
|
| 29 |
+
"waist"
|
| 30 |
+
],
|
| 31 |
+
"sin_cos_embedding_keys": null,
|
| 32 |
+
"mean_std_embedding_keys": null,
|
| 33 |
+
"action_configs": null
|
| 34 |
+
},
|
| 35 |
+
"action": {
|
| 36 |
+
"delta_indices": [
|
| 37 |
+
0,
|
| 38 |
+
1,
|
| 39 |
+
2,
|
| 40 |
+
3,
|
| 41 |
+
4,
|
| 42 |
+
5,
|
| 43 |
+
6,
|
| 44 |
+
7,
|
| 45 |
+
8,
|
| 46 |
+
9,
|
| 47 |
+
10,
|
| 48 |
+
11,
|
| 49 |
+
12,
|
| 50 |
+
13,
|
| 51 |
+
14,
|
| 52 |
+
15,
|
| 53 |
+
16,
|
| 54 |
+
17,
|
| 55 |
+
18,
|
| 56 |
+
19,
|
| 57 |
+
20,
|
| 58 |
+
21,
|
| 59 |
+
22,
|
| 60 |
+
23,
|
| 61 |
+
24,
|
| 62 |
+
25,
|
| 63 |
+
26,
|
| 64 |
+
27,
|
| 65 |
+
28,
|
| 66 |
+
29,
|
| 67 |
+
30,
|
| 68 |
+
31,
|
| 69 |
+
32,
|
| 70 |
+
33,
|
| 71 |
+
34,
|
| 72 |
+
35,
|
| 73 |
+
36,
|
| 74 |
+
37,
|
| 75 |
+
38,
|
| 76 |
+
39
|
| 77 |
+
],
|
| 78 |
+
"modality_keys": [
|
| 79 |
+
"left_wrist_eef_9d",
|
| 80 |
+
"right_wrist_eef_9d",
|
| 81 |
+
"left_hand",
|
| 82 |
+
"right_hand",
|
| 83 |
+
"left_arm",
|
| 84 |
+
"right_arm",
|
| 85 |
+
"waist",
|
| 86 |
+
"base_height_command",
|
| 87 |
+
"navigate_command"
|
| 88 |
+
],
|
| 89 |
+
"sin_cos_embedding_keys": null,
|
| 90 |
+
"mean_std_embedding_keys": null,
|
| 91 |
+
"action_configs": [
|
| 92 |
+
{
|
| 93 |
+
"rep": "RELATIVE",
|
| 94 |
+
"type": "EEF",
|
| 95 |
+
"format": "XYZ_ROT6D",
|
| 96 |
+
"state_key": "left_wrist_eef_9d"
|
| 97 |
+
},
|
| 98 |
+
{
|
| 99 |
+
"rep": "RELATIVE",
|
| 100 |
+
"type": "EEF",
|
| 101 |
+
"format": "XYZ_ROT6D",
|
| 102 |
+
"state_key": "right_wrist_eef_9d"
|
| 103 |
+
},
|
| 104 |
+
{
|
| 105 |
+
"rep": "ABSOLUTE",
|
| 106 |
+
"type": "NON_EEF",
|
| 107 |
+
"format": "DEFAULT",
|
| 108 |
+
"state_key": "left_hand"
|
| 109 |
+
},
|
| 110 |
+
{
|
| 111 |
+
"rep": "ABSOLUTE",
|
| 112 |
+
"type": "NON_EEF",
|
| 113 |
+
"format": "DEFAULT",
|
| 114 |
+
"state_key": "right_hand"
|
| 115 |
+
},
|
| 116 |
+
{
|
| 117 |
+
"rep": "RELATIVE",
|
| 118 |
+
"type": "NON_EEF",
|
| 119 |
+
"format": "DEFAULT",
|
| 120 |
+
"state_key": "left_arm"
|
| 121 |
+
},
|
| 122 |
+
{
|
| 123 |
+
"rep": "RELATIVE",
|
| 124 |
+
"type": "NON_EEF",
|
| 125 |
+
"format": "DEFAULT",
|
| 126 |
+
"state_key": "right_arm"
|
| 127 |
+
},
|
| 128 |
+
{
|
| 129 |
+
"rep": "ABSOLUTE",
|
| 130 |
+
"type": "NON_EEF",
|
| 131 |
+
"format": "DEFAULT",
|
| 132 |
+
"state_key": "waist"
|
| 133 |
+
},
|
| 134 |
+
{
|
| 135 |
+
"rep": "ABSOLUTE",
|
| 136 |
+
"type": "NON_EEF",
|
| 137 |
+
"format": "DEFAULT",
|
| 138 |
+
"state_key": "base_height_command"
|
| 139 |
+
},
|
| 140 |
+
{
|
| 141 |
+
"rep": "ABSOLUTE",
|
| 142 |
+
"type": "NON_EEF",
|
| 143 |
+
"format": "DEFAULT",
|
| 144 |
+
"state_key": "navigate_command"
|
| 145 |
+
}
|
| 146 |
+
]
|
| 147 |
+
},
|
| 148 |
+
"language": {
|
| 149 |
+
"delta_indices": [
|
| 150 |
+
0
|
| 151 |
+
],
|
| 152 |
+
"modality_keys": [
|
| 153 |
+
"annotation.human.task_description"
|
| 154 |
+
],
|
| 155 |
+
"sin_cos_embedding_keys": null,
|
| 156 |
+
"mean_std_embedding_keys": null,
|
| 157 |
+
"action_configs": null
|
| 158 |
+
}
|
| 159 |
+
},
|
| 160 |
+
"real_r1_pro_sharpa_relative_eef_mecka": {
|
| 161 |
+
"video": {
|
| 162 |
+
"delta_indices": [
|
| 163 |
+
-30,
|
| 164 |
+
0
|
| 165 |
+
],
|
| 166 |
+
"modality_keys": [
|
| 167 |
+
"ego_view_cropratio_res320x240_freq30"
|
| 168 |
+
],
|
| 169 |
+
"sin_cos_embedding_keys": null,
|
| 170 |
+
"mean_std_embedding_keys": null,
|
| 171 |
+
"action_configs": null
|
| 172 |
+
},
|
| 173 |
+
"state": {
|
| 174 |
+
"delta_indices": [
|
| 175 |
+
0
|
| 176 |
+
],
|
| 177 |
+
"modality_keys": [
|
| 178 |
+
"left_wrist_eef",
|
| 179 |
+
"right_wrist_eef",
|
| 180 |
+
"left_hand_joints",
|
| 181 |
+
"right_hand_joints"
|
| 182 |
+
],
|
| 183 |
+
"sin_cos_embedding_keys": null,
|
| 184 |
+
"mean_std_embedding_keys": null,
|
| 185 |
+
"action_configs": null
|
| 186 |
+
},
|
| 187 |
+
"action": {
|
| 188 |
+
"delta_indices": [
|
| 189 |
+
0,
|
| 190 |
+
1,
|
| 191 |
+
2,
|
| 192 |
+
3,
|
| 193 |
+
4,
|
| 194 |
+
5,
|
| 195 |
+
6,
|
| 196 |
+
7,
|
| 197 |
+
8,
|
| 198 |
+
9,
|
| 199 |
+
10,
|
| 200 |
+
11,
|
| 201 |
+
12,
|
| 202 |
+
13,
|
| 203 |
+
14,
|
| 204 |
+
15,
|
| 205 |
+
16,
|
| 206 |
+
17,
|
| 207 |
+
18,
|
| 208 |
+
19,
|
| 209 |
+
20,
|
| 210 |
+
21,
|
| 211 |
+
22,
|
| 212 |
+
23,
|
| 213 |
+
24,
|
| 214 |
+
25,
|
| 215 |
+
26,
|
| 216 |
+
27,
|
| 217 |
+
28,
|
| 218 |
+
29,
|
| 219 |
+
30,
|
| 220 |
+
31,
|
| 221 |
+
32,
|
| 222 |
+
33,
|
| 223 |
+
34,
|
| 224 |
+
35,
|
| 225 |
+
36,
|
| 226 |
+
37,
|
| 227 |
+
38,
|
| 228 |
+
39
|
| 229 |
+
],
|
| 230 |
+
"modality_keys": [
|
| 231 |
+
"left_wrist_eef",
|
| 232 |
+
"right_wrist_eef",
|
| 233 |
+
"left_hand_joints",
|
| 234 |
+
"right_hand_joints"
|
| 235 |
+
],
|
| 236 |
+
"sin_cos_embedding_keys": null,
|
| 237 |
+
"mean_std_embedding_keys": null,
|
| 238 |
+
"action_configs": [
|
| 239 |
+
{
|
| 240 |
+
"rep": "RELATIVE",
|
| 241 |
+
"type": "EEF",
|
| 242 |
+
"format": "XYZ_ROT6D",
|
| 243 |
+
"state_key": "left_wrist_eef"
|
| 244 |
+
},
|
| 245 |
+
{
|
| 246 |
+
"rep": "RELATIVE",
|
| 247 |
+
"type": "EEF",
|
| 248 |
+
"format": "XYZ_ROT6D",
|
| 249 |
+
"state_key": "right_wrist_eef"
|
| 250 |
+
},
|
| 251 |
+
{
|
| 252 |
+
"rep": "ABSOLUTE",
|
| 253 |
+
"type": "NON_EEF",
|
| 254 |
+
"format": "DEFAULT",
|
| 255 |
+
"state_key": "left_hand_joints"
|
| 256 |
+
},
|
| 257 |
+
{
|
| 258 |
+
"rep": "ABSOLUTE",
|
| 259 |
+
"type": "NON_EEF",
|
| 260 |
+
"format": "DEFAULT",
|
| 261 |
+
"state_key": "right_hand_joints"
|
| 262 |
+
}
|
| 263 |
+
]
|
| 264 |
+
},
|
| 265 |
+
"language": {
|
| 266 |
+
"delta_indices": [
|
| 267 |
+
0
|
| 268 |
+
],
|
| 269 |
+
"modality_keys": [
|
| 270 |
+
"annotation.human.coarse_action"
|
| 271 |
+
],
|
| 272 |
+
"sin_cos_embedding_keys": null,
|
| 273 |
+
"mean_std_embedding_keys": null,
|
| 274 |
+
"action_configs": null
|
| 275 |
+
}
|
| 276 |
+
},
|
| 277 |
+
"real_r1_pro_sharpa_relative_eef_human": {
|
| 278 |
+
"video": {
|
| 279 |
+
"delta_indices": [
|
| 280 |
+
-20,
|
| 281 |
+
0
|
| 282 |
+
],
|
| 283 |
+
"modality_keys": [
|
| 284 |
+
"ego_view_res320x240_freq20",
|
| 285 |
+
"left_wrist_view_res320x240_freq20",
|
| 286 |
+
"right_wrist_view_res320x240_freq20"
|
| 287 |
+
],
|
| 288 |
+
"sin_cos_embedding_keys": null,
|
| 289 |
+
"mean_std_embedding_keys": null,
|
| 290 |
+
"action_configs": null
|
| 291 |
+
},
|
| 292 |
+
"state": {
|
| 293 |
+
"delta_indices": [
|
| 294 |
+
0
|
| 295 |
+
],
|
| 296 |
+
"modality_keys": [
|
| 297 |
+
"left_wrist_eef",
|
| 298 |
+
"right_wrist_eef",
|
| 299 |
+
"left_hand_joints",
|
| 300 |
+
"right_hand_joints"
|
| 301 |
+
],
|
| 302 |
+
"sin_cos_embedding_keys": null,
|
| 303 |
+
"mean_std_embedding_keys": null,
|
| 304 |
+
"action_configs": null
|
| 305 |
+
},
|
| 306 |
+
"action": {
|
| 307 |
+
"delta_indices": [
|
| 308 |
+
0,
|
| 309 |
+
1,
|
| 310 |
+
2,
|
| 311 |
+
3,
|
| 312 |
+
4,
|
| 313 |
+
5,
|
| 314 |
+
6,
|
| 315 |
+
7,
|
| 316 |
+
8,
|
| 317 |
+
9,
|
| 318 |
+
10,
|
| 319 |
+
11,
|
| 320 |
+
12,
|
| 321 |
+
13,
|
| 322 |
+
14,
|
| 323 |
+
15,
|
| 324 |
+
16,
|
| 325 |
+
17,
|
| 326 |
+
18,
|
| 327 |
+
19,
|
| 328 |
+
20,
|
| 329 |
+
21,
|
| 330 |
+
22,
|
| 331 |
+
23,
|
| 332 |
+
24,
|
| 333 |
+
25,
|
| 334 |
+
26,
|
| 335 |
+
27,
|
| 336 |
+
28,
|
| 337 |
+
29,
|
| 338 |
+
30,
|
| 339 |
+
31,
|
| 340 |
+
32,
|
| 341 |
+
33,
|
| 342 |
+
34,
|
| 343 |
+
35,
|
| 344 |
+
36,
|
| 345 |
+
37,
|
| 346 |
+
38,
|
| 347 |
+
39
|
| 348 |
+
],
|
| 349 |
+
"modality_keys": [
|
| 350 |
+
"left_wrist_eef",
|
| 351 |
+
"right_wrist_eef",
|
| 352 |
+
"left_hand_joints",
|
| 353 |
+
"right_hand_joints"
|
| 354 |
+
],
|
| 355 |
+
"sin_cos_embedding_keys": null,
|
| 356 |
+
"mean_std_embedding_keys": null,
|
| 357 |
+
"action_configs": [
|
| 358 |
+
{
|
| 359 |
+
"rep": "RELATIVE",
|
| 360 |
+
"type": "EEF",
|
| 361 |
+
"format": "XYZ_ROT6D",
|
| 362 |
+
"state_key": "left_wrist_eef"
|
| 363 |
+
},
|
| 364 |
+
{
|
| 365 |
+
"rep": "RELATIVE",
|
| 366 |
+
"type": "EEF",
|
| 367 |
+
"format": "XYZ_ROT6D",
|
| 368 |
+
"state_key": "right_wrist_eef"
|
| 369 |
+
},
|
| 370 |
+
{
|
| 371 |
+
"rep": "ABSOLUTE",
|
| 372 |
+
"type": "NON_EEF",
|
| 373 |
+
"format": "DEFAULT",
|
| 374 |
+
"state_key": "left_hand_joints"
|
| 375 |
+
},
|
| 376 |
+
{
|
| 377 |
+
"rep": "ABSOLUTE",
|
| 378 |
+
"type": "NON_EEF",
|
| 379 |
+
"format": "DEFAULT",
|
| 380 |
+
"state_key": "right_hand_joints"
|
| 381 |
+
}
|
| 382 |
+
]
|
| 383 |
+
},
|
| 384 |
+
"language": {
|
| 385 |
+
"delta_indices": [
|
| 386 |
+
0
|
| 387 |
+
],
|
| 388 |
+
"modality_keys": [
|
| 389 |
+
"annotation.human.coarse_action"
|
| 390 |
+
],
|
| 391 |
+
"sin_cos_embedding_keys": null,
|
| 392 |
+
"mean_std_embedding_keys": null,
|
| 393 |
+
"action_configs": null
|
| 394 |
+
}
|
| 395 |
+
},
|
| 396 |
+
"real_r1_pro_sharpa_relative_eef": {
|
| 397 |
+
"video": {
|
| 398 |
+
"delta_indices": [
|
| 399 |
+
-20,
|
| 400 |
+
0
|
| 401 |
+
],
|
| 402 |
+
"modality_keys": [
|
| 403 |
+
"ego_view_res320x240_freq20",
|
| 404 |
+
"left_wrist_view_res320x240_freq20",
|
| 405 |
+
"right_wrist_view_res320x240_freq20"
|
| 406 |
+
],
|
| 407 |
+
"sin_cos_embedding_keys": null,
|
| 408 |
+
"mean_std_embedding_keys": null,
|
| 409 |
+
"action_configs": null
|
| 410 |
+
},
|
| 411 |
+
"state": {
|
| 412 |
+
"delta_indices": [
|
| 413 |
+
0
|
| 414 |
+
],
|
| 415 |
+
"modality_keys": [
|
| 416 |
+
"left_wrist_eef",
|
| 417 |
+
"right_wrist_eef",
|
| 418 |
+
"left_hand_joints",
|
| 419 |
+
"right_hand_joints"
|
| 420 |
+
],
|
| 421 |
+
"sin_cos_embedding_keys": null,
|
| 422 |
+
"mean_std_embedding_keys": null,
|
| 423 |
+
"action_configs": null
|
| 424 |
+
},
|
| 425 |
+
"action": {
|
| 426 |
+
"delta_indices": [
|
| 427 |
+
0,
|
| 428 |
+
1,
|
| 429 |
+
2,
|
| 430 |
+
3,
|
| 431 |
+
4,
|
| 432 |
+
5,
|
| 433 |
+
6,
|
| 434 |
+
7,
|
| 435 |
+
8,
|
| 436 |
+
9,
|
| 437 |
+
10,
|
| 438 |
+
11,
|
| 439 |
+
12,
|
| 440 |
+
13,
|
| 441 |
+
14,
|
| 442 |
+
15,
|
| 443 |
+
16,
|
| 444 |
+
17,
|
| 445 |
+
18,
|
| 446 |
+
19,
|
| 447 |
+
20,
|
| 448 |
+
21,
|
| 449 |
+
22,
|
| 450 |
+
23,
|
| 451 |
+
24,
|
| 452 |
+
25,
|
| 453 |
+
26,
|
| 454 |
+
27,
|
| 455 |
+
28,
|
| 456 |
+
29,
|
| 457 |
+
30,
|
| 458 |
+
31,
|
| 459 |
+
32,
|
| 460 |
+
33,
|
| 461 |
+
34,
|
| 462 |
+
35,
|
| 463 |
+
36,
|
| 464 |
+
37,
|
| 465 |
+
38,
|
| 466 |
+
39
|
| 467 |
+
],
|
| 468 |
+
"modality_keys": [
|
| 469 |
+
"left_wrist_eef",
|
| 470 |
+
"right_wrist_eef",
|
| 471 |
+
"left_hand_joints",
|
| 472 |
+
"right_hand_joints"
|
| 473 |
+
],
|
| 474 |
+
"sin_cos_embedding_keys": null,
|
| 475 |
+
"mean_std_embedding_keys": null,
|
| 476 |
+
"action_configs": [
|
| 477 |
+
{
|
| 478 |
+
"rep": "RELATIVE",
|
| 479 |
+
"type": "EEF",
|
| 480 |
+
"format": "XYZ_ROT6D",
|
| 481 |
+
"state_key": "left_wrist_eef"
|
| 482 |
+
},
|
| 483 |
+
{
|
| 484 |
+
"rep": "RELATIVE",
|
| 485 |
+
"type": "EEF",
|
| 486 |
+
"format": "XYZ_ROT6D",
|
| 487 |
+
"state_key": "right_wrist_eef"
|
| 488 |
+
},
|
| 489 |
+
{
|
| 490 |
+
"rep": "ABSOLUTE",
|
| 491 |
+
"type": "NON_EEF",
|
| 492 |
+
"format": "DEFAULT",
|
| 493 |
+
"state_key": "left_hand_joints"
|
| 494 |
+
},
|
| 495 |
+
{
|
| 496 |
+
"rep": "ABSOLUTE",
|
| 497 |
+
"type": "NON_EEF",
|
| 498 |
+
"format": "DEFAULT",
|
| 499 |
+
"state_key": "right_hand_joints"
|
| 500 |
+
}
|
| 501 |
+
]
|
| 502 |
+
},
|
| 503 |
+
"language": {
|
| 504 |
+
"delta_indices": [
|
| 505 |
+
0
|
| 506 |
+
],
|
| 507 |
+
"modality_keys": [
|
| 508 |
+
"annotation.human.coarse_action"
|
| 509 |
+
],
|
| 510 |
+
"sin_cos_embedding_keys": null,
|
| 511 |
+
"mean_std_embedding_keys": null,
|
| 512 |
+
"action_configs": null
|
| 513 |
+
}
|
| 514 |
+
},
|
| 515 |
+
"xdof_relative_eef_relative_joint": {
|
| 516 |
+
"video": {
|
| 517 |
+
"delta_indices": [
|
| 518 |
+
-30,
|
| 519 |
+
0
|
| 520 |
+
],
|
| 521 |
+
"modality_keys": [
|
| 522 |
+
"top_camera-images-rgb_320_240",
|
| 523 |
+
"left_camera-images-rgb_320_240",
|
| 524 |
+
"right_camera-images-rgb_320_240"
|
| 525 |
+
],
|
| 526 |
+
"sin_cos_embedding_keys": null,
|
| 527 |
+
"mean_std_embedding_keys": null,
|
| 528 |
+
"action_configs": null
|
| 529 |
+
},
|
| 530 |
+
"state": {
|
| 531 |
+
"delta_indices": [
|
| 532 |
+
0
|
| 533 |
+
],
|
| 534 |
+
"modality_keys": [
|
| 535 |
+
"left_wrist_eef",
|
| 536 |
+
"right_wrist_eef",
|
| 537 |
+
"left_gripper_pos",
|
| 538 |
+
"right_gripper_pos",
|
| 539 |
+
"left_joint_pos",
|
| 540 |
+
"right_joint_pos"
|
| 541 |
+
],
|
| 542 |
+
"sin_cos_embedding_keys": null,
|
| 543 |
+
"mean_std_embedding_keys": null,
|
| 544 |
+
"action_configs": null
|
| 545 |
+
},
|
| 546 |
+
"action": {
|
| 547 |
+
"delta_indices": [
|
| 548 |
+
0,
|
| 549 |
+
1,
|
| 550 |
+
2,
|
| 551 |
+
3,
|
| 552 |
+
4,
|
| 553 |
+
5,
|
| 554 |
+
6,
|
| 555 |
+
7,
|
| 556 |
+
8,
|
| 557 |
+
9,
|
| 558 |
+
10,
|
| 559 |
+
11,
|
| 560 |
+
12,
|
| 561 |
+
13,
|
| 562 |
+
14,
|
| 563 |
+
15,
|
| 564 |
+
16,
|
| 565 |
+
17,
|
| 566 |
+
18,
|
| 567 |
+
19,
|
| 568 |
+
20,
|
| 569 |
+
21,
|
| 570 |
+
22,
|
| 571 |
+
23,
|
| 572 |
+
24,
|
| 573 |
+
25,
|
| 574 |
+
26,
|
| 575 |
+
27,
|
| 576 |
+
28,
|
| 577 |
+
29,
|
| 578 |
+
30,
|
| 579 |
+
31,
|
| 580 |
+
32,
|
| 581 |
+
33,
|
| 582 |
+
34,
|
| 583 |
+
35,
|
| 584 |
+
36,
|
| 585 |
+
37,
|
| 586 |
+
38,
|
| 587 |
+
39
|
| 588 |
+
],
|
| 589 |
+
"modality_keys": [
|
| 590 |
+
"left_wrist_eef",
|
| 591 |
+
"right_wrist_eef",
|
| 592 |
+
"left_gripper_pos",
|
| 593 |
+
"right_gripper_pos",
|
| 594 |
+
"left_joint_pos",
|
| 595 |
+
"right_joint_pos"
|
| 596 |
+
],
|
| 597 |
+
"sin_cos_embedding_keys": null,
|
| 598 |
+
"mean_std_embedding_keys": null,
|
| 599 |
+
"action_configs": [
|
| 600 |
+
{
|
| 601 |
+
"rep": "RELATIVE",
|
| 602 |
+
"type": "EEF",
|
| 603 |
+
"format": "XYZ_ROT6D",
|
| 604 |
+
"state_key": "left_wrist_eef"
|
| 605 |
+
},
|
| 606 |
+
{
|
| 607 |
+
"rep": "RELATIVE",
|
| 608 |
+
"type": "EEF",
|
| 609 |
+
"format": "XYZ_ROT6D",
|
| 610 |
+
"state_key": "right_wrist_eef"
|
| 611 |
+
},
|
| 612 |
+
{
|
| 613 |
+
"rep": "ABSOLUTE",
|
| 614 |
+
"type": "NON_EEF",
|
| 615 |
+
"format": "DEFAULT",
|
| 616 |
+
"state_key": "left_gripper_pos"
|
| 617 |
+
},
|
| 618 |
+
{
|
| 619 |
+
"rep": "ABSOLUTE",
|
| 620 |
+
"type": "NON_EEF",
|
| 621 |
+
"format": "DEFAULT",
|
| 622 |
+
"state_key": "right_gripper_pos"
|
| 623 |
+
},
|
| 624 |
+
{
|
| 625 |
+
"rep": "RELATIVE",
|
| 626 |
+
"type": "NON_EEF",
|
| 627 |
+
"format": "DEFAULT",
|
| 628 |
+
"state_key": "left_joint_pos"
|
| 629 |
+
},
|
| 630 |
+
{
|
| 631 |
+
"rep": "RELATIVE",
|
| 632 |
+
"type": "NON_EEF",
|
| 633 |
+
"format": "DEFAULT",
|
| 634 |
+
"state_key": "right_joint_pos"
|
| 635 |
+
}
|
| 636 |
+
]
|
| 637 |
+
},
|
| 638 |
+
"language": {
|
| 639 |
+
"delta_indices": [
|
| 640 |
+
0
|
| 641 |
+
],
|
| 642 |
+
"modality_keys": [
|
| 643 |
+
"annotation.task"
|
| 644 |
+
],
|
| 645 |
+
"sin_cos_embedding_keys": null,
|
| 646 |
+
"mean_std_embedding_keys": null,
|
| 647 |
+
"action_configs": null
|
| 648 |
+
}
|
| 649 |
+
},
|
| 650 |
+
"real_r1_pro_sharpa_relative_eef_maxinsights": {
|
| 651 |
+
"video": {
|
| 652 |
+
"delta_indices": [
|
| 653 |
+
-30,
|
| 654 |
+
0
|
| 655 |
+
],
|
| 656 |
+
"modality_keys": [
|
| 657 |
+
"ego_view_cropratio_res320x240_freq30"
|
| 658 |
+
],
|
| 659 |
+
"sin_cos_embedding_keys": null,
|
| 660 |
+
"mean_std_embedding_keys": null,
|
| 661 |
+
"action_configs": null
|
| 662 |
+
},
|
| 663 |
+
"state": {
|
| 664 |
+
"delta_indices": [
|
| 665 |
+
0
|
| 666 |
+
],
|
| 667 |
+
"modality_keys": [
|
| 668 |
+
"left_wrist_eef",
|
| 669 |
+
"right_wrist_eef",
|
| 670 |
+
"left_hand_joints",
|
| 671 |
+
"right_hand_joints"
|
| 672 |
+
],
|
| 673 |
+
"sin_cos_embedding_keys": null,
|
| 674 |
+
"mean_std_embedding_keys": null,
|
| 675 |
+
"action_configs": null
|
| 676 |
+
},
|
| 677 |
+
"action": {
|
| 678 |
+
"delta_indices": [
|
| 679 |
+
0,
|
| 680 |
+
1,
|
| 681 |
+
2,
|
| 682 |
+
3,
|
| 683 |
+
4,
|
| 684 |
+
5,
|
| 685 |
+
6,
|
| 686 |
+
7,
|
| 687 |
+
8,
|
| 688 |
+
9,
|
| 689 |
+
10,
|
| 690 |
+
11,
|
| 691 |
+
12,
|
| 692 |
+
13,
|
| 693 |
+
14,
|
| 694 |
+
15,
|
| 695 |
+
16,
|
| 696 |
+
17,
|
| 697 |
+
18,
|
| 698 |
+
19,
|
| 699 |
+
20,
|
| 700 |
+
21,
|
| 701 |
+
22,
|
| 702 |
+
23,
|
| 703 |
+
24,
|
| 704 |
+
25,
|
| 705 |
+
26,
|
| 706 |
+
27,
|
| 707 |
+
28,
|
| 708 |
+
29,
|
| 709 |
+
30,
|
| 710 |
+
31,
|
| 711 |
+
32,
|
| 712 |
+
33,
|
| 713 |
+
34,
|
| 714 |
+
35,
|
| 715 |
+
36,
|
| 716 |
+
37,
|
| 717 |
+
38,
|
| 718 |
+
39
|
| 719 |
+
],
|
| 720 |
+
"modality_keys": [
|
| 721 |
+
"left_wrist_eef",
|
| 722 |
+
"right_wrist_eef",
|
| 723 |
+
"left_hand_joints",
|
| 724 |
+
"right_hand_joints"
|
| 725 |
+
],
|
| 726 |
+
"sin_cos_embedding_keys": null,
|
| 727 |
+
"mean_std_embedding_keys": null,
|
| 728 |
+
"action_configs": [
|
| 729 |
+
{
|
| 730 |
+
"rep": "RELATIVE",
|
| 731 |
+
"type": "EEF",
|
| 732 |
+
"format": "XYZ_ROT6D",
|
| 733 |
+
"state_key": "left_wrist_eef"
|
| 734 |
+
},
|
| 735 |
+
{
|
| 736 |
+
"rep": "RELATIVE",
|
| 737 |
+
"type": "EEF",
|
| 738 |
+
"format": "XYZ_ROT6D",
|
| 739 |
+
"state_key": "right_wrist_eef"
|
| 740 |
+
},
|
| 741 |
+
{
|
| 742 |
+
"rep": "ABSOLUTE",
|
| 743 |
+
"type": "NON_EEF",
|
| 744 |
+
"format": "DEFAULT",
|
| 745 |
+
"state_key": "left_hand_joints"
|
| 746 |
+
},
|
| 747 |
+
{
|
| 748 |
+
"rep": "ABSOLUTE",
|
| 749 |
+
"type": "NON_EEF",
|
| 750 |
+
"format": "DEFAULT",
|
| 751 |
+
"state_key": "right_hand_joints"
|
| 752 |
+
}
|
| 753 |
+
]
|
| 754 |
+
},
|
| 755 |
+
"language": {
|
| 756 |
+
"delta_indices": [
|
| 757 |
+
0
|
| 758 |
+
],
|
| 759 |
+
"modality_keys": [
|
| 760 |
+
"annotation.human.coarse_action"
|
| 761 |
+
],
|
| 762 |
+
"sin_cos_embedding_keys": null,
|
| 763 |
+
"mean_std_embedding_keys": null,
|
| 764 |
+
"action_configs": null
|
| 765 |
+
}
|
| 766 |
+
},
|
| 767 |
+
"xdof_relative_eef_relative_joint_subtask": {
|
| 768 |
+
"video": {
|
| 769 |
+
"delta_indices": [
|
| 770 |
+
-30,
|
| 771 |
+
0
|
| 772 |
+
],
|
| 773 |
+
"modality_keys": [
|
| 774 |
+
"top_camera-images-rgb_320_240",
|
| 775 |
+
"left_camera-images-rgb_320_240",
|
| 776 |
+
"right_camera-images-rgb_320_240"
|
| 777 |
+
],
|
| 778 |
+
"sin_cos_embedding_keys": null,
|
| 779 |
+
"mean_std_embedding_keys": null,
|
| 780 |
+
"action_configs": null
|
| 781 |
+
},
|
| 782 |
+
"state": {
|
| 783 |
+
"delta_indices": [
|
| 784 |
+
0
|
| 785 |
+
],
|
| 786 |
+
"modality_keys": [
|
| 787 |
+
"left_wrist_eef",
|
| 788 |
+
"right_wrist_eef",
|
| 789 |
+
"left_gripper_pos",
|
| 790 |
+
"right_gripper_pos",
|
| 791 |
+
"left_joint_pos",
|
| 792 |
+
"right_joint_pos"
|
| 793 |
+
],
|
| 794 |
+
"sin_cos_embedding_keys": null,
|
| 795 |
+
"mean_std_embedding_keys": null,
|
| 796 |
+
"action_configs": null
|
| 797 |
+
},
|
| 798 |
+
"action": {
|
| 799 |
+
"delta_indices": [
|
| 800 |
+
0,
|
| 801 |
+
1,
|
| 802 |
+
2,
|
| 803 |
+
3,
|
| 804 |
+
4,
|
| 805 |
+
5,
|
| 806 |
+
6,
|
| 807 |
+
7,
|
| 808 |
+
8,
|
| 809 |
+
9,
|
| 810 |
+
10,
|
| 811 |
+
11,
|
| 812 |
+
12,
|
| 813 |
+
13,
|
| 814 |
+
14,
|
| 815 |
+
15,
|
| 816 |
+
16,
|
| 817 |
+
17,
|
| 818 |
+
18,
|
| 819 |
+
19,
|
| 820 |
+
20,
|
| 821 |
+
21,
|
| 822 |
+
22,
|
| 823 |
+
23,
|
| 824 |
+
24,
|
| 825 |
+
25,
|
| 826 |
+
26,
|
| 827 |
+
27,
|
| 828 |
+
28,
|
| 829 |
+
29,
|
| 830 |
+
30,
|
| 831 |
+
31,
|
| 832 |
+
32,
|
| 833 |
+
33,
|
| 834 |
+
34,
|
| 835 |
+
35,
|
| 836 |
+
36,
|
| 837 |
+
37,
|
| 838 |
+
38,
|
| 839 |
+
39
|
| 840 |
+
],
|
| 841 |
+
"modality_keys": [
|
| 842 |
+
"left_wrist_eef",
|
| 843 |
+
"right_wrist_eef",
|
| 844 |
+
"left_gripper_pos",
|
| 845 |
+
"right_gripper_pos",
|
| 846 |
+
"left_joint_pos",
|
| 847 |
+
"right_joint_pos"
|
| 848 |
+
],
|
| 849 |
+
"sin_cos_embedding_keys": null,
|
| 850 |
+
"mean_std_embedding_keys": null,
|
| 851 |
+
"action_configs": [
|
| 852 |
+
{
|
| 853 |
+
"rep": "RELATIVE",
|
| 854 |
+
"type": "EEF",
|
| 855 |
+
"format": "XYZ_ROT6D",
|
| 856 |
+
"state_key": "left_wrist_eef"
|
| 857 |
+
},
|
| 858 |
+
{
|
| 859 |
+
"rep": "RELATIVE",
|
| 860 |
+
"type": "EEF",
|
| 861 |
+
"format": "XYZ_ROT6D",
|
| 862 |
+
"state_key": "right_wrist_eef"
|
| 863 |
+
},
|
| 864 |
+
{
|
| 865 |
+
"rep": "ABSOLUTE",
|
| 866 |
+
"type": "NON_EEF",
|
| 867 |
+
"format": "DEFAULT",
|
| 868 |
+
"state_key": "left_gripper_pos"
|
| 869 |
+
},
|
| 870 |
+
{
|
| 871 |
+
"rep": "ABSOLUTE",
|
| 872 |
+
"type": "NON_EEF",
|
| 873 |
+
"format": "DEFAULT",
|
| 874 |
+
"state_key": "right_gripper_pos"
|
| 875 |
+
},
|
| 876 |
+
{
|
| 877 |
+
"rep": "RELATIVE",
|
| 878 |
+
"type": "NON_EEF",
|
| 879 |
+
"format": "DEFAULT",
|
| 880 |
+
"state_key": "left_joint_pos"
|
| 881 |
+
},
|
| 882 |
+
{
|
| 883 |
+
"rep": "RELATIVE",
|
| 884 |
+
"type": "NON_EEF",
|
| 885 |
+
"format": "DEFAULT",
|
| 886 |
+
"state_key": "right_joint_pos"
|
| 887 |
+
}
|
| 888 |
+
]
|
| 889 |
+
},
|
| 890 |
+
"language": {
|
| 891 |
+
"delta_indices": [
|
| 892 |
+
0
|
| 893 |
+
],
|
| 894 |
+
"modality_keys": [
|
| 895 |
+
"annotation.sub_task"
|
| 896 |
+
],
|
| 897 |
+
"sin_cos_embedding_keys": null,
|
| 898 |
+
"mean_std_embedding_keys": null,
|
| 899 |
+
"action_configs": null
|
| 900 |
+
}
|
| 901 |
+
},
|
| 902 |
+
"oxe_droid_relative_eef_relative_joint": {
|
| 903 |
+
"video": {
|
| 904 |
+
"delta_indices": [
|
| 905 |
+
-15,
|
| 906 |
+
0
|
| 907 |
+
],
|
| 908 |
+
"modality_keys": [
|
| 909 |
+
"exterior_image_1_left",
|
| 910 |
+
"wrist_image_left"
|
| 911 |
+
],
|
| 912 |
+
"sin_cos_embedding_keys": null,
|
| 913 |
+
"mean_std_embedding_keys": null,
|
| 914 |
+
"action_configs": null
|
| 915 |
+
},
|
| 916 |
+
"state": {
|
| 917 |
+
"delta_indices": [
|
| 918 |
+
0
|
| 919 |
+
],
|
| 920 |
+
"modality_keys": [
|
| 921 |
+
"eef_9d",
|
| 922 |
+
"gripper_position",
|
| 923 |
+
"joint_position"
|
| 924 |
+
],
|
| 925 |
+
"sin_cos_embedding_keys": null,
|
| 926 |
+
"mean_std_embedding_keys": null,
|
| 927 |
+
"action_configs": null
|
| 928 |
+
},
|
| 929 |
+
"action": {
|
| 930 |
+
"delta_indices": [
|
| 931 |
+
0,
|
| 932 |
+
1,
|
| 933 |
+
2,
|
| 934 |
+
3,
|
| 935 |
+
4,
|
| 936 |
+
5,
|
| 937 |
+
6,
|
| 938 |
+
7,
|
| 939 |
+
8,
|
| 940 |
+
9,
|
| 941 |
+
10,
|
| 942 |
+
11,
|
| 943 |
+
12,
|
| 944 |
+
13,
|
| 945 |
+
14,
|
| 946 |
+
15,
|
| 947 |
+
16,
|
| 948 |
+
17,
|
| 949 |
+
18,
|
| 950 |
+
19,
|
| 951 |
+
20,
|
| 952 |
+
21,
|
| 953 |
+
22,
|
| 954 |
+
23,
|
| 955 |
+
24,
|
| 956 |
+
25,
|
| 957 |
+
26,
|
| 958 |
+
27,
|
| 959 |
+
28,
|
| 960 |
+
29,
|
| 961 |
+
30,
|
| 962 |
+
31,
|
| 963 |
+
32,
|
| 964 |
+
33,
|
| 965 |
+
34,
|
| 966 |
+
35,
|
| 967 |
+
36,
|
| 968 |
+
37,
|
| 969 |
+
38,
|
| 970 |
+
39
|
| 971 |
+
],
|
| 972 |
+
"modality_keys": [
|
| 973 |
+
"eef_9d",
|
| 974 |
+
"gripper_position",
|
| 975 |
+
"joint_position"
|
| 976 |
+
],
|
| 977 |
+
"sin_cos_embedding_keys": null,
|
| 978 |
+
"mean_std_embedding_keys": null,
|
| 979 |
+
"action_configs": [
|
| 980 |
+
{
|
| 981 |
+
"rep": "RELATIVE",
|
| 982 |
+
"type": "EEF",
|
| 983 |
+
"format": "XYZ_ROT6D",
|
| 984 |
+
"state_key": "eef_9d"
|
| 985 |
+
},
|
| 986 |
+
{
|
| 987 |
+
"rep": "ABSOLUTE",
|
| 988 |
+
"type": "NON_EEF",
|
| 989 |
+
"format": "DEFAULT",
|
| 990 |
+
"state_key": "gripper_position"
|
| 991 |
+
},
|
| 992 |
+
{
|
| 993 |
+
"rep": "RELATIVE",
|
| 994 |
+
"type": "NON_EEF",
|
| 995 |
+
"format": "DEFAULT",
|
| 996 |
+
"state_key": "joint_position"
|
| 997 |
+
}
|
| 998 |
+
]
|
| 999 |
+
},
|
| 1000 |
+
"language": {
|
| 1001 |
+
"delta_indices": [
|
| 1002 |
+
0
|
| 1003 |
+
],
|
| 1004 |
+
"modality_keys": [
|
| 1005 |
+
"annotation.language.language_instruction"
|
| 1006 |
+
],
|
| 1007 |
+
"sin_cos_embedding_keys": null,
|
| 1008 |
+
"mean_std_embedding_keys": null,
|
| 1009 |
+
"action_configs": null
|
| 1010 |
+
}
|
| 1011 |
+
},
|
| 1012 |
+
"new_embodiment": {
|
| 1013 |
+
"video": {
|
| 1014 |
+
"delta_indices": [
|
| 1015 |
+
0
|
| 1016 |
+
],
|
| 1017 |
+
"modality_keys": [
|
| 1018 |
+
"observation.images.head",
|
| 1019 |
+
"observation.images.left_wrist",
|
| 1020 |
+
"observation.images.right_wrist"
|
| 1021 |
+
],
|
| 1022 |
+
"sin_cos_embedding_keys": null,
|
| 1023 |
+
"mean_std_embedding_keys": null,
|
| 1024 |
+
"action_configs": null
|
| 1025 |
+
},
|
| 1026 |
+
"state": {
|
| 1027 |
+
"delta_indices": [
|
| 1028 |
+
0
|
| 1029 |
+
],
|
| 1030 |
+
"modality_keys": [
|
| 1031 |
+
"body"
|
| 1032 |
+
],
|
| 1033 |
+
"sin_cos_embedding_keys": null,
|
| 1034 |
+
"mean_std_embedding_keys": null,
|
| 1035 |
+
"action_configs": null
|
| 1036 |
+
},
|
| 1037 |
+
"action": {
|
| 1038 |
+
"delta_indices": [
|
| 1039 |
+
0,
|
| 1040 |
+
1,
|
| 1041 |
+
2,
|
| 1042 |
+
3,
|
| 1043 |
+
4,
|
| 1044 |
+
5,
|
| 1045 |
+
6,
|
| 1046 |
+
7,
|
| 1047 |
+
8,
|
| 1048 |
+
9,
|
| 1049 |
+
10,
|
| 1050 |
+
11,
|
| 1051 |
+
12,
|
| 1052 |
+
13,
|
| 1053 |
+
14,
|
| 1054 |
+
15,
|
| 1055 |
+
16,
|
| 1056 |
+
17,
|
| 1057 |
+
18,
|
| 1058 |
+
19,
|
| 1059 |
+
20,
|
| 1060 |
+
21,
|
| 1061 |
+
22,
|
| 1062 |
+
23,
|
| 1063 |
+
24,
|
| 1064 |
+
25,
|
| 1065 |
+
26,
|
| 1066 |
+
27,
|
| 1067 |
+
28,
|
| 1068 |
+
29,
|
| 1069 |
+
30,
|
| 1070 |
+
31,
|
| 1071 |
+
32,
|
| 1072 |
+
33,
|
| 1073 |
+
34,
|
| 1074 |
+
35,
|
| 1075 |
+
36,
|
| 1076 |
+
37,
|
| 1077 |
+
38,
|
| 1078 |
+
39
|
| 1079 |
+
],
|
| 1080 |
+
"modality_keys": [
|
| 1081 |
+
"body"
|
| 1082 |
+
],
|
| 1083 |
+
"sin_cos_embedding_keys": null,
|
| 1084 |
+
"mean_std_embedding_keys": null,
|
| 1085 |
+
"action_configs": [
|
| 1086 |
+
{
|
| 1087 |
+
"rep": "ABSOLUTE",
|
| 1088 |
+
"type": "NON_EEF",
|
| 1089 |
+
"format": "DEFAULT",
|
| 1090 |
+
"state_key": null
|
| 1091 |
+
}
|
| 1092 |
+
]
|
| 1093 |
+
},
|
| 1094 |
+
"language": {
|
| 1095 |
+
"delta_indices": [
|
| 1096 |
+
0
|
| 1097 |
+
],
|
| 1098 |
+
"modality_keys": [
|
| 1099 |
+
"task"
|
| 1100 |
+
],
|
| 1101 |
+
"sin_cos_embedding_keys": null,
|
| 1102 |
+
"mean_std_embedding_keys": null,
|
| 1103 |
+
"action_configs": null
|
| 1104 |
+
}
|
| 1105 |
+
}
|
| 1106 |
+
},
|
| 1107 |
+
"image_crop_size": [
|
| 1108 |
+
230,
|
| 1109 |
+
230
|
| 1110 |
+
],
|
| 1111 |
+
"image_target_size": [
|
| 1112 |
+
256,
|
| 1113 |
+
256
|
| 1114 |
+
],
|
| 1115 |
+
"use_albumentations": true,
|
| 1116 |
+
"random_rotation_angle": 0,
|
| 1117 |
+
"color_jitter_params": {
|
| 1118 |
+
"brightness": 0.3,
|
| 1119 |
+
"contrast": 0.4,
|
| 1120 |
+
"saturation": 0.5,
|
| 1121 |
+
"hue": 0.08
|
| 1122 |
+
},
|
| 1123 |
+
"shortest_image_edge": 256,
|
| 1124 |
+
"crop_fraction": 0.95,
|
| 1125 |
+
"letter_box_transform": false,
|
| 1126 |
+
"model_name": "nvidia/Cosmos-Reason2-2B",
|
| 1127 |
+
"model_type": "qwen",
|
| 1128 |
+
"formalize_language": true,
|
| 1129 |
+
"max_state_dim": 132,
|
| 1130 |
+
"max_action_dim": 132,
|
| 1131 |
+
"max_action_horizon": 40,
|
| 1132 |
+
"use_percentiles": true,
|
| 1133 |
+
"use_mean_std": false,
|
| 1134 |
+
"clip_outliers": true,
|
| 1135 |
+
"apply_sincos_state_encoding": false,
|
| 1136 |
+
"use_relative_action": false,
|
| 1137 |
+
"exclude_state": false,
|
| 1138 |
+
"state_dropout_prob": 0.2
|
| 1139 |
+
}
|
| 1140 |
+
}
|
autostep_v2_bs128_30k/gr00t_n17/checkpoints/checkpoint-30K/provenance/configs/Bs128_30k/data/dataset_autostep_v2.yaml
ADDED
|
@@ -0,0 +1,31 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
source:
|
| 2 |
+
repo_id: RwH-Lab/rwh-auto-step-v2
|
| 3 |
+
revision: 7ccc8d62be1d6ff997267dcf55b831cbe865eedd
|
| 4 |
+
root: /home/work/mntvol/data/RwH-Kuavo-v2/rwh-auto-step-v2
|
| 5 |
+
sampling:
|
| 6 |
+
source_fps: 30
|
| 7 |
+
fps: 10
|
| 8 |
+
split_at_task_change: true
|
| 9 |
+
state:
|
| 10 |
+
source: observation.state
|
| 11 |
+
indices:
|
| 12 |
+
- 0
|
| 13 |
+
- 20
|
| 14 |
+
invalid_stored_zero: keep_as_input
|
| 15 |
+
action:
|
| 16 |
+
source: action
|
| 17 |
+
indices:
|
| 18 |
+
- 0
|
| 19 |
+
- 23
|
| 20 |
+
validity_feature: validity.action
|
| 21 |
+
invalid: mask_loss
|
| 22 |
+
statistics:
|
| 23 |
+
state: all_inputs
|
| 24 |
+
action: valid_targets
|
| 25 |
+
video:
|
| 26 |
+
mode: original
|
| 27 |
+
alignment: timestamp
|
| 28 |
+
adapters:
|
| 29 |
+
gr00t_n17:
|
| 30 |
+
enabled: true
|
| 31 |
+
type: letools_native_v3
|
autostep_v2_bs128_30k/gr00t_n17/checkpoints/checkpoint-30K/provenance/configs/Bs128_30k/train/gr00t_n17_autostep_v2.yaml
ADDED
|
@@ -0,0 +1,57 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
dataset:
|
| 2 |
+
config: ../data/dataset_autostep_v2.yaml
|
| 3 |
+
training:
|
| 4 |
+
resume: false
|
| 5 |
+
output_dir: /home/work/mntvol/RwH-Kuavo_ckpt/autostep_v2_bs128_30k/gr00t_n17
|
| 6 |
+
job_name: autostep_v2_bs128_30k_gr00t_n17
|
| 7 |
+
batch_size: 128
|
| 8 |
+
global_batch_size: 128
|
| 9 |
+
gradient_accumulation_steps: 1
|
| 10 |
+
steps: 30000
|
| 11 |
+
save_freq: 10000
|
| 12 |
+
log_freq: 100
|
| 13 |
+
num_workers: 6
|
| 14 |
+
seed: 42
|
| 15 |
+
optimizer:
|
| 16 |
+
lr: 0.0001
|
| 17 |
+
betas:
|
| 18 |
+
- 0.9
|
| 19 |
+
- 0.999
|
| 20 |
+
weight_decay: 1.0e-05
|
| 21 |
+
grad_clip_norm: 1.0
|
| 22 |
+
scheduler:
|
| 23 |
+
type: cosine
|
| 24 |
+
warmup_ratio: 0.05
|
| 25 |
+
decay_lr: 0.0
|
| 26 |
+
shard_size: 1024
|
| 27 |
+
episode_sampling_rate: 0.1
|
| 28 |
+
num_shards_per_epoch: 100000
|
| 29 |
+
policy:
|
| 30 |
+
type: groot_n17
|
| 31 |
+
base_model_path: nvidia/GR00T-N1.7-3B
|
| 32 |
+
base_revision: 2fc962b973bccdd5d8ce4f67cc63b264d6886495
|
| 33 |
+
runtime_root: /home/work/workspace/RwH-Kuavo_V2/LeTools-Learning/kuavo_model/external_models/gr00tn1d7
|
| 34 |
+
embodiment_tag: new_embodiment
|
| 35 |
+
chunk_size: 40
|
| 36 |
+
n_action_steps: 40
|
| 37 |
+
max_state_dim: 132
|
| 38 |
+
max_action_dim: 132
|
| 39 |
+
tune_llm: false
|
| 40 |
+
tune_visual: false
|
| 41 |
+
tune_projector: true
|
| 42 |
+
tune_diffusion_model: true
|
| 43 |
+
tune_vlln: true
|
| 44 |
+
use_bf16: true
|
| 45 |
+
use_relative_action: false
|
| 46 |
+
state_dropout_prob: 0.2
|
| 47 |
+
loss:
|
| 48 |
+
validity_feature: validity.action
|
| 49 |
+
dimension_mask: true
|
| 50 |
+
exclude_action_padding: true
|
| 51 |
+
reduction: sum_valid_elements_div_valid_count
|
| 52 |
+
wandb:
|
| 53 |
+
enable: true
|
| 54 |
+
project: kuavo
|
| 55 |
+
entity: minje227_hyu-hanyang-university
|
| 56 |
+
mode: online
|
| 57 |
+
disable_artifact: true
|
autostep_v2_bs128_30k/gr00t_n17/checkpoints/checkpoint-30K/provenance/configs/Bs128_30k/train/launch.sh
ADDED
|
@@ -0,0 +1,18 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
#!/usr/bin/env bash
|
| 2 |
+
source /home/work/workspace/session-bootstrap.sh >/dev/null
|
| 3 |
+
set -euo pipefail
|
| 4 |
+
task_dir="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
|
| 5 |
+
project="$(cd "$task_dir/../../.." && pwd)"
|
| 6 |
+
export RWH_TRAIN_CONFIG_DIR="$task_dir"
|
| 7 |
+
export PYTHONPATH="$project:$project/../LeTools-Learning/third_party/lerobot/src"
|
| 8 |
+
export WANDB_API_KEY="$(cat /home/work/.local/share/wandb/kuavo_api_key)"
|
| 9 |
+
export WANDB__SERVICE_WAIT=120
|
| 10 |
+
export TOKENIZERS_PARALLELISM=false OMP_NUM_THREADS=8 MKL_NUM_THREADS=8 HF_HUB_OFFLINE=1 TRANSFORMERS_OFFLINE=1
|
| 11 |
+
policy="$1"
|
| 12 |
+
shift
|
| 13 |
+
cd "$project"
|
| 14 |
+
if [ "$policy" = gr00t_n17 ]; then
|
| 15 |
+
exec /home/work/.local/miniconda3/envs/rwh_n17_letools/bin/python -m scripts.train.run_n17 "$@"
|
| 16 |
+
else
|
| 17 |
+
exec /home/work/.local/miniconda3/envs/rwh/bin/python -m scripts.train.run_lerobot "$policy" "$@"
|
| 18 |
+
fi
|
autostep_v2_bs128_30k/gr00t_n17/checkpoints/checkpoint-30K/provenance/data_audit/data_manifest.json
ADDED
|
@@ -0,0 +1,56 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"fingerprint": "81f3787f08f2a596f6f22aa087eb414d27264756b661467a3a88b59495401888",
|
| 3 |
+
"config": {
|
| 4 |
+
"source": {
|
| 5 |
+
"repo_id": "RwH-Lab/rwh-auto-step-v2",
|
| 6 |
+
"revision": "7ccc8d62be1d6ff997267dcf55b831cbe865eedd",
|
| 7 |
+
"root": "/home/work/mntvol/data/RwH-Kuavo-v2/rwh-auto-step-v2"
|
| 8 |
+
},
|
| 9 |
+
"sampling": {
|
| 10 |
+
"source_fps": 30,
|
| 11 |
+
"fps": 10,
|
| 12 |
+
"split_at_task_change": true
|
| 13 |
+
},
|
| 14 |
+
"state": {
|
| 15 |
+
"source": "observation.state",
|
| 16 |
+
"indices": [
|
| 17 |
+
0,
|
| 18 |
+
20
|
| 19 |
+
],
|
| 20 |
+
"invalid_stored_zero": "keep_as_input"
|
| 21 |
+
},
|
| 22 |
+
"action": {
|
| 23 |
+
"source": "action",
|
| 24 |
+
"indices": [
|
| 25 |
+
0,
|
| 26 |
+
23
|
| 27 |
+
],
|
| 28 |
+
"validity_feature": "validity.action",
|
| 29 |
+
"invalid": "mask_loss"
|
| 30 |
+
},
|
| 31 |
+
"statistics": {
|
| 32 |
+
"state": "all_inputs",
|
| 33 |
+
"action": "valid_targets"
|
| 34 |
+
},
|
| 35 |
+
"video": {
|
| 36 |
+
"mode": "original",
|
| 37 |
+
"alignment": "timestamp"
|
| 38 |
+
},
|
| 39 |
+
"adapters": {
|
| 40 |
+
"gr00t_n17": {
|
| 41 |
+
"enabled": true,
|
| 42 |
+
"type": "letools_native_v3"
|
| 43 |
+
}
|
| 44 |
+
}
|
| 45 |
+
},
|
| 46 |
+
"source_rows": 72256,
|
| 47 |
+
"selected_rows": 24099,
|
| 48 |
+
"task_episodes": 170,
|
| 49 |
+
"eligible_anchors": {
|
| 50 |
+
"gr00t_n15": 21549,
|
| 51 |
+
"gr00t_n17": 17469,
|
| 52 |
+
"multi_task_dit": 18829
|
| 53 |
+
},
|
| 54 |
+
"invalid_action_elements": 12,
|
| 55 |
+
"original_torso": true
|
| 56 |
+
}
|
autostep_v2_bs128_30k/gr00t_n17/checkpoints/checkpoint-30K/provenance/data_audit/statistics.json
ADDED
|
@@ -0,0 +1,335 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"observation.state": {
|
| 3 |
+
"mean": [
|
| 4 |
+
0.21196168186950348,
|
| 5 |
+
-0.7514868040925324,
|
| 6 |
+
0.5390654384994545,
|
| 7 |
+
0.00019716837657020593,
|
| 8 |
+
-0.7646202199218703,
|
| 9 |
+
-0.15427511241404404,
|
| 10 |
+
-0.09935649553034477,
|
| 11 |
+
-1.10718588663773,
|
| 12 |
+
-1.2239546234753425,
|
| 13 |
+
-0.7037268158015755,
|
| 14 |
+
0.4223936951295517,
|
| 15 |
+
-0.8214044226929247,
|
| 16 |
+
0.13009621515421746,
|
| 17 |
+
0.1399642855361991,
|
| 18 |
+
-0.9799681805261878,
|
| 19 |
+
1.2945560363497624,
|
| 20 |
+
0.6619886975048248,
|
| 21 |
+
0.24624616878157482,
|
| 22 |
+
0.45670749518567855,
|
| 23 |
+
0.45710977127333363
|
| 24 |
+
],
|
| 25 |
+
"std": [
|
| 26 |
+
0.16394003010843358,
|
| 27 |
+
0.3611505305466768,
|
| 28 |
+
0.23261554074762886,
|
| 29 |
+
2.1892267621294564e-05,
|
| 30 |
+
0.9256915089791072,
|
| 31 |
+
0.2060212315583477,
|
| 32 |
+
0.10228258351389043,
|
| 33 |
+
0.8544703996723307,
|
| 34 |
+
0.16341012016826634,
|
| 35 |
+
0.44911625498193086,
|
| 36 |
+
0.2717423453912273,
|
| 37 |
+
0.9605204688205453,
|
| 38 |
+
0.19753839451148422,
|
| 39 |
+
0.09507557900175964,
|
| 40 |
+
0.942503897282871,
|
| 41 |
+
0.1406902834138152,
|
| 42 |
+
0.42246422789535903,
|
| 43 |
+
0.18565328903744022,
|
| 44 |
+
0.4182900296106432,
|
| 45 |
+
0.43909710239351607
|
| 46 |
+
],
|
| 47 |
+
"min": [
|
| 48 |
+
0.03855503723025322,
|
| 49 |
+
-1.976850986480713,
|
| 50 |
+
-0.10788518190383911,
|
| 51 |
+
0.00013844996283296496,
|
| 52 |
+
-1.930004596710205,
|
| 53 |
+
-0.33190345764160156,
|
| 54 |
+
-0.32552143931388855,
|
| 55 |
+
-2.442972183227539,
|
| 56 |
+
-1.4680404663085938,
|
| 57 |
+
-1.2988954782485962,
|
| 58 |
+
-0.00035752542316913605,
|
| 59 |
+
-1.8960524797439575,
|
| 60 |
+
-0.6130184531211853,
|
| 61 |
+
-0.12195385247468948,
|
| 62 |
+
-2.442214012145996,
|
| 63 |
+
0.7672629952430725,
|
| 64 |
+
0.00378138548694551,
|
| 65 |
+
0.00040852295933291316,
|
| 66 |
+
0.0,
|
| 67 |
+
0.0
|
| 68 |
+
],
|
| 69 |
+
"max": [
|
| 70 |
+
0.8783406019210815,
|
| 71 |
+
-0.021845076233148575,
|
| 72 |
+
1.0994032621383667,
|
| 73 |
+
0.00029990068287588656,
|
| 74 |
+
0.7006514668464661,
|
| 75 |
+
0.17508120834827423,
|
| 76 |
+
0.18718431890010834,
|
| 77 |
+
-0.30937686562538147,
|
| 78 |
+
-0.9190890192985535,
|
| 79 |
+
-0.001868014456704259,
|
| 80 |
+
0.6851634383201599,
|
| 81 |
+
0.7017965316772461,
|
| 82 |
+
0.331140398979187,
|
| 83 |
+
0.4411070644855499,
|
| 84 |
+
-0.13656805455684662,
|
| 85 |
+
1.5035191774368286,
|
| 86 |
+
1.3030930757522583,
|
| 87 |
+
0.6927947402000427,
|
| 88 |
+
0.9329638481140137,
|
| 89 |
+
0.9411078691482544
|
| 90 |
+
],
|
| 91 |
+
"q01": [
|
| 92 |
+
0.039413705468177795,
|
| 93 |
+
-1.9746426343917847,
|
| 94 |
+
-0.10192423313856125,
|
| 95 |
+
0.00015722897660452873,
|
| 96 |
+
-1.8041181564331055,
|
| 97 |
+
-0.3307555019855499,
|
| 98 |
+
-0.24922363460063934,
|
| 99 |
+
-2.4425907135009766,
|
| 100 |
+
-1.4638391733169556,
|
| 101 |
+
-1.294311285018921,
|
| 102 |
+
-0.0003532645641826093,
|
| 103 |
+
-1.8342540264129639,
|
| 104 |
+
-0.17660866677761078,
|
| 105 |
+
-0.11966221034526825,
|
| 106 |
+
-2.441826105117798,
|
| 107 |
+
0.7733715772628784,
|
| 108 |
+
0.004162844270467758,
|
| 109 |
+
0.000793116691056639,
|
| 110 |
+
0.05745922774076462,
|
| 111 |
+
0.026526950299739838
|
| 112 |
+
],
|
| 113 |
+
"q99": [
|
| 114 |
+
0.8756298422813416,
|
| 115 |
+
-0.02880830317735672,
|
| 116 |
+
1.0977506637573242,
|
| 117 |
+
0.00026508700102567673,
|
| 118 |
+
0.6976009607315063,
|
| 119 |
+
0.17469997704029083,
|
| 120 |
+
0.18414869904518127,
|
| 121 |
+
-0.3097602128982544,
|
| 122 |
+
-0.9221421480178833,
|
| 123 |
+
-0.0018737635109573603,
|
| 124 |
+
0.6847780346870422,
|
| 125 |
+
0.6991260051727295,
|
| 126 |
+
0.32122400403022766,
|
| 127 |
+
0.2988153398036957,
|
| 128 |
+
-0.13694968819618225,
|
| 129 |
+
1.5019869804382324,
|
| 130 |
+
1.3027067184448242,
|
| 131 |
+
0.6878299117088318,
|
| 132 |
+
0.9186827540397644,
|
| 133 |
+
0.9404404759407043
|
| 134 |
+
],
|
| 135 |
+
"count": [
|
| 136 |
+
24099,
|
| 137 |
+
24099,
|
| 138 |
+
24099,
|
| 139 |
+
24099,
|
| 140 |
+
24099,
|
| 141 |
+
24099,
|
| 142 |
+
24099,
|
| 143 |
+
24099,
|
| 144 |
+
24099,
|
| 145 |
+
24099,
|
| 146 |
+
24099,
|
| 147 |
+
24099,
|
| 148 |
+
24099,
|
| 149 |
+
24099,
|
| 150 |
+
24099,
|
| 151 |
+
24099,
|
| 152 |
+
24099,
|
| 153 |
+
24099,
|
| 154 |
+
24099,
|
| 155 |
+
24099
|
| 156 |
+
]
|
| 157 |
+
},
|
| 158 |
+
"action": {
|
| 159 |
+
"mean": [
|
| 160 |
+
0.21199114325154228,
|
| 161 |
+
-0.7515651376243128,
|
| 162 |
+
0.5394359501007526,
|
| 163 |
+
-4.724065802401234e-11,
|
| 164 |
+
-0.7640375334495146,
|
| 165 |
+
-0.15636381779229616,
|
| 166 |
+
-0.09991505538804454,
|
| 167 |
+
-1.1060487120382727,
|
| 168 |
+
-1.2256422669537617,
|
| 169 |
+
-0.703843592485072,
|
| 170 |
+
0.4241832441010222,
|
| 171 |
+
-0.821803362483175,
|
| 172 |
+
0.1328330199614216,
|
| 173 |
+
0.1392625335852382,
|
| 174 |
+
-0.9792303322460836,
|
| 175 |
+
1.2943495028869478,
|
| 176 |
+
0.6577485896044258,
|
| 177 |
+
0.2449609850843616,
|
| 178 |
+
0.4816522577702455,
|
| 179 |
+
0.47454964992457654,
|
| 180 |
+
0.0022111699295641937,
|
| 181 |
+
-0.0006924764198898383,
|
| 182 |
+
0.007395968592941293
|
| 183 |
+
],
|
| 184 |
+
"std": [
|
| 185 |
+
0.16390777791086045,
|
| 186 |
+
0.3610721294060178,
|
| 187 |
+
0.23259818549440509,
|
| 188 |
+
3.217892825913958e-10,
|
| 189 |
+
0.9245446140013067,
|
| 190 |
+
0.20716480111407665,
|
| 191 |
+
0.10180602005088588,
|
| 192 |
+
0.8544051181680116,
|
| 193 |
+
0.16472923903734873,
|
| 194 |
+
0.4495591082747891,
|
| 195 |
+
0.2724493404733809,
|
| 196 |
+
0.9592625162743048,
|
| 197 |
+
0.19810646003690374,
|
| 198 |
+
0.09492076203273465,
|
| 199 |
+
0.9425144906293679,
|
| 200 |
+
0.141260488648769,
|
| 201 |
+
0.4226395478504563,
|
| 202 |
+
0.18648967656964804,
|
| 203 |
+
0.47961121197640094,
|
| 204 |
+
0.48583091244298615,
|
| 205 |
+
0.07022875316058003,
|
| 206 |
+
0.049274146506525694,
|
| 207 |
+
0.15042563237977474
|
| 208 |
+
],
|
| 209 |
+
"min": [
|
| 210 |
+
0.03928852453827858,
|
| 211 |
+
-1.974665880203247,
|
| 212 |
+
-0.1092473641037941,
|
| 213 |
+
-3.523909386515811e-09,
|
| 214 |
+
-1.928551435470581,
|
| 215 |
+
-0.3490658402442932,
|
| 216 |
+
-0.3255215287208557,
|
| 217 |
+
-2.4434609413146973,
|
| 218 |
+
-1.4736061096191406,
|
| 219 |
+
-1.3089969158172607,
|
| 220 |
+
-0.00035663045127876103,
|
| 221 |
+
-1.8954039812088013,
|
| 222 |
+
-0.612984836101532,
|
| 223 |
+
-0.1222471222281456,
|
| 224 |
+
-2.4434609413146973,
|
| 225 |
+
0.7676474452018738,
|
| 226 |
+
0.0,
|
| 227 |
+
0.0,
|
| 228 |
+
0.0,
|
| 229 |
+
0.0,
|
| 230 |
+
-0.37572357058525085,
|
| 231 |
+
-0.3798280656337738,
|
| 232 |
+
-0.7853981852531433
|
| 233 |
+
],
|
| 234 |
+
"max": [
|
| 235 |
+
0.8758663535118103,
|
| 236 |
+
-0.024403074756264687,
|
| 237 |
+
1.0989824533462524,
|
| 238 |
+
2.085197126078242e-09,
|
| 239 |
+
0.6981316804885864,
|
| 240 |
+
0.1747012734413147,
|
| 241 |
+
0.18565650284290314,
|
| 242 |
+
-0.30905622243881226,
|
| 243 |
+
-0.917204737663269,
|
| 244 |
+
0.0,
|
| 245 |
+
0.6981316804885864,
|
| 246 |
+
0.7017980217933655,
|
| 247 |
+
0.3490658402442932,
|
| 248 |
+
0.4399641156196594,
|
| 249 |
+
-0.136188805103302,
|
| 250 |
+
1.5064144134521484,
|
| 251 |
+
1.3089969158172607,
|
| 252 |
+
0.6982267498970032,
|
| 253 |
+
1.0,
|
| 254 |
+
1.0,
|
| 255 |
+
0.37999626994132996,
|
| 256 |
+
0.29002830386161804,
|
| 257 |
+
0.6108652353286743
|
| 258 |
+
],
|
| 259 |
+
"q01": [
|
| 260 |
+
0.03963063284754753,
|
| 261 |
+
-1.9746354818344116,
|
| 262 |
+
-0.10297044366598129,
|
| 263 |
+
-1.3552187150267514e-09,
|
| 264 |
+
-1.8033545017242432,
|
| 265 |
+
-0.3431340157985687,
|
| 266 |
+
-0.2492240071296692,
|
| 267 |
+
-2.4434609413146973,
|
| 268 |
+
-1.4668941497802734,
|
| 269 |
+
-1.2981292009353638,
|
| 270 |
+
-0.0003514575364533812,
|
| 271 |
+
-1.8342512845993042,
|
| 272 |
+
-0.1766083687543869,
|
| 273 |
+
-0.11851905286312103,
|
| 274 |
+
-2.4434609413146973,
|
| 275 |
+
0.7733729481697083,
|
| 276 |
+
0.0,
|
| 277 |
+
0.0,
|
| 278 |
+
0.0,
|
| 279 |
+
0.0,
|
| 280 |
+
-0.2992435097694397,
|
| 281 |
+
-0.23223118484020233,
|
| 282 |
+
-0.7853588461875916
|
| 283 |
+
],
|
| 284 |
+
"q99": [
|
| 285 |
+
0.8756532669067383,
|
| 286 |
+
-0.028718173503875732,
|
| 287 |
+
1.0989824533462524,
|
| 288 |
+
9.750755758375362e-10,
|
| 289 |
+
0.6981316804885864,
|
| 290 |
+
0.1745329201221466,
|
| 291 |
+
0.18336854875087738,
|
| 292 |
+
-0.3101411461830139,
|
| 293 |
+
-0.920646071434021,
|
| 294 |
+
0.0,
|
| 295 |
+
0.6981316804885864,
|
| 296 |
+
0.6991260051727295,
|
| 297 |
+
0.32731497287750244,
|
| 298 |
+
0.2969474792480469,
|
| 299 |
+
-0.13694968819618225,
|
| 300 |
+
1.5035138130187988,
|
| 301 |
+
1.302707314491272,
|
| 302 |
+
0.692794144153595,
|
| 303 |
+
1.0,
|
| 304 |
+
1.0,
|
| 305 |
+
0.28374183177948,
|
| 306 |
+
0.22523708641529083,
|
| 307 |
+
0.6108652353286743
|
| 308 |
+
],
|
| 309 |
+
"count": [
|
| 310 |
+
24099,
|
| 311 |
+
24099,
|
| 312 |
+
24099,
|
| 313 |
+
24099,
|
| 314 |
+
24099,
|
| 315 |
+
24099,
|
| 316 |
+
24099,
|
| 317 |
+
24099,
|
| 318 |
+
24099,
|
| 319 |
+
24099,
|
| 320 |
+
24099,
|
| 321 |
+
24099,
|
| 322 |
+
24099,
|
| 323 |
+
24099,
|
| 324 |
+
24099,
|
| 325 |
+
24099,
|
| 326 |
+
24099,
|
| 327 |
+
24099,
|
| 328 |
+
24099,
|
| 329 |
+
24099,
|
| 330 |
+
24095,
|
| 331 |
+
24095,
|
| 332 |
+
24095
|
| 333 |
+
]
|
| 334 |
+
}
|
| 335 |
+
}
|
autostep_v2_bs128_30k/gr00t_n17/checkpoints/checkpoint-30K/provenance/launch_contract.json
ADDED
|
@@ -0,0 +1,28 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"recorded_at": 1793172963.93596,
|
| 3 |
+
"source_download_verification": {
|
| 4 |
+
"state": "complete",
|
| 5 |
+
"pid": 1624429,
|
| 6 |
+
"repo_id": "RwH-Lab/rwh-auto-step-v2",
|
| 7 |
+
"revision": "7ccc8d62be1d6ff997267dcf55b831cbe865eedd",
|
| 8 |
+
"root": "/home/work/mntvol/data/RwH-Kuavo-v2/rwh-auto-step-v2",
|
| 9 |
+
"started_at": 1793171063.8235075,
|
| 10 |
+
"updated_at": 1793171185.3980815,
|
| 11 |
+
"verified_files": 191,
|
| 12 |
+
"verified_bytes": 2971536288,
|
| 13 |
+
"elapsed_s": 121.57455372810364
|
| 14 |
+
},
|
| 15 |
+
"files": {
|
| 16 |
+
"configs/Bs128_30k/train/gr00t_n17_autostep_v2.yaml": "0d9729c76c36792e060368c79a8c53d8b0657d3fb55394acecefca250a56f4c8",
|
| 17 |
+
"configs/Bs128_30k/data/dataset_autostep_v2.yaml": "71778a003f0eb52cf5154879a831c63f1391d534792217649e588f3c376ec3c7",
|
| 18 |
+
"scripts/train/run_n17.py": "f1a726e523a4a71e59de2d1d69cda3dd7f341dd257b98586c8673d166b6598ec",
|
| 19 |
+
"scripts/data/view.py": "ca7f03c7fca94575b9e089bed2727981406791b8ddbf382b61059bc3bbf937e6",
|
| 20 |
+
"scripts/adapters/groot_v3.py": "f23c095efaefcaa3bbeb20487ba70141520783ffd820cf39354b9ee1a93463b1"
|
| 21 |
+
},
|
| 22 |
+
"gpu": 0,
|
| 23 |
+
"batch_size": 128,
|
| 24 |
+
"steps": 30000,
|
| 25 |
+
"base_model": "nvidia/GR00T-N1.7-3B",
|
| 26 |
+
"base_revision": "2fc962b973bccdd5d8ce4f67cc63b264d6886495",
|
| 27 |
+
"gpu_switch_record": "/home/work/mntvol/RwH-Kuavo_ckpt/autostep_v2_bs128_30k/_audit/gpu_switch.json"
|
| 28 |
+
}
|
autostep_v2_bs128_30k/gr00t_n17/checkpoints/checkpoint-30K/provenance/scripts/__init__.py
ADDED
|
@@ -0,0 +1 @@
|
|
|
|
|
|
|
| 1 |
+
"""Reusable RwH data views and LeTools training entry points."""
|
autostep_v2_bs128_30k/gr00t_n17/checkpoints/checkpoint-30K/provenance/scripts/adapters/__init__.py
ADDED
|
@@ -0,0 +1 @@
|
|
|
|
|
|
|
| 1 |
+
"""Small compatibility adapters; model and trainer implementations stay upstream."""
|
autostep_v2_bs128_30k/gr00t_n17/checkpoints/checkpoint-30K/provenance/scripts/adapters/groot_v3.py
ADDED
|
@@ -0,0 +1,86 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
"""Adapt v3 row/timestamp views to the bundled native episode interface."""
|
| 2 |
+
from pathlib import Path
|
| 3 |
+
import numpy as np
|
| 4 |
+
import pandas as pd
|
| 5 |
+
|
| 6 |
+
|
| 7 |
+
def install(view):
|
| 8 |
+
import torch
|
| 9 |
+
from PIL import Image
|
| 10 |
+
import gr00t.data.dataset.sharded_single_step_dataset as single
|
| 11 |
+
import gr00t.data.dataset.factory as factory
|
| 12 |
+
from gr00t.utils.video_utils import get_frames_by_timestamps
|
| 13 |
+
|
| 14 |
+
class EpisodeLoader:
|
| 15 |
+
def __init__(self, dataset_path, modality_configs, video_backend, video_backend_kwargs=None):
|
| 16 |
+
if Path(dataset_path).resolve()!=view.root: raise ValueError('Unexpected source dataset')
|
| 17 |
+
self.modality_configs=modality_configs
|
| 18 |
+
self.video_backend=video_backend
|
| 19 |
+
self.episode_lengths=[s['end']-s['start'] for s in view.segments]
|
| 20 |
+
def get_episode_length(self,index): return self.episode_lengths[index]
|
| 21 |
+
def __len__(self): return len(self.episode_lengths)
|
| 22 |
+
def get_initial_actions(self): return []
|
| 23 |
+
def get_dataset_statistics(self):
|
| 24 |
+
return {kind:{'body':{k:v for k,v in view.stats[feature].items() if k!='count'}}
|
| 25 |
+
for kind,feature in [('state','observation.state'),('action','action')]}
|
| 26 |
+
def __getitem__(self,index):
|
| 27 |
+
s=view.segments[index]; rows=np.arange(s['start'],s['end'])
|
| 28 |
+
df=pd.DataFrame({'state.body':list(view.state[rows]),'action.body':list(view.action[rows]),
|
| 29 |
+
'rwh.validity':list(view.validity[rows]),'rwh.anchor':rows,
|
| 30 |
+
'language.task':[view.tasks[s['task']]]*len(rows)})
|
| 31 |
+
for camera in view.cameras:
|
| 32 |
+
needed=getattr(self,'requested',{}).get(index,np.arange(len(rows)))
|
| 33 |
+
locations=[view.video_location(rows[i],camera) for i in needed]
|
| 34 |
+
if not locations:
|
| 35 |
+
df['video.'+camera]=[None]*len(rows)
|
| 36 |
+
continue
|
| 37 |
+
assert len({str(p) for p,t in locations})==1
|
| 38 |
+
frames=get_frames_by_timestamps(str(locations[0][0]),np.asarray([t for p,t in locations]),
|
| 39 |
+
video_backend=self.video_backend)
|
| 40 |
+
assert len(frames)==len(needed)
|
| 41 |
+
values=[None]*len(rows)
|
| 42 |
+
for i,frame in zip(needed,frames):values[int(i)]=Image.fromarray(np.ascontiguousarray(frame))
|
| 43 |
+
df['video.'+camera]=values
|
| 44 |
+
return df
|
| 45 |
+
|
| 46 |
+
# Scoped to this process and exact source root. Upstream files remain clean.
|
| 47 |
+
single.LeRobotEpisodeLoader=EpisodeLoader
|
| 48 |
+
original_shard=single.ShardedSingleStepDataset.get_shard
|
| 49 |
+
def shard(self,index):
|
| 50 |
+
# Keep native shard scheduling, processing and prefetch. Only v3 I/O gets
|
| 51 |
+
# the exact requested frame indices so unused full episodes aren't decoded.
|
| 52 |
+
deltas=np.asarray(self.modality_configs['video'].delta_indices)
|
| 53 |
+
requested={}
|
| 54 |
+
for ep,steps in self.sharded_episodes[index]:
|
| 55 |
+
needed=(np.asarray(steps)[:,None]+deltas[None,:]).reshape(-1)
|
| 56 |
+
requested.setdefault(int(ep),[]).extend(needed.tolist())
|
| 57 |
+
self.episode_loader.requested={k:np.unique(v).astype(int) for k,v in requested.items()}
|
| 58 |
+
try:return original_shard(self,index)
|
| 59 |
+
finally:self.episode_loader.requested={}
|
| 60 |
+
single.ShardedSingleStepDataset.get_shard=shard
|
| 61 |
+
original=single.ShardedSingleStepDataset.get_datapoint
|
| 62 |
+
def datapoint(self,episode_data,step_index):
|
| 63 |
+
encoded=original(self,episode_data,step_index)
|
| 64 |
+
indices=step_index+np.asarray(self.modality_configs['action'].delta_indices)
|
| 65 |
+
validity=np.stack(episode_data['rwh.validity'].iloc[indices])
|
| 66 |
+
mask=encoded['action_mask']
|
| 67 |
+
padded=torch.zeros_like(mask,dtype=torch.bool)
|
| 68 |
+
padded[:len(indices),:23]=torch.as_tensor(validity,device=mask.device)
|
| 69 |
+
encoded['action_mask']=mask.bool() & padded
|
| 70 |
+
return encoded
|
| 71 |
+
single.ShardedSingleStepDataset.get_datapoint=datapoint
|
| 72 |
+
def validated_stats(path,*args):
|
| 73 |
+
if Path(path).resolve()!=view.root: raise ValueError('Unexpected statistics source')
|
| 74 |
+
return view.stats
|
| 75 |
+
factory.generate_stats=validated_stats
|
| 76 |
+
factory.generate_rel_stats=validated_stats
|
| 77 |
+
|
| 78 |
+
|
| 79 |
+
def modality(view,horizon):
|
| 80 |
+
from gr00t.data.types import ModalityConfig, ActionConfig, ActionRepresentation, ActionType, ActionFormat
|
| 81 |
+
return {'new_embodiment':{
|
| 82 |
+
'video':ModalityConfig(delta_indices=[0],modality_keys=view.cameras),
|
| 83 |
+
'state':ModalityConfig(delta_indices=[0],modality_keys=['body']),
|
| 84 |
+
'action':ModalityConfig(delta_indices=list(range(horizon)),modality_keys=['body'],
|
| 85 |
+
action_configs=[ActionConfig(rep=ActionRepresentation.ABSOLUTE,type=ActionType.NON_EEF,format=ActionFormat.DEFAULT)]),
|
| 86 |
+
'language':ModalityConfig(delta_indices=[0],modality_keys=['task'])}}
|
autostep_v2_bs128_30k/gr00t_n17/checkpoints/checkpoint-30K/provenance/scripts/data/__init__.py
ADDED
|
@@ -0,0 +1 @@
|
|
|
|
|
|
|
| 1 |
+
"""Read-only source dataset views."""
|
autostep_v2_bs128_30k/gr00t_n17/checkpoints/checkpoint-30K/provenance/scripts/data/prepare.py
ADDED
|
@@ -0,0 +1,18 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
"""Read and audit the configured logical view without exporting a dataset."""
|
| 2 |
+
import argparse
|
| 3 |
+
import json
|
| 4 |
+
from scripts.data.view import DataView
|
| 5 |
+
|
| 6 |
+
|
| 7 |
+
def main():
|
| 8 |
+
parser = argparse.ArgumentParser()
|
| 9 |
+
parser.add_argument('--config', required=True)
|
| 10 |
+
parser.add_argument('--audit-dir', required=True)
|
| 11 |
+
args = parser.parse_args()
|
| 12 |
+
view = DataView(args.config)
|
| 13 |
+
view.save_audit(args.audit_dir)
|
| 14 |
+
print(json.dumps(view.manifest(), indent=2))
|
| 15 |
+
|
| 16 |
+
|
| 17 |
+
if __name__ == '__main__':
|
| 18 |
+
main()
|
autostep_v2_bs128_30k/gr00t_n17/checkpoints/checkpoint-30K/provenance/scripts/data/validate.py
ADDED
|
@@ -0,0 +1,67 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
"""Exhaustive index/target checks plus sampled native video comparisons."""
|
| 2 |
+
import argparse
|
| 3 |
+
import hashlib
|
| 4 |
+
import json
|
| 5 |
+
from pathlib import Path
|
| 6 |
+
import numpy as np
|
| 7 |
+
from scripts.data.view import DataView, statistics
|
| 8 |
+
|
| 9 |
+
|
| 10 |
+
def source_inventory(root):
|
| 11 |
+
return {str(p.relative_to(root)):[p.stat().st_size,p.stat().st_mtime_ns] for p in root.rglob('*') if p.is_file()}
|
| 12 |
+
|
| 13 |
+
|
| 14 |
+
def main():
|
| 15 |
+
p=argparse.ArgumentParser();p.add_argument('--config',required=True);p.add_argument('--output',required=True)
|
| 16 |
+
args=p.parse_args();view=DataView(args.config)
|
| 17 |
+
inventory=source_inventory(view.root)
|
| 18 |
+
assert np.array_equal(view.action,view.raw['action'][view.rows,:23])
|
| 19 |
+
assert np.array_equal(view.validity,view.raw['validity.action'][view.rows,:23])
|
| 20 |
+
assert np.array_equal(view.state,view.raw['observation.state'][view.rows,:20])
|
| 21 |
+
assert view.validity[:,:4].all()
|
| 22 |
+
anchor_report={}
|
| 23 |
+
for policy in ['gr00t_n15','gr00t_n17','multi_task_dit']:
|
| 24 |
+
anchors=view.anchors(policy);idx=anchors[:,None]+view.offsets(policy)[None,:]
|
| 25 |
+
assert (view.episode[idx]==view.episode[anchors,None]).all()
|
| 26 |
+
assert (view.task[idx]==view.task[anchors,None]).all()
|
| 27 |
+
assert (np.diff(view.rows[idx],axis=1)==view.stride).all()
|
| 28 |
+
anchor_report[policy]={'count':len(anchors),'first':int(anchors[0]),'last':int(anchors[-1])}
|
| 29 |
+
for j in range(23):
|
| 30 |
+
actual=view.action[view.validity[:,j],j]
|
| 31 |
+
assert np.isclose(view.stats['action']['mean'][j],actual.mean(dtype=np.float64))
|
| 32 |
+
assert view.stats['action']['count'][j]==len(actual)
|
| 33 |
+
# Native LeRobot video decode is compared against an independent PyAV seek.
|
| 34 |
+
import av
|
| 35 |
+
from lerobot.datasets.lerobot_dataset import LeRobotDataset
|
| 36 |
+
native=LeRobotDataset(view.config['source']['repo_id'],root=view.root,video_backend='pyav',
|
| 37 |
+
return_uint8=True,tolerance_s=1/30+.001)
|
| 38 |
+
rng=np.random.default_rng(42)
|
| 39 |
+
chosen=set(rng.choice(len(view.rows),24,replace=False).tolist())
|
| 40 |
+
for s in view.segments[::max(1,len(view.segments)//12)]:chosen.update([s['start'],s['end']-1])
|
| 41 |
+
checks=[]
|
| 42 |
+
for i in sorted(chosen):
|
| 43 |
+
sample=native[int(view.rows[i])]
|
| 44 |
+
for camera in view.cameras:
|
| 45 |
+
path,timestamp=view.video_location(i,camera)
|
| 46 |
+
with av.open(str(path)) as container:
|
| 47 |
+
stream=container.streams.video[0];stream.thread_type='NONE';stream.codec_context.thread_count=1
|
| 48 |
+
container.seek(max(0,int(timestamp/stream.time_base)),stream=stream,backward=True)
|
| 49 |
+
decoded=[]
|
| 50 |
+
for frame in container.decode(stream):
|
| 51 |
+
decoded.append((abs(frame.time-timestamp),frame.time,frame.to_ndarray(format='rgb24')))
|
| 52 |
+
if frame.time>=timestamp:break
|
| 53 |
+
distance,pts,image=min(decoded,key=lambda x:x[0])
|
| 54 |
+
assert distance<1/60+.001,(i,camera,timestamp,pts)
|
| 55 |
+
other=sample[camera].permute(1,2,0).numpy()
|
| 56 |
+
assert np.array_equal(image,other),(i,camera,int(np.max(np.abs(image.astype(int)-other))))
|
| 57 |
+
checks.append({'row':i,'source_row':int(view.rows[i]),'camera':camera,
|
| 58 |
+
'timestamp':timestamp,'decoded_pts':pts,'sha256':hashlib.sha256(image.tobytes()).hexdigest()})
|
| 59 |
+
assert inventory==source_inventory(view.root),'Source dataset was mutated'
|
| 60 |
+
out=Path(args.output);view.save_audit(out)
|
| 61 |
+
report={'status':'passed','manifest':view.manifest(),'anchors':anchor_report,'video_checks':checks,
|
| 62 |
+
'all_selected_targets_equal_source':True,'all_anchors_task_safe':True,'source_unchanged':True}
|
| 63 |
+
(out/'data_validation.json').write_text(json.dumps(report,indent=2)+'\n')
|
| 64 |
+
print(json.dumps({**report,'video_checks':len(checks)},indent=2))
|
| 65 |
+
|
| 66 |
+
|
| 67 |
+
if __name__=='__main__':main()
|
autostep_v2_bs128_30k/gr00t_n17/checkpoints/checkpoint-30K/provenance/scripts/data/view.py
ADDED
|
@@ -0,0 +1,166 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
"""A small in-memory index over original v3 files; never exports another dataset."""
|
| 2 |
+
from pathlib import Path
|
| 3 |
+
import copy
|
| 4 |
+
import hashlib
|
| 5 |
+
import json
|
| 6 |
+
import numpy as np
|
| 7 |
+
import pyarrow.parquet as pq
|
| 8 |
+
import yaml
|
| 9 |
+
|
| 10 |
+
|
| 11 |
+
def load_config(path):
|
| 12 |
+
path = Path(path).resolve()
|
| 13 |
+
cfg = yaml.safe_load(path.read_text())
|
| 14 |
+
allowed = {'source', 'sampling', 'state', 'action', 'statistics', 'video', 'adapters'}
|
| 15 |
+
if set(cfg) - allowed:
|
| 16 |
+
raise ValueError(f'Unknown data keys: {set(cfg)-allowed}')
|
| 17 |
+
if 'output' in cfg or cfg['action']['source'] != 'action':
|
| 18 |
+
raise ValueError('Expected original actions and a read-only view')
|
| 19 |
+
if cfg['state']['indices'] != [0, 20] or cfg['action']['indices'] != [0, 23]:
|
| 20 |
+
raise ValueError('Expected state20/action23')
|
| 21 |
+
if not cfg['sampling']['split_at_task_change']:
|
| 22 |
+
raise ValueError('Task boundaries are mandatory')
|
| 23 |
+
if cfg['action']['invalid'] != 'mask_loss' or cfg['statistics'] != {'state':'all_inputs', 'action':'valid_targets'}:
|
| 24 |
+
raise ValueError('Unsupported validity/statistics contract')
|
| 25 |
+
if cfg['video'] != {'mode':'original', 'alignment':'timestamp'}:
|
| 26 |
+
raise ValueError('Videos must use original timestamps')
|
| 27 |
+
if cfg['adapters'] != {'gr00t_n17':{'enabled':True,'type':'letools_native_v3'}}:
|
| 28 |
+
raise ValueError('This v3 contract requires the enabled native N1.7 adapter')
|
| 29 |
+
return cfg
|
| 30 |
+
|
| 31 |
+
|
| 32 |
+
def statistics(values, mask=None):
|
| 33 |
+
result = {k: [] for k in ['mean', 'std', 'min', 'max', 'q01', 'q99', 'count']}
|
| 34 |
+
for d in range(values.shape[1]):
|
| 35 |
+
a = values[:, d] if mask is None else values[mask[:, d], d]
|
| 36 |
+
if not len(a) or not np.isfinite(a).all():
|
| 37 |
+
raise ValueError(f'No finite valid statistics for dimension {d}')
|
| 38 |
+
for k, v in zip(result, [a.mean(dtype=np.float64), a.std(dtype=np.float64), a.min(), a.max(),
|
| 39 |
+
np.quantile(a, .01), np.quantile(a, .99), len(a)]):
|
| 40 |
+
result[k].append(float(v) if k != 'count' else int(v))
|
| 41 |
+
return result
|
| 42 |
+
|
| 43 |
+
|
| 44 |
+
class DataView:
|
| 45 |
+
def __init__(self, config_path):
|
| 46 |
+
self.config_path = str(Path(config_path).resolve())
|
| 47 |
+
self.config = load_config(config_path)
|
| 48 |
+
self.root = Path(self.config['source']['root']).resolve()
|
| 49 |
+
self.info = json.loads((self.root/'meta/info.json').read_text())
|
| 50 |
+
assert self.info['codebase_version'].startswith('v3')
|
| 51 |
+
sf, fps = self.config['sampling']['source_fps'], self.config['sampling']['fps']
|
| 52 |
+
if self.info['fps'] != sf or sf % fps:
|
| 53 |
+
raise ValueError('This adapter requires an exact integer FPS stride')
|
| 54 |
+
self.stride, self.fps = sf // fps, fps
|
| 55 |
+
if self.stride < 1: raise ValueError('Cannot upsample')
|
| 56 |
+
files = sorted((self.root/'data').rglob('*.parquet'))
|
| 57 |
+
table = pq.read_table(files)
|
| 58 |
+
order = np.argsort(table['index'].to_numpy())
|
| 59 |
+
self.raw = {k: np.asarray(table[k].to_pylist())[order] for k in
|
| 60 |
+
['index', 'episode_index', 'frame_index', 'timestamp', 'task_index',
|
| 61 |
+
self.config['state']['source'], 'action', self.config['action']['validity_feature']]}
|
| 62 |
+
assert np.array_equal(self.raw['index'], np.arange(len(order)))
|
| 63 |
+
self.episode_meta = {int(r['episode_index']): r for r in
|
| 64 |
+
pq.read_table(sorted((self.root/'meta/episodes').rglob('*.parquet'))).to_pylist()}
|
| 65 |
+
tasks = pq.read_table(self.root/'meta/tasks.parquet').to_pandas()
|
| 66 |
+
self.tasks = {int(row.task_index): str(label) for label, row in tasks.iterrows()}
|
| 67 |
+
self.cameras = [k for k,v in self.info['features'].items() if v['dtype']=='video']
|
| 68 |
+
self.segments, selection = [], []
|
| 69 |
+
for ep in sorted(self.episode_meta):
|
| 70 |
+
rows = np.flatnonzero(self.raw['episode_index']==ep)
|
| 71 |
+
assert np.all(np.diff(rows)==1)
|
| 72 |
+
task = self.raw['task_index'][rows]
|
| 73 |
+
edges = np.r_[0, np.flatnonzero(task[1:]!=task[:-1])+1, len(rows)]
|
| 74 |
+
selected = rows[::self.stride]
|
| 75 |
+
for lo, hi in zip(edges[:-1], edges[1:]):
|
| 76 |
+
chosen = selected[(selected>=rows[lo]) & (selected<rows[hi-1]+1)]
|
| 77 |
+
if not len(chosen): continue
|
| 78 |
+
start = len(selection); selection.extend(chosen.tolist())
|
| 79 |
+
self.segments.append(dict(start=start, end=len(selection), source_episode=ep,
|
| 80 |
+
source_from=int(rows[lo]), source_to=int(rows[hi-1])+1,
|
| 81 |
+
task=int(task[lo])))
|
| 82 |
+
self.rows = np.asarray(selection, dtype=np.int64)
|
| 83 |
+
self.state = self.raw[self.config['state']['source']][self.rows, :20].astype(np.float32)
|
| 84 |
+
self.action = self.raw['action'][self.rows, :23].astype(np.float32)
|
| 85 |
+
self.validity = self.raw[self.config['action']['validity_feature']][self.rows,:23].astype(bool)
|
| 86 |
+
self.task = self.raw['task_index'][self.rows].astype(np.int64)
|
| 87 |
+
self.episode = np.empty(len(self.rows), dtype=np.int64)
|
| 88 |
+
for i,s in enumerate(self.segments): self.episode[s['start']:s['end']] = i
|
| 89 |
+
assert np.isfinite(self.state).all() and np.isfinite(self.action).all()
|
| 90 |
+
self.stats = {'observation.state': statistics(self.state), 'action': statistics(self.action,self.validity)}
|
| 91 |
+
signature = [(str(p.relative_to(self.root)),p.stat().st_size,p.stat().st_mtime_ns) for p in files]
|
| 92 |
+
self.fingerprint = hashlib.sha256(json.dumps([self.config,signature],sort_keys=True).encode()).hexdigest()
|
| 93 |
+
|
| 94 |
+
def anchors(self, policy):
|
| 95 |
+
history, future = {'gr00t_n15':(0,15),'gr00t_n17':(0,39),'multi_task_dit':(1,30)}[policy]
|
| 96 |
+
return np.concatenate([np.arange(s['start']+history,s['end']-future,dtype=np.int64)
|
| 97 |
+
for s in self.segments if s['end']-s['start']>history+future])
|
| 98 |
+
|
| 99 |
+
def offsets(self, policy):
|
| 100 |
+
return np.arange(-1,31) if policy=='multi_task_dit' else np.arange(40 if policy=='gr00t_n17' else 16)
|
| 101 |
+
|
| 102 |
+
def video_location(self, row, camera):
|
| 103 |
+
raw = int(self.rows[row]); ep = int(self.raw['episode_index'][raw]); meta=self.episode_meta[ep]
|
| 104 |
+
prefix='videos/'+camera+'/'
|
| 105 |
+
path=self.root/self.info['video_path'].format(video_key=camera,
|
| 106 |
+
chunk_index=int(meta[prefix+'chunk_index']),file_index=int(meta[prefix+'file_index']))
|
| 107 |
+
return path, float(meta[prefix+'from_timestamp'])+float(self.raw['timestamp'][raw])
|
| 108 |
+
|
| 109 |
+
def manifest(self):
|
| 110 |
+
return dict(fingerprint=self.fingerprint, config=self.config, source_rows=len(self.raw['index']),
|
| 111 |
+
selected_rows=len(self.rows), task_episodes=len(self.segments),
|
| 112 |
+
eligible_anchors={p:len(self.anchors(p)) for p in ['gr00t_n15','gr00t_n17','multi_task_dit']},
|
| 113 |
+
invalid_action_elements=int((~self.validity).sum()), original_torso=True)
|
| 114 |
+
|
| 115 |
+
def save_audit(self, directory):
|
| 116 |
+
directory=Path(directory); directory.mkdir(parents=True,exist_ok=True)
|
| 117 |
+
(directory/'data_manifest.json').write_text(json.dumps(self.manifest(),indent=2)+'\n')
|
| 118 |
+
(directory/'statistics.json').write_text(json.dumps(self.stats,indent=2)+'\n')
|
| 119 |
+
np.savez(directory/'indices.npz',source_rows=self.rows,logical_episode=self.episode)
|
| 120 |
+
|
| 121 |
+
|
| 122 |
+
def lerobot_dataset(view, policy):
|
| 123 |
+
import torch
|
| 124 |
+
from lerobot.datasets.lerobot_dataset import LeRobotDataset
|
| 125 |
+
from lerobot.utils.constants import IMAGENET_STATS
|
| 126 |
+
from scripts.adapters.language import letterbox224
|
| 127 |
+
native=LeRobotDataset(view.config['source']['repo_id'],root=view.root,video_backend='pyav',
|
| 128 |
+
return_uint8=True,tolerance_s=1/30+.001)
|
| 129 |
+
meta=copy.copy(native.meta)
|
| 130 |
+
meta.info=copy.deepcopy(meta.info)
|
| 131 |
+
meta.info.update(fps=view.fps,total_frames=len(view.rows),total_episodes=len(view.segments))
|
| 132 |
+
meta.info['features']={k:copy.deepcopy(view.info['features'][k]) for k in
|
| 133 |
+
[*view.cameras,'observation.state','action']}
|
| 134 |
+
meta.info['features']['observation.state']['shape']=[20]
|
| 135 |
+
meta.info['features']['observation.state']['names']=view.info['features']['observation.state']['names'][:20]
|
| 136 |
+
from datasets import Dataset
|
| 137 |
+
meta.episodes=Dataset.from_list([{'episode_index':i,'length':s['end']-s['start'],
|
| 138 |
+
'dataset_from_index':s['start'],'dataset_to_index':s['end'],
|
| 139 |
+
'tasks':[view.tasks[s['task']]]} for i,s in enumerate(view.segments)])
|
| 140 |
+
for camera in view.cameras: meta.info['features'][camera]['shape']=[224,224,3]
|
| 141 |
+
meta.stats={k:{s:torch.tensor(v) for s,v in stats.items()} for k,stats in view.stats.items()}
|
| 142 |
+
for camera in view.cameras:
|
| 143 |
+
meta.stats[camera]={k:torch.tensor(v,dtype=torch.float32) for k,v in IMAGENET_STATS.items()}
|
| 144 |
+
class ViewedDataset(torch.utils.data.Dataset):
|
| 145 |
+
def __init__(self):
|
| 146 |
+
self.meta=meta; self.root=view.root; self.episodes=None
|
| 147 |
+
self.num_frames=len(view.rows); self.num_episodes=len(view.segments)
|
| 148 |
+
self.features=meta.features; self.fps=view.fps; self.repo_id=native.repo_id
|
| 149 |
+
def __len__(self): return self.num_frames
|
| 150 |
+
def __getitem__(self,index):
|
| 151 |
+
index=int(index); offsets=view.offsets(policy); seq=index+offsets
|
| 152 |
+
if seq.min()<0 or seq.max()>=len(view.rows) or np.any(view.episode[seq]!=view.episode[index]):
|
| 153 |
+
raise ValueError('Sampling crossed a task boundary')
|
| 154 |
+
obs_indices=[index-1,index] if policy=='multi_task_dit' else [index]
|
| 155 |
+
frames=[native[int(view.rows[i])] for i in obs_indices]
|
| 156 |
+
result={k:torch.stack([letterbox224(f[k]) for f in frames]) for k in view.cameras}
|
| 157 |
+
states=torch.tensor(view.state[obs_indices])
|
| 158 |
+
if policy!='multi_task_dit':
|
| 159 |
+
result={k:v[0] for k,v in result.items()}; states=states[0]
|
| 160 |
+
result.update({'observation.state':states,'action':torch.tensor(view.action[seq]),
|
| 161 |
+
'action_is_pad':torch.zeros(len(seq),dtype=torch.bool), 'index':torch.tensor(index),
|
| 162 |
+
'episode_index':torch.tensor(view.episode[index]), 'task_index':torch.tensor(view.task[index]),
|
| 163 |
+
'timestamp':torch.tensor(index/view.fps,dtype=torch.float32),
|
| 164 |
+
'task':view.tasks[int(view.task[index])]})
|
| 165 |
+
return result
|
| 166 |
+
return ViewedDataset()
|
autostep_v2_bs128_30k/gr00t_n17/checkpoints/checkpoint-30K/provenance/scripts/train/__init__.py
ADDED
|
@@ -0,0 +1 @@
|
|
|
|
|
|
|
| 1 |
+
"""LeTools training launchers."""
|
autostep_v2_bs128_30k/gr00t_n17/checkpoints/checkpoint-30K/provenance/scripts/train/checkpoints.py
ADDED
|
@@ -0,0 +1,35 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
"""Validate the native Transformers checkpoint layout, including weight shards."""
|
| 2 |
+
import json
|
| 3 |
+
from pathlib import Path
|
| 4 |
+
|
| 5 |
+
|
| 6 |
+
def weight_files(model_dir):
|
| 7 |
+
root=Path(model_dir)
|
| 8 |
+
index=root/'model.safetensors.index.json'
|
| 9 |
+
if index.exists():
|
| 10 |
+
mapping=json.loads(index.read_text())['weight_map']
|
| 11 |
+
names=sorted(set(mapping.values()))
|
| 12 |
+
if not names or any(Path(n).name!=n for n in names):raise ValueError('Unsafe/empty weight index')
|
| 13 |
+
files=[root/n for n in names]
|
| 14 |
+
if not all(p.is_file() and p.stat().st_size>0 for p in files):raise ValueError('Missing model shard')
|
| 15 |
+
from safetensors import safe_open
|
| 16 |
+
for path in files:
|
| 17 |
+
with safe_open(str(path),framework='pt',device='cpu') as sf:
|
| 18 |
+
if set(sf.keys())!={key for key,name in mapping.items() if name==path.name}:
|
| 19 |
+
raise ValueError('Weight index and shard keys differ: '+path.name)
|
| 20 |
+
else:
|
| 21 |
+
files=[root/'model.safetensors']
|
| 22 |
+
if not files[0].is_file():raise ValueError('No safetensors weights or shard index')
|
| 23 |
+
return files
|
| 24 |
+
|
| 25 |
+
|
| 26 |
+
def validate_native_checkpoint(path,step):
|
| 27 |
+
path=Path(path)
|
| 28 |
+
for name in ['config.json','processor_config.json','statistics.json','embodiment_id.json','trainer_state.json']:
|
| 29 |
+
if not (path/name).exists():
|
| 30 |
+
raise ValueError('Missing native checkpoint file: '+name)
|
| 31 |
+
processor=json.loads((path/'processor_config.json').read_text())
|
| 32 |
+
if not processor['processor_kwargs']['modality_configs']:raise ValueError('Missing saved modalities')
|
| 33 |
+
state=json.loads((path/'trainer_state.json').read_text())
|
| 34 |
+
if state['global_step']!=step:raise ValueError('Wrong native saved step')
|
| 35 |
+
return {'step':step,'weight_files':[p.name for p in weight_files(path)]}
|
autostep_v2_bs128_30k/gr00t_n17/checkpoints/checkpoint-30K/provenance/scripts/train/hf_watcher.py
ADDED
|
@@ -0,0 +1,371 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
"""Publish verified 10k/20k/30k checkpoints from successful Kuavo runs.
|
| 2 |
+
|
| 3 |
+
CPU only. Never modifies a trainer checkpoint, last symlink, or optimizer state.
|
| 4 |
+
All paths/revisions are pinned in hf_watcher.json before starting this process.
|
| 5 |
+
"""
|
| 6 |
+
from __future__ import annotations
|
| 7 |
+
|
| 8 |
+
import argparse
|
| 9 |
+
import fcntl
|
| 10 |
+
import hashlib
|
| 11 |
+
import json
|
| 12 |
+
import math
|
| 13 |
+
import os
|
| 14 |
+
from pathlib import Path
|
| 15 |
+
import re
|
| 16 |
+
import time
|
| 17 |
+
|
| 18 |
+
POLICIES = ("multi_task_dit", "gr00t_n15", "gr00t_n17")
|
| 19 |
+
STEPS = (10000, 20000, 30000)
|
| 20 |
+
|
| 21 |
+
|
| 22 |
+
def checkpoint_plan(cfg):
|
| 23 |
+
configured = cfg.get("steps_by_policy", {p: list(STEPS) for p in POLICIES})
|
| 24 |
+
if not configured or set(configured) - set(POLICIES):
|
| 25 |
+
raise ValueError("Expected checkpoint steps for known policies")
|
| 26 |
+
for policy, steps in configured.items():
|
| 27 |
+
if not steps or len(steps) != len(set(steps)) or any(s not in STEPS for s in steps):
|
| 28 |
+
raise ValueError("Invalid checkpoint steps: " + policy)
|
| 29 |
+
return [(policy, step) for step in STEPS for policy in POLICIES if step in configured.get(policy, [])]
|
| 30 |
+
|
| 31 |
+
|
| 32 |
+
def remote_prefix(cfg, policy, step):
|
| 33 |
+
root = cfg["remote_root"]
|
| 34 |
+
if not isinstance(root, str) or not re.fullmatch(r"[A-Za-z0-9][A-Za-z0-9_.-]*", root):
|
| 35 |
+
raise ValueError("Expected one safe HF experiment directory name")
|
| 36 |
+
return f"{root}/{policy}/checkpoints/checkpoint-{step // 1000}K"
|
| 37 |
+
|
| 38 |
+
|
| 39 |
+
class NotReady(Exception):
|
| 40 |
+
pass
|
| 41 |
+
|
| 42 |
+
|
| 43 |
+
def read(path):
|
| 44 |
+
return json.loads(Path(path).read_text())
|
| 45 |
+
|
| 46 |
+
|
| 47 |
+
def atomic_json(path, value):
|
| 48 |
+
path = Path(path)
|
| 49 |
+
path.parent.mkdir(parents=True, exist_ok=True)
|
| 50 |
+
temporary = path.with_suffix(path.suffix + ".tmp")
|
| 51 |
+
temporary.write_text(json.dumps(value, indent=2, ensure_ascii=False) + "\n")
|
| 52 |
+
temporary.replace(path)
|
| 53 |
+
|
| 54 |
+
|
| 55 |
+
def signature(root):
|
| 56 |
+
"""A stable native snapshot includes training_state, even though it is not uploaded."""
|
| 57 |
+
values = {}
|
| 58 |
+
for f in sorted(root.rglob("*")):
|
| 59 |
+
if f.is_symlink():
|
| 60 |
+
raise ValueError("Checkpoint contains an unexpected symlink")
|
| 61 |
+
if f.is_file():
|
| 62 |
+
s = f.stat()
|
| 63 |
+
values[str(f.relative_to(root))] = [s.st_size, s.st_mtime_ns]
|
| 64 |
+
return values
|
| 65 |
+
|
| 66 |
+
|
| 67 |
+
def completed_source(cfg, policy, step):
|
| 68 |
+
output = Path(cfg["output_root"]) / policy
|
| 69 |
+
result = read(cfg["results"][policy])
|
| 70 |
+
if result.get("status") != "passed":
|
| 71 |
+
raise NotReady("Waiting for successful final training and inference verification")
|
| 72 |
+
for key, expected in cfg.get("expected_run", {}).items():
|
| 73 |
+
if result.get(key) != expected:
|
| 74 |
+
raise ValueError("Unexpected production run field: " + key)
|
| 75 |
+
if result["policy"] != policy or result["smoke"] or result["steps_requested"] != 30000:
|
| 76 |
+
raise ValueError("Unexpected production identity")
|
| 77 |
+
if Path(result["output_dir"]) != output:
|
| 78 |
+
raise ValueError("Wrong production output")
|
| 79 |
+
final = Path(result["checkpoint"])
|
| 80 |
+
expected_final = output / ("checkpoint-30000" if policy == "gr00t_n17" else "checkpoints/030000")
|
| 81 |
+
if final.resolve() != expected_final.resolve():
|
| 82 |
+
raise ValueError("Unexpected final checkpoint location")
|
| 83 |
+
checkpoint = output / (f"checkpoint-{step}" if policy == "gr00t_n17" else f"checkpoints/{step:06d}")
|
| 84 |
+
if checkpoint.resolve().parent not in [output.resolve(), (output/"checkpoints").resolve()]:
|
| 85 |
+
raise ValueError("Checkpoint outside this run")
|
| 86 |
+
if policy == "gr00t_n17":
|
| 87 |
+
if read(final/"rwh_save_complete.json")["step"] != 30000:
|
| 88 |
+
raise ValueError("Native save marker step mismatch")
|
| 89 |
+
from scripts.train.checkpoints import validate_native_checkpoint
|
| 90 |
+
validate_native_checkpoint(checkpoint, step)
|
| 91 |
+
model = checkpoint
|
| 92 |
+
required = ["config.json", "processor_config.json", "statistics.json", "embodiment_id.json"]
|
| 93 |
+
else:
|
| 94 |
+
if read(checkpoint/"training_state/training_step.json")["step"] != step:
|
| 95 |
+
raise ValueError("LeRobot training step mismatch")
|
| 96 |
+
model = checkpoint/"pretrained_model"
|
| 97 |
+
required = ["config.json", "train_config.json", "policy_preprocessor.json", "policy_postprocessor.json", "model.safetensors"]
|
| 98 |
+
train = read(model/"train_config.json")
|
| 99 |
+
for key,value in {"steps":30000,"batch_size":128,"seed":42,"save_freq":10000}.items():
|
| 100 |
+
if train[key] != value: raise ValueError("Wrong training contract: "+key)
|
| 101 |
+
for name in required:
|
| 102 |
+
if not (model/name).is_file(): raise ValueError("Missing checkpoint file: "+name)
|
| 103 |
+
from scripts.train.checkpoints import weight_files
|
| 104 |
+
weight_files(model)
|
| 105 |
+
snapshot=signature(checkpoint)
|
| 106 |
+
newest=max(mtime for size,mtime in snapshot.values())/1e9
|
| 107 |
+
if time.time()-newest < cfg["stable_seconds"]: raise NotReady("Waiting for stability interval")
|
| 108 |
+
return checkpoint,snapshot
|
| 109 |
+
|
| 110 |
+
|
| 111 |
+
def file_info(path):
|
| 112 |
+
size = path.stat().st_size
|
| 113 |
+
sha256 = hashlib.sha256()
|
| 114 |
+
blob = hashlib.sha1(f"blob {size}\0".encode())
|
| 115 |
+
with path.open("rb") as stream:
|
| 116 |
+
for chunk in iter(lambda: stream.read(8 * 1024 * 1024), b""):
|
| 117 |
+
sha256.update(chunk)
|
| 118 |
+
blob.update(chunk)
|
| 119 |
+
return {"size": size, "sha256": sha256.hexdigest(), "git_blob_sha1": blob.hexdigest()}
|
| 120 |
+
|
| 121 |
+
|
| 122 |
+
def validate_tensors(model):
|
| 123 |
+
import torch
|
| 124 |
+
from safetensors import safe_open
|
| 125 |
+
torch.set_num_threads(2)
|
| 126 |
+
count = 0
|
| 127 |
+
for file in sorted(model.glob("*.safetensors")):
|
| 128 |
+
with safe_open(str(file), framework="pt", device="cpu") as tensors:
|
| 129 |
+
for name in tensors.keys():
|
| 130 |
+
tensor = tensors.get_slice(name)
|
| 131 |
+
shape = tensor.get_shape()
|
| 132 |
+
if not shape:
|
| 133 |
+
chunks = [tensors.get_tensor(name)]
|
| 134 |
+
else:
|
| 135 |
+
rows = max(1, 8_000_000 // max(1, math.prod(shape[1:])))
|
| 136 |
+
chunks = (tensor[i:i + rows] for i in range(0, shape[0], rows))
|
| 137 |
+
for chunk in chunks:
|
| 138 |
+
if not torch.isfinite(chunk).all().item():
|
| 139 |
+
raise ValueError("Nonfinite saved tensor: " + name)
|
| 140 |
+
count += 1
|
| 141 |
+
if not count:
|
| 142 |
+
raise ValueError("No saved tensors")
|
| 143 |
+
return count
|
| 144 |
+
|
| 145 |
+
|
| 146 |
+
def compare_remote(api, repo, prefix, revision, files):
|
| 147 |
+
"""LFS SHA256 or Git object SHA1 proves actual content, beyond size equality."""
|
| 148 |
+
from huggingface_hub.utils import EntryNotFoundError
|
| 149 |
+
try:
|
| 150 |
+
entries = list(api.list_repo_tree(repo, path_in_repo=prefix, revision=revision,
|
| 151 |
+
repo_type="model", recursive=True))
|
| 152 |
+
except EntryNotFoundError:
|
| 153 |
+
entries = []
|
| 154 |
+
remote = {entry.path[len(prefix) + 1:]: entry for entry in entries if hasattr(entry, "size")}
|
| 155 |
+
missing = sorted(set(files) - set(remote))
|
| 156 |
+
extra = sorted(set(remote) - set(files))
|
| 157 |
+
mismatch = []
|
| 158 |
+
for name in sorted(set(files) & set(remote)):
|
| 159 |
+
expected, actual = files[name], remote[name]
|
| 160 |
+
lfs = actual.lfs
|
| 161 |
+
remote_hash = lfs.sha256 if lfs is not None else actual.blob_id
|
| 162 |
+
expected_hash = expected["sha256"] if lfs is not None else expected["git_blob_sha1"]
|
| 163 |
+
if actual.size != expected["size"] or remote_hash != expected_hash:
|
| 164 |
+
mismatch.append(name)
|
| 165 |
+
return {"missing": missing, "extra": extra, "mismatch": mismatch,
|
| 166 |
+
"file_count": len(remote), "bytes": sum(x.size for x in remote.values())}
|
| 167 |
+
|
| 168 |
+
|
| 169 |
+
def matches(check):
|
| 170 |
+
return not any(check[k] for k in ("missing", "extra", "mismatch"))
|
| 171 |
+
|
| 172 |
+
|
| 173 |
+
def payload(cfg, policy, step, checkpoint, snapshot):
|
| 174 |
+
prefix = remote_prefix(cfg, policy, step)
|
| 175 |
+
work=Path(cfg["state_dir"])/"payloads"/cfg["remote_root"]/policy/str(step)
|
| 176 |
+
manifest_path=work/"checkpoint_manifest.json"
|
| 177 |
+
model=checkpoint if policy=="gr00t_n17" else checkpoint/"pretrained_model"
|
| 178 |
+
def included(f):
|
| 179 |
+
rel=f.relative_to(checkpoint)
|
| 180 |
+
return not any(x in ("training_state",) for x in rel.parts) and not f.name.startswith(("optimizer", "scheduler", "rng_state"))
|
| 181 |
+
if manifest_path.exists():
|
| 182 |
+
manifest=read(manifest_path)
|
| 183 |
+
if manifest["source_signature"] != snapshot or manifest["remote_prefix"] != prefix:
|
| 184 |
+
raise ValueError("Saved checkpoint or destination changed")
|
| 185 |
+
else:
|
| 186 |
+
count=validate_tensors(model)
|
| 187 |
+
provenance=work/"provenance";provenance.mkdir(parents=True,exist_ok=True)
|
| 188 |
+
for source,target in cfg["provenance"][policy].items():
|
| 189 |
+
dest=provenance/target;dest.parent.mkdir(parents=True,exist_ok=True)
|
| 190 |
+
dest.write_bytes(Path(source).read_bytes())
|
| 191 |
+
(work/"README.md").write_text(
|
| 192 |
+
f"# {policy}: original-action Kuavo, checkpoint {step} of 30000 updates\n\n"
|
| 193 |
+
"Original v3 data; 30Hz to 10Hz logical view; state20/action23; seed42; actual batch128; accumulation1.\n"
|
| 194 |
+
"Torso targets are the original action commands. Invalid action elements are masked.\n"
|
| 195 |
+
"N1.5/DiT use LeTools-pinned LeRobot; N1.7 uses bundled native model, processor, sharded loader and Trainer.\n"
|
| 196 |
+
"LeTools revision: 6d2002613241ae2a32c5d8fd753f80caa1383fba. See provenance for adapter and YAML.\n"
|
| 197 |
+
"N1.7 uses native checkpoint layout. Optimizer/RNG states stay on the training server.\n")
|
| 198 |
+
if cfg.get("dataset_source"):
|
| 199 |
+
with (work/"README.md").open("a") as summary:
|
| 200 |
+
source = cfg["dataset_source"]
|
| 201 |
+
summary.write(f"\nDataset: `{source['repo_id']}` at revision `{source['revision']}`.\n")
|
| 202 |
+
paths={str(f.relative_to(checkpoint)):f for f in checkpoint.rglob("*") if f.is_file() and included(f)}
|
| 203 |
+
paths.update({str(f.relative_to(work)):f for f in work.rglob("*") if f.is_file()})
|
| 204 |
+
manifest={"schema":2,"policy":policy,"step":step,"repo_id":cfg["repo_id"],
|
| 205 |
+
"remote_prefix":prefix,
|
| 206 |
+
"source_checkpoint":str(checkpoint),"source_signature":snapshot,"tensor_count":count,
|
| 207 |
+
"training_state_included":False,"files":{n:file_info(f) for n,f in sorted(paths.items())}}
|
| 208 |
+
if signature(checkpoint)!=snapshot: raise ValueError("Checkpoint changed during validation")
|
| 209 |
+
atomic_json(manifest_path,manifest)
|
| 210 |
+
paths={name:work/name if name.startswith("provenance/") or name=="README.md" else checkpoint/name
|
| 211 |
+
for name in manifest["files"]}
|
| 212 |
+
paths["checkpoint_manifest.json"]=manifest_path
|
| 213 |
+
files=dict(manifest["files"]);files["checkpoint_manifest.json"]=file_info(manifest_path)
|
| 214 |
+
return manifest,paths,files
|
| 215 |
+
|
| 216 |
+
|
| 217 |
+
def publish(api, cfg, policy, step):
|
| 218 |
+
from huggingface_hub import CommitOperationAdd
|
| 219 |
+
checkpoint, snapshot = completed_source(cfg, policy, step)
|
| 220 |
+
manifest, paths, files = payload(cfg, policy, step, checkpoint, snapshot)
|
| 221 |
+
prefix, repo = manifest["remote_prefix"], cfg["repo_id"]
|
| 222 |
+
print(json.dumps({"mapping": {"local": str(checkpoint), "HF": repo + "/" + prefix},
|
| 223 |
+
"step": step, "files": len(files)}), flush=True)
|
| 224 |
+
head = api.model_info(repo).sha
|
| 225 |
+
before = compare_remote(api, repo, prefix, head, files)
|
| 226 |
+
if before["extra"] or before["mismatch"]:
|
| 227 |
+
raise ValueError("Existing HF checkpoint conflicts with the validated local manifest")
|
| 228 |
+
uploaded = False
|
| 229 |
+
if before["missing"]:
|
| 230 |
+
if signature(checkpoint) != snapshot:
|
| 231 |
+
raise ValueError("Checkpoint changed before upload")
|
| 232 |
+
commit = api.create_commit(
|
| 233 |
+
repo, repo_type="model", parent_commit=head, num_threads=2,
|
| 234 |
+
operations=[CommitOperationAdd(path_in_repo=prefix + "/" + name,
|
| 235 |
+
path_or_fileobj=str(paths[name])) for name in before["missing"]],
|
| 236 |
+
commit_message=f"Kuavo original action: {policy} checkpoint {step // 1000}K")
|
| 237 |
+
head = commit.oid
|
| 238 |
+
uploaded = True
|
| 239 |
+
check = compare_remote(api, repo, prefix, head, files)
|
| 240 |
+
if not matches(check):
|
| 241 |
+
raise ValueError("Remote manifest verification failed: " + json.dumps(check))
|
| 242 |
+
if signature(checkpoint) != snapshot:
|
| 243 |
+
raise ValueError("Source changed during upload; do not mark verified")
|
| 244 |
+
return {"status": "verified", "policy": policy, "step": step, "revision": head,
|
| 245 |
+
"prefix": prefix, "uploaded_this_attempt": uploaded, "verification": check,
|
| 246 |
+
"verified_at": time.time(), "manifest": str(Path(cfg["state_dir"]) / "payloads" /
|
| 247 |
+
cfg["remote_root"] / policy / str(step) / "checkpoint_manifest.json")}
|
| 248 |
+
|
| 249 |
+
|
| 250 |
+
def once(api, cfg, state, selected=None, selected_step=None, audit=False):
|
| 251 |
+
plan = checkpoint_plan(cfg)
|
| 252 |
+
for step in STEPS:
|
| 253 |
+
for policy in POLICIES:
|
| 254 |
+
if (policy, step) not in plan:
|
| 255 |
+
continue
|
| 256 |
+
if selected and policy != selected or selected_step and step != selected_step:
|
| 257 |
+
continue
|
| 258 |
+
key = f"{policy}/{step}"
|
| 259 |
+
prior = state["checkpoints"].get(key, {})
|
| 260 |
+
same_destination = prior.get("prefix") == remote_prefix(cfg, policy, step)
|
| 261 |
+
if prior.get("status") == "verified" and same_destination and not audit:
|
| 262 |
+
continue
|
| 263 |
+
if prior.get("retry_after", 0) > time.time() and same_destination and not audit:
|
| 264 |
+
continue
|
| 265 |
+
state.update(current=key, heartbeat=time.time(), status="running")
|
| 266 |
+
atomic_json(Path(cfg["state_dir"]) / "status.json", state)
|
| 267 |
+
try:
|
| 268 |
+
record = publish(api, cfg, policy, step)
|
| 269 |
+
except (NotReady, FileNotFoundError, json.JSONDecodeError) as exc:
|
| 270 |
+
record = {"status": "waiting", "reason": str(exc), "checked_at": time.time()}
|
| 271 |
+
except Exception as exc:
|
| 272 |
+
# Never serialize tokens/headers or HTTP exception payloads.
|
| 273 |
+
attempts = prior.get("attempts", 0) + 1
|
| 274 |
+
record = {"status": "error", "error_type": type(exc).__name__,
|
| 275 |
+
"reason": str(exc)[:500] if isinstance(exc, ValueError) else "See service type; retry scheduled",
|
| 276 |
+
"attempts": attempts, "retry_after": time.time() + min(1800, 60 * 2 ** min(attempts, 5))}
|
| 277 |
+
record["prefix"] = remote_prefix(cfg, policy, step)
|
| 278 |
+
state["checkpoints"][key] = record
|
| 279 |
+
state["heartbeat"] = time.time()
|
| 280 |
+
atomic_json(Path(cfg["state_dir"]) / "status.json", state)
|
| 281 |
+
print(json.dumps({"checkpoint": key, **record}), flush=True)
|
| 282 |
+
state["current"] = None
|
| 283 |
+
records = [state["checkpoints"].get(f"{p}/{s}", {}) for p, s in plan]
|
| 284 |
+
state["expected_checkpoints"] = len(plan)
|
| 285 |
+
state["verified_count"] = sum(x.get("status") == "verified" and
|
| 286 |
+
x.get("prefix") == remote_prefix(cfg, p, s) for (p, s), x in zip(plan, records))
|
| 287 |
+
if state["verified_count"] == len(plan):
|
| 288 |
+
# Recheck the entire set against ONE current HF revision before declaring
|
| 289 |
+
# completion, including checkpoints uploaded many hours earlier.
|
| 290 |
+
try:
|
| 291 |
+
revision = api.model_info(cfg["repo_id"]).sha
|
| 292 |
+
for record in records:
|
| 293 |
+
manifest_path = Path(record["manifest"])
|
| 294 |
+
manifest = read(manifest_path)
|
| 295 |
+
files = dict(manifest["files"])
|
| 296 |
+
files["checkpoint_manifest.json"] = file_info(manifest_path)
|
| 297 |
+
check = compare_remote(api, cfg["repo_id"], record["prefix"], revision, files)
|
| 298 |
+
if not matches(check):
|
| 299 |
+
raise ValueError("Completion audit failed: " + record["prefix"])
|
| 300 |
+
state["completion_audit"] = {"status": "verified", "revision": revision,
|
| 301 |
+
"checkpoints": len(plan), "missing": 0, "extra": 0, "mismatch": 0}
|
| 302 |
+
except Exception as exc:
|
| 303 |
+
state["completion_audit"] = {"status": "error", "error_type": type(exc).__name__}
|
| 304 |
+
state["status"] = "complete" if state["verified_count"] == len(plan) and state.get("completion_audit", {}).get("status") == "verified" else "watching"
|
| 305 |
+
state["attention"] = [k for k, v in state["checkpoints"].items() if v.get("status") == "error"]
|
| 306 |
+
if state.get("completion_audit", {}).get("status") == "error":
|
| 307 |
+
state["attention"].append("completion_audit_failed")
|
| 308 |
+
for policy, path in cfg["results"].items():
|
| 309 |
+
try:
|
| 310 |
+
if read(path)["status"] == "failed":
|
| 311 |
+
state["attention"].append(policy + ":training_failed")
|
| 312 |
+
except (FileNotFoundError, json.JSONDecodeError):
|
| 313 |
+
try:
|
| 314 |
+
controller = read(Path(cfg["run_dir"]) / "controller_status.json")
|
| 315 |
+
if controller.get("jobs", {}).get(policy, {}).get("state") == "failed":
|
| 316 |
+
state["attention"].append(policy + ":training_failed")
|
| 317 |
+
except (FileNotFoundError, json.JSONDecodeError):
|
| 318 |
+
pass
|
| 319 |
+
state["heartbeat"] = time.time()
|
| 320 |
+
atomic_json(Path(cfg["state_dir"]) / "status.json", state)
|
| 321 |
+
|
| 322 |
+
|
| 323 |
+
def main():
|
| 324 |
+
parser = argparse.ArgumentParser(__doc__)
|
| 325 |
+
parser.add_argument("--config", required=True, type=Path)
|
| 326 |
+
parser.add_argument("--once", action="store_true")
|
| 327 |
+
parser.add_argument("--policy", choices=POLICIES)
|
| 328 |
+
parser.add_argument("--step", type=int, choices=STEPS)
|
| 329 |
+
parser.add_argument("--audit-remote", action="store_true")
|
| 330 |
+
args = parser.parse_args()
|
| 331 |
+
os.environ["CUDA_VISIBLE_DEVICES"] = ""
|
| 332 |
+
os.environ["HF_HUB_OFFLINE"] = "0"
|
| 333 |
+
os.environ["TRANSFORMERS_OFFLINE"] = "0"
|
| 334 |
+
os.environ["HF_HUB_DISABLE_PROGRESS_BARS"] = "1"
|
| 335 |
+
os.environ["OMP_NUM_THREADS"] = "2"
|
| 336 |
+
os.environ["MKL_NUM_THREADS"] = "2"
|
| 337 |
+
from huggingface_hub import HfApi
|
| 338 |
+
cfg = read(args.config)
|
| 339 |
+
state_dir = Path(cfg["state_dir"])
|
| 340 |
+
state_dir.mkdir(parents=True, exist_ok=True)
|
| 341 |
+
with (state_dir / "watcher.lock").open("w") as lock:
|
| 342 |
+
fcntl.flock(lock, fcntl.LOCK_EX | fcntl.LOCK_NB)
|
| 343 |
+
token_file = cfg.get('token_file')
|
| 344 |
+
api = HfApi(token=Path(token_file).read_text().strip() if token_file else None)
|
| 345 |
+
identity = api.whoami()
|
| 346 |
+
if identity["name"] != cfg["hf_account"]:
|
| 347 |
+
raise ValueError("Unexpected HF account")
|
| 348 |
+
if not any(o["name"] == "RwH-Lab" and o.get("roleInOrg") in ("write", "admin")
|
| 349 |
+
for o in identity.get("orgs", [])):
|
| 350 |
+
raise ValueError("Missing HF organization write permission")
|
| 351 |
+
# User authorized a new checkpoint repo; keep private on creation.
|
| 352 |
+
api.create_repo(cfg["repo_id"], repo_type="model", private=True, exist_ok=True)
|
| 353 |
+
if not api.model_info(cfg["repo_id"]).private:
|
| 354 |
+
raise ValueError("Expected a private checkpoint repository")
|
| 355 |
+
status_path = state_dir / "status.json"
|
| 356 |
+
state = read(status_path) if status_path.exists() else {"schema": 1, "checkpoints": {}}
|
| 357 |
+
if state.get("run_dir", cfg["run_dir"]) != cfg["run_dir"]:
|
| 358 |
+
raise ValueError("State belongs to a different run")
|
| 359 |
+
state.update(repo_id=cfg["repo_id"], run_dir=cfg["run_dir"], pid=os.getpid(),
|
| 360 |
+
expected_checkpoints=len(checkpoint_plan(cfg)), remote_root=cfg["remote_root"],
|
| 361 |
+
started_at=state.get("started_at", time.time()))
|
| 362 |
+
while True:
|
| 363 |
+
once(api, cfg, state, args.policy, args.step, args.audit_remote)
|
| 364 |
+
if args.once or state["status"] == "complete":
|
| 365 |
+
break
|
| 366 |
+
time.sleep(cfg["poll_seconds"])
|
| 367 |
+
print("HF_WATCHER_EXIT", state["status"], state["verified_count"], flush=True)
|
| 368 |
+
|
| 369 |
+
|
| 370 |
+
if __name__ == "__main__":
|
| 371 |
+
main()
|
autostep_v2_bs128_30k/gr00t_n17/checkpoints/checkpoint-30K/provenance/scripts/train/native_assets.py
ADDED
|
@@ -0,0 +1,48 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
"""Pin local assets while preserving the native backbone-name routing contract."""
|
| 2 |
+
import os
|
| 3 |
+
from pathlib import Path
|
| 4 |
+
|
| 5 |
+
|
| 6 |
+
def pinned_cosmos_name():
|
| 7 |
+
from huggingface_hub import try_to_load_from_cache
|
| 8 |
+
name = "nvidia/Cosmos-Reason2-2B"
|
| 9 |
+
config = try_to_load_from_cache(name, "config.json", revision="main")
|
| 10 |
+
if not isinstance(config, str) or Path(config).parent.name != "9ce19a195e423419c349abfc86fd07178b230561":
|
| 11 |
+
raise RuntimeError("Cosmos cached main differs from the approved revision")
|
| 12 |
+
return name
|
| 13 |
+
|
| 14 |
+
|
| 15 |
+
def local_cosmos():
|
| 16 |
+
from huggingface_hub import snapshot_download
|
| 17 |
+
revision = "9ce19a195e423419c349abfc86fd07178b230561"
|
| 18 |
+
snapshot = Path(snapshot_download("nvidia/Cosmos-Reason2-2B", revision=revision, local_files_only=True))
|
| 19 |
+
# Native get_backbone_cls recognizes the nvidia/Cosmos-Reason2 substring.
|
| 20 |
+
alias = Path(os.environ.get("HF_HOME", str(Path.home()/".cache/huggingface"))) / "rwh_pinned/nvidia/Cosmos-Reason2-2B" / revision
|
| 21 |
+
alias.parent.mkdir(parents=True, exist_ok=True)
|
| 22 |
+
if alias.is_symlink():
|
| 23 |
+
if alias.resolve() != snapshot.resolve():
|
| 24 |
+
raise RuntimeError("Pinned Cosmos alias points at another snapshot")
|
| 25 |
+
elif alias.exists():
|
| 26 |
+
raise RuntimeError("Pinned Cosmos alias is occupied")
|
| 27 |
+
else:
|
| 28 |
+
try: alias.symlink_to(snapshot, target_is_directory=True)
|
| 29 |
+
except FileExistsError:
|
| 30 |
+
if alias.resolve() != snapshot.resolve(): raise
|
| 31 |
+
return str(alias)
|
| 32 |
+
|
| 33 |
+
|
| 34 |
+
def verify_legacy_refs(policy):
|
| 35 |
+
from huggingface_hub import try_to_load_from_cache
|
| 36 |
+
assets = {
|
| 37 |
+
"nvidia/GR00T-N1.5-3B": "869830fc749c35f34771aa5209f923ac57e4564e",
|
| 38 |
+
"openai/clip-vit-base-patch16": "57c216476eefef5ab752ec549e440a49ae4ae5f3",
|
| 39 |
+
"lerobot/eagle2hg-processor-groot-n1p5": "baf604d8a5caf26fda5cc545f141bc1814156237",
|
| 40 |
+
}
|
| 41 |
+
repos = ["openai/clip-vit-base-patch16"] if policy == "multi_task_dit" else ["nvidia/GR00T-N1.5-3B"]
|
| 42 |
+
for repo in repos:
|
| 43 |
+
current = try_to_load_from_cache(repo, "config.json", revision="main")
|
| 44 |
+
pinned = try_to_load_from_cache(repo, "config.json", revision=assets[repo])
|
| 45 |
+
# DiT receives the explicit pinned CLIP snapshot path in its config.
|
| 46 |
+
if not isinstance(pinned, str) or (policy != "multi_task_dit" and
|
| 47 |
+
(not isinstance(current, str) or Path(current).parent.name != assets[repo])):
|
| 48 |
+
raise RuntimeError("Cached main is not the approved model revision: " + repo)
|
autostep_v2_bs128_30k/gr00t_n17/checkpoints/checkpoint-30K/provenance/scripts/train/run_n17.py
ADDED
|
@@ -0,0 +1,173 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
"""Configure and call LeTools' bundled N1.7 pipeline and HF Trainer."""
|
| 2 |
+
import argparse
|
| 3 |
+
import json
|
| 4 |
+
import math
|
| 5 |
+
import os
|
| 6 |
+
from pathlib import Path
|
| 7 |
+
import sys
|
| 8 |
+
import time
|
| 9 |
+
|
| 10 |
+
ROOT=Path(os.environ['RWH_TRAIN_CONFIG_DIR']).resolve()
|
| 11 |
+
LETOOLS=ROOT.parents[3]/'LeTools-Learning'
|
| 12 |
+
|
| 13 |
+
|
| 14 |
+
def write(path,value):
|
| 15 |
+
path=Path(path); path.parent.mkdir(parents=True,exist_ok=True)
|
| 16 |
+
tmp=path.with_suffix('.tmp');tmp.write_text(json.dumps(value,indent=2,ensure_ascii=False)+'\n');tmp.replace(path)
|
| 17 |
+
|
| 18 |
+
|
| 19 |
+
def main():
|
| 20 |
+
parser=argparse.ArgumentParser()
|
| 21 |
+
parser.add_argument('--config',type=Path,default=ROOT/'gr00t_n17.yaml')
|
| 22 |
+
parser.add_argument('--gpu',required=True,choices=['0','1'])
|
| 23 |
+
parser.add_argument('--smoke-steps',type=int,default=0)
|
| 24 |
+
args=parser.parse_args()
|
| 25 |
+
os.environ['CUDA_VISIBLE_DEVICES']=args.gpu
|
| 26 |
+
os.environ['NO_ALBUMENTATIONS_UPDATE']='1'
|
| 27 |
+
sys.path.insert(0,str(LETOOLS/'kuavo_model/external_models/gr00tn1d7'))
|
| 28 |
+
import yaml
|
| 29 |
+
import numpy as np
|
| 30 |
+
import torch
|
| 31 |
+
torch.set_num_threads(8)
|
| 32 |
+
if args.smoke_steps:
|
| 33 |
+
import faulthandler
|
| 34 |
+
faulthandler.dump_traceback_later(90, repeat=True)
|
| 35 |
+
from huggingface_hub import snapshot_download
|
| 36 |
+
from transformers import TrainerCallback
|
| 37 |
+
from scripts.data.view import DataView
|
| 38 |
+
from scripts.adapters import groot_v3
|
| 39 |
+
from gr00t.configs.base_config import get_default_config
|
| 40 |
+
from gr00t.configs.data.data_config import SingleDatasetConfig
|
| 41 |
+
import gr00t.experiment.experiment as experiment
|
| 42 |
+
from scripts.train.native_assets import pinned_cosmos_name
|
| 43 |
+
|
| 44 |
+
config_path=args.config.expanduser().resolve()
|
| 45 |
+
simple=yaml.safe_load(config_path.read_text())
|
| 46 |
+
view=DataView(config_path.parent/simple['dataset']['config'])
|
| 47 |
+
t,p=simple['training'],simple['policy']
|
| 48 |
+
if t['batch_size']!=128 or t['gradient_accumulation_steps']!=1 or t['resume']:
|
| 49 |
+
raise ValueError('Expected fresh actual batch128/accumulation1')
|
| 50 |
+
if p['chunk_size']!=40 or p['n_action_steps']!=40: raise ValueError('Expected 40 steps')
|
| 51 |
+
if int(os.environ.get('WORLD_SIZE','1'))!=1: raise ValueError('One GPU per model required')
|
| 52 |
+
output=Path(t['output_dir'])
|
| 53 |
+
if args.smoke_steps: output=output.parent/'_validation'/('gr00t_n17_'+time.strftime('%Y%m%d_%H%M%S'))
|
| 54 |
+
if output.exists(): raise FileExistsError(str(output))
|
| 55 |
+
log=Path(os.environ.get('RWH_RESULT_DIR',str(output.parent/'_runs/gr00t_n17')))
|
| 56 |
+
report={'status':'loading','policy':'gr00t_n17','gpu':int(args.gpu),'batch_size':128,
|
| 57 |
+
'gradient_accumulation_steps':1,'steps_requested':args.smoke_steps or t['steps'],
|
| 58 |
+
'smoke':bool(args.smoke_steps),'output_dir':str(output),'measurements':[],
|
| 59 |
+
'data_fingerprint':view.fingerprint,'source_revision':view.config['source']['revision'],
|
| 60 |
+
'train_config':str(config_path),
|
| 61 |
+
'runtime_root':str(LETOOLS/'kuavo_model/external_models/gr00tn1d7'),
|
| 62 |
+
'trainer_source':experiment.__file__}
|
| 63 |
+
result=log/'result.json';write(result,report);view.save_audit(log/'data')
|
| 64 |
+
base=snapshot_download(p['base_model_path'],revision=p['base_revision'],local_files_only=True)
|
| 65 |
+
cfg=get_default_config(); cfg.load_config_path=None
|
| 66 |
+
cfg.data.datasets=[SingleDatasetConfig(dataset_paths=[str(view.root)],mix_ratio=1.0,embodiment_tag='new_embodiment')]
|
| 67 |
+
cfg.data.modality_configs=groot_v3.modality(view,40)
|
| 68 |
+
cfg.data.seed=42;cfg.data.allow_padding=False;cfg.data.video_backend='torchcodec'
|
| 69 |
+
cfg.data.shard_size=t['shard_size'];cfg.data.episode_sampling_rate=t['episode_sampling_rate']
|
| 70 |
+
cfg.data.num_shards_per_epoch=t['num_shards_per_epoch'];cfg.data.download_cache=False
|
| 71 |
+
cfg.data.multiprocessing_context='fork'
|
| 72 |
+
cfg.model.model_name=pinned_cosmos_name()
|
| 73 |
+
cfg.model.action_horizon=40
|
| 74 |
+
cfg.model.tune_llm=False;cfg.model.tune_visual=False
|
| 75 |
+
cfg.model.tune_projector=True;cfg.model.tune_diffusion_model=True;cfg.model.tune_vlln=True
|
| 76 |
+
cfg.model.load_bf16=True;cfg.model.backbone_trainable_params_fp32=True
|
| 77 |
+
cfg.model.state_dropout_prob=p['state_dropout_prob']
|
| 78 |
+
cfg.model.use_relative_action=False
|
| 79 |
+
cfg.training.output_dir=str(output);cfg.training.experiment_name=None
|
| 80 |
+
cfg.training.start_from_checkpoint=base
|
| 81 |
+
cfg.training.global_batch_size=128;cfg.training.gradient_accumulation_steps=1
|
| 82 |
+
cfg.training.num_gpus=1;cfg.training.max_steps=args.smoke_steps or t['steps']
|
| 83 |
+
cfg.training.save_steps=args.smoke_steps or t['save_freq'];cfg.training.save_total_limit=3
|
| 84 |
+
cfg.training.learning_rate=t['optimizer']['lr'];cfg.training.optim='adamw_torch'
|
| 85 |
+
cfg.training.weight_decay=t['optimizer']['weight_decay']
|
| 86 |
+
cfg.training.max_grad_norm=t['optimizer']['grad_clip_norm']
|
| 87 |
+
cfg.training.warmup_ratio=t['scheduler']['warmup_ratio'];cfg.training.logging_steps=1 if args.smoke_steps else t['log_freq']
|
| 88 |
+
cfg.training.dataloader_num_workers=t['num_workers']
|
| 89 |
+
cfg.training.transformers_local_files_only=True
|
| 90 |
+
cfg.training.use_wandb=bool(simple['wandb']['enable'] and not args.smoke_steps)
|
| 91 |
+
cfg.training.wandb_project=simple['wandb']['project']
|
| 92 |
+
os.environ['WANDB_ENTITY']=simple['wandb']['entity']
|
| 93 |
+
groot_v3.install(view)
|
| 94 |
+
original=experiment.Gr00tTrainer
|
| 95 |
+
holder={}
|
| 96 |
+
|
| 97 |
+
class Progress(TrainerCallback):
|
| 98 |
+
def on_step_end(self,args,state,control,**kw):
|
| 99 |
+
loss=float(holder['last_loss'])
|
| 100 |
+
if not math.isfinite(loss): raise ValueError('Nonfinite loss')
|
| 101 |
+
report.update(status='running',last_step=state.global_step,heartbeat=time.time())
|
| 102 |
+
if state.global_step==1 or state.global_step%args.logging_steps==0 or report['smoke']:
|
| 103 |
+
row={'step':state.global_step,'loss':loss,'elapsed_s':time.time()-holder['started'],
|
| 104 |
+
'max_allocated_gib':torch.cuda.max_memory_allocated()/2**30}
|
| 105 |
+
report['measurements'].append(row);write(result,report)
|
| 106 |
+
print('RWH_STEP',json.dumps(row),flush=True)
|
| 107 |
+
def on_log(self,args,state,control,logs=None,**kw):
|
| 108 |
+
if logs and 'grad_norm' in logs and not math.isfinite(float(logs['grad_norm'])):
|
| 109 |
+
raise ValueError('Nonfinite gradient norm')
|
| 110 |
+
def on_train_end(self,args,state,control,**kw):
|
| 111 |
+
from scripts.train.checkpoints import validate_native_checkpoint
|
| 112 |
+
path=Path(args.output_dir)/f'checkpoint-{state.global_step}'
|
| 113 |
+
validate_native_checkpoint(path,state.global_step)
|
| 114 |
+
write(path/'rwh_save_complete.json',{'step':state.global_step,'fingerprint':view.fingerprint})
|
| 115 |
+
report.update(checkpoint=str(path),saved_step=state.global_step);write(result,report)
|
| 116 |
+
|
| 117 |
+
def create_trainer(*a,**kw):
|
| 118 |
+
trainer=original(*a,**kw)
|
| 119 |
+
# The instance is the unmodified native Gr00tTrainer. Only add a standard
|
| 120 |
+
# progress callback and observational model hooks; no loss/update overrides.
|
| 121 |
+
trainer.add_callback(Progress());holder['trainer']=trainer
|
| 122 |
+
report.update(parameters=sum(p.numel() for p in trainer.model.parameters()),
|
| 123 |
+
trainable_parameters=sum(p.numel() for p in trainer.model.parameters() if p.requires_grad))
|
| 124 |
+
def before(model,inputs,kwargs):
|
| 125 |
+
b=kwargs.get('inputs',inputs[0] if inputs else None)
|
| 126 |
+
if b['action'].shape!=(128,40,132):raise ValueError(f"Invalid batch: {b['action'].shape}")
|
| 127 |
+
if b['action_mask'][...,23:].any():raise ValueError('Padding entered loss')
|
| 128 |
+
holder['mask']=b['action_mask']
|
| 129 |
+
if 'batch_shapes' not in report:
|
| 130 |
+
report['batch_shapes']={k:list(v.shape) for k,v in b.items() if torch.is_tensor(v)}
|
| 131 |
+
report['masked_target_elements']=int((~b['action_mask'][...,:23].bool()).sum())
|
| 132 |
+
write(result,report);print('RWH_NATIVE_FIRST_BATCH',report['batch_shapes'],flush=True)
|
| 133 |
+
def after(model,inputs,output):
|
| 134 |
+
holder['last_loss']=output['loss'].detach().float().cpu().item()
|
| 135 |
+
holder['before_handle']=trainer.model.register_forward_pre_hook(before,with_kwargs=True)
|
| 136 |
+
holder['after_handle']=trainer.model.register_forward_hook(after)
|
| 137 |
+
if args.smoke_steps:
|
| 138 |
+
def capture(module,inputs,output):
|
| 139 |
+
if not output.requires_grad:return
|
| 140 |
+
mask=holder['mask'].bool()
|
| 141 |
+
def check(grad):
|
| 142 |
+
tail=grad[:,-mask.shape[1]:]
|
| 143 |
+
if tail[~mask].abs().sum().item()!=0 or tail[mask].abs().sum().item()==0:
|
| 144 |
+
raise ValueError('Native action-mask gradient check failed')
|
| 145 |
+
report['masked_gradient_checks']=report.get('masked_gradient_checks',0)+1
|
| 146 |
+
return grad
|
| 147 |
+
output.register_hook(check)
|
| 148 |
+
trainer.model.action_head.action_decoder.register_forward_hook(capture)
|
| 149 |
+
return trainer
|
| 150 |
+
experiment.Gr00tTrainer=create_trainer
|
| 151 |
+
try:
|
| 152 |
+
holder['started']=time.time();experiment.run(cfg)
|
| 153 |
+
holder['before_handle'].remove();holder['after_handle'].remove()
|
| 154 |
+
trainer=holder['trainer'];model=trainer.model;model.eval()
|
| 155 |
+
processor=trainer.train_dataset.processor;processor.eval()
|
| 156 |
+
ds=trainer.train_dataset.datasets[0]
|
| 157 |
+
i=next(i for i,s in enumerate(view.segments) if s['end']-s['start']>=40)
|
| 158 |
+
df=ds.episode_loader[i]
|
| 159 |
+
sample=ds.get_datapoint(df,0)
|
| 160 |
+
one=processor.collator([sample])['inputs'];one.pop('action',None);one.pop('action_mask',None)
|
| 161 |
+
one={k:v.to(model.device) for k,v in one.items()}
|
| 162 |
+
with torch.inference_mode(),torch.autocast('cuda',dtype=torch.bfloat16):
|
| 163 |
+
predicted=model.get_action(one)['action_pred']
|
| 164 |
+
if predicted.shape!=(1,40,132) or not torch.isfinite(predicted).all():raise ValueError('Invalid prediction')
|
| 165 |
+
report.update(status='passed',prediction_shape=list(predicted.shape),completed_at=time.time())
|
| 166 |
+
except BaseException as exc:
|
| 167 |
+
report.update(status='failed',error_type=type(exc).__name__,error=str(exc)[-2500:]);raise
|
| 168 |
+
finally:
|
| 169 |
+
write(result,report)
|
| 170 |
+
if args.smoke_steps:faulthandler.cancel_dump_traceback_later()
|
| 171 |
+
|
| 172 |
+
|
| 173 |
+
if __name__=='__main__': main()
|
autostep_v2_bs128_30k/gr00t_n17/checkpoints/checkpoint-30K/rwh_save_complete.json
ADDED
|
@@ -0,0 +1,4 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"step": 30000,
|
| 3 |
+
"fingerprint": "81f3787f08f2a596f6f22aa087eb414d27264756b661467a3a88b59495401888"
|
| 4 |
+
}
|
autostep_v2_bs128_30k/gr00t_n17/checkpoints/checkpoint-30K/statistics.json
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
autostep_v2_bs128_30k/gr00t_n17/checkpoints/checkpoint-30K/trainer_state.json
ADDED
|
@@ -0,0 +1,1834 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"best_global_step": null,
|
| 3 |
+
"best_metric": null,
|
| 4 |
+
"best_model_checkpoint": null,
|
| 5 |
+
"epoch": 1.0,
|
| 6 |
+
"eval_steps": 500,
|
| 7 |
+
"global_step": 30000,
|
| 8 |
+
"is_hyper_param_search": false,
|
| 9 |
+
"is_local_process_zero": true,
|
| 10 |
+
"is_world_process_zero": true,
|
| 11 |
+
"log_history": [
|
| 12 |
+
{
|
| 13 |
+
"grad_norm": 0.5429198145866394,
|
| 14 |
+
"learning_rate": 6.6e-06,
|
| 15 |
+
"loss": 1.2813,
|
| 16 |
+
"step": 100
|
| 17 |
+
},
|
| 18 |
+
{
|
| 19 |
+
"grad_norm": 1.0767709016799927,
|
| 20 |
+
"learning_rate": 1.3266666666666666e-05,
|
| 21 |
+
"loss": 0.9844,
|
| 22 |
+
"step": 200
|
| 23 |
+
},
|
| 24 |
+
{
|
| 25 |
+
"grad_norm": 1.1089611053466797,
|
| 26 |
+
"learning_rate": 1.9933333333333334e-05,
|
| 27 |
+
"loss": 0.3765,
|
| 28 |
+
"step": 300
|
| 29 |
+
},
|
| 30 |
+
{
|
| 31 |
+
"grad_norm": 1.4204347133636475,
|
| 32 |
+
"learning_rate": 2.6600000000000003e-05,
|
| 33 |
+
"loss": 0.1337,
|
| 34 |
+
"step": 400
|
| 35 |
+
},
|
| 36 |
+
{
|
| 37 |
+
"grad_norm": 0.8840071558952332,
|
| 38 |
+
"learning_rate": 3.326666666666667e-05,
|
| 39 |
+
"loss": 0.0933,
|
| 40 |
+
"step": 500
|
| 41 |
+
},
|
| 42 |
+
{
|
| 43 |
+
"grad_norm": 1.0456535816192627,
|
| 44 |
+
"learning_rate": 3.993333333333333e-05,
|
| 45 |
+
"loss": 0.0785,
|
| 46 |
+
"step": 600
|
| 47 |
+
},
|
| 48 |
+
{
|
| 49 |
+
"grad_norm": 1.394777774810791,
|
| 50 |
+
"learning_rate": 4.660000000000001e-05,
|
| 51 |
+
"loss": 0.0726,
|
| 52 |
+
"step": 700
|
| 53 |
+
},
|
| 54 |
+
{
|
| 55 |
+
"grad_norm": 1.324737310409546,
|
| 56 |
+
"learning_rate": 5.326666666666666e-05,
|
| 57 |
+
"loss": 0.0665,
|
| 58 |
+
"step": 800
|
| 59 |
+
},
|
| 60 |
+
{
|
| 61 |
+
"grad_norm": 1.1846684217453003,
|
| 62 |
+
"learning_rate": 5.9933333333333345e-05,
|
| 63 |
+
"loss": 0.063,
|
| 64 |
+
"step": 900
|
| 65 |
+
},
|
| 66 |
+
{
|
| 67 |
+
"grad_norm": 0.949142575263977,
|
| 68 |
+
"learning_rate": 6.66e-05,
|
| 69 |
+
"loss": 0.0576,
|
| 70 |
+
"step": 1000
|
| 71 |
+
},
|
| 72 |
+
{
|
| 73 |
+
"grad_norm": 0.9336966872215271,
|
| 74 |
+
"learning_rate": 7.326666666666667e-05,
|
| 75 |
+
"loss": 0.0542,
|
| 76 |
+
"step": 1100
|
| 77 |
+
},
|
| 78 |
+
{
|
| 79 |
+
"grad_norm": 1.1244068145751953,
|
| 80 |
+
"learning_rate": 7.993333333333334e-05,
|
| 81 |
+
"loss": 0.0513,
|
| 82 |
+
"step": 1200
|
| 83 |
+
},
|
| 84 |
+
{
|
| 85 |
+
"grad_norm": 0.8189104795455933,
|
| 86 |
+
"learning_rate": 8.66e-05,
|
| 87 |
+
"loss": 0.0498,
|
| 88 |
+
"step": 1300
|
| 89 |
+
},
|
| 90 |
+
{
|
| 91 |
+
"grad_norm": 0.773094892501831,
|
| 92 |
+
"learning_rate": 9.326666666666667e-05,
|
| 93 |
+
"loss": 0.0454,
|
| 94 |
+
"step": 1400
|
| 95 |
+
},
|
| 96 |
+
{
|
| 97 |
+
"grad_norm": 0.7437832951545715,
|
| 98 |
+
"learning_rate": 9.993333333333334e-05,
|
| 99 |
+
"loss": 0.0453,
|
| 100 |
+
"step": 1500
|
| 101 |
+
},
|
| 102 |
+
{
|
| 103 |
+
"grad_norm": 0.6425749063491821,
|
| 104 |
+
"learning_rate": 9.999702274445236e-05,
|
| 105 |
+
"loss": 0.0418,
|
| 106 |
+
"step": 1600
|
| 107 |
+
},
|
| 108 |
+
{
|
| 109 |
+
"grad_norm": 0.6502171158790588,
|
| 110 |
+
"learning_rate": 9.998797074387361e-05,
|
| 111 |
+
"loss": 0.0391,
|
| 112 |
+
"step": 1700
|
| 113 |
+
},
|
| 114 |
+
{
|
| 115 |
+
"grad_norm": 0.7395640015602112,
|
| 116 |
+
"learning_rate": 9.997284479438253e-05,
|
| 117 |
+
"loss": 0.038,
|
| 118 |
+
"step": 1800
|
| 119 |
+
},
|
| 120 |
+
{
|
| 121 |
+
"grad_norm": 0.6537087559700012,
|
| 122 |
+
"learning_rate": 9.995164673390625e-05,
|
| 123 |
+
"loss": 0.0358,
|
| 124 |
+
"step": 1900
|
| 125 |
+
},
|
| 126 |
+
{
|
| 127 |
+
"grad_norm": 0.6031444668769836,
|
| 128 |
+
"learning_rate": 9.992437913818312e-05,
|
| 129 |
+
"loss": 0.0346,
|
| 130 |
+
"step": 2000
|
| 131 |
+
},
|
| 132 |
+
{
|
| 133 |
+
"grad_norm": 0.5211590528488159,
|
| 134 |
+
"learning_rate": 9.989104532044994e-05,
|
| 135 |
+
"loss": 0.0323,
|
| 136 |
+
"step": 2100
|
| 137 |
+
},
|
| 138 |
+
{
|
| 139 |
+
"grad_norm": 0.4885427951812744,
|
| 140 |
+
"learning_rate": 9.985164933103929e-05,
|
| 141 |
+
"loss": 0.0308,
|
| 142 |
+
"step": 2200
|
| 143 |
+
},
|
| 144 |
+
{
|
| 145 |
+
"grad_norm": 0.6266382932662964,
|
| 146 |
+
"learning_rate": 9.980619595688737e-05,
|
| 147 |
+
"loss": 0.0302,
|
| 148 |
+
"step": 2300
|
| 149 |
+
},
|
| 150 |
+
{
|
| 151 |
+
"grad_norm": 0.4790324866771698,
|
| 152 |
+
"learning_rate": 9.975469072095237e-05,
|
| 153 |
+
"loss": 0.0292,
|
| 154 |
+
"step": 2400
|
| 155 |
+
},
|
| 156 |
+
{
|
| 157 |
+
"grad_norm": 0.43936362862586975,
|
| 158 |
+
"learning_rate": 9.969713988154339e-05,
|
| 159 |
+
"loss": 0.0285,
|
| 160 |
+
"step": 2500
|
| 161 |
+
},
|
| 162 |
+
{
|
| 163 |
+
"grad_norm": 0.485794335603714,
|
| 164 |
+
"learning_rate": 9.963355043155996e-05,
|
| 165 |
+
"loss": 0.0265,
|
| 166 |
+
"step": 2600
|
| 167 |
+
},
|
| 168 |
+
{
|
| 169 |
+
"grad_norm": 0.41582396626472473,
|
| 170 |
+
"learning_rate": 9.956393009764244e-05,
|
| 171 |
+
"loss": 0.0244,
|
| 172 |
+
"step": 2700
|
| 173 |
+
},
|
| 174 |
+
{
|
| 175 |
+
"grad_norm": 0.4837326407432556,
|
| 176 |
+
"learning_rate": 9.948828733923305e-05,
|
| 177 |
+
"loss": 0.0238,
|
| 178 |
+
"step": 2800
|
| 179 |
+
},
|
| 180 |
+
{
|
| 181 |
+
"grad_norm": 0.35663583874702454,
|
| 182 |
+
"learning_rate": 9.94066313475481e-05,
|
| 183 |
+
"loss": 0.0236,
|
| 184 |
+
"step": 2900
|
| 185 |
+
},
|
| 186 |
+
{
|
| 187 |
+
"grad_norm": 0.4235639274120331,
|
| 188 |
+
"learning_rate": 9.931897204446104e-05,
|
| 189 |
+
"loss": 0.0231,
|
| 190 |
+
"step": 3000
|
| 191 |
+
},
|
| 192 |
+
{
|
| 193 |
+
"grad_norm": 0.518033504486084,
|
| 194 |
+
"learning_rate": 9.9225320081297e-05,
|
| 195 |
+
"loss": 0.0217,
|
| 196 |
+
"step": 3100
|
| 197 |
+
},
|
| 198 |
+
{
|
| 199 |
+
"grad_norm": 0.4571470022201538,
|
| 200 |
+
"learning_rate": 9.912568683753853e-05,
|
| 201 |
+
"loss": 0.0207,
|
| 202 |
+
"step": 3200
|
| 203 |
+
},
|
| 204 |
+
{
|
| 205 |
+
"grad_norm": 0.40404418110847473,
|
| 206 |
+
"learning_rate": 9.902008441944286e-05,
|
| 207 |
+
"loss": 0.021,
|
| 208 |
+
"step": 3300
|
| 209 |
+
},
|
| 210 |
+
{
|
| 211 |
+
"grad_norm": 0.3548375368118286,
|
| 212 |
+
"learning_rate": 9.890852565857092e-05,
|
| 213 |
+
"loss": 0.0215,
|
| 214 |
+
"step": 3400
|
| 215 |
+
},
|
| 216 |
+
{
|
| 217 |
+
"grad_norm": 0.4166719615459442,
|
| 218 |
+
"learning_rate": 9.879102411022817e-05,
|
| 219 |
+
"loss": 0.0195,
|
| 220 |
+
"step": 3500
|
| 221 |
+
},
|
| 222 |
+
{
|
| 223 |
+
"grad_norm": 0.3508767783641815,
|
| 224 |
+
"learning_rate": 9.866759405181765e-05,
|
| 225 |
+
"loss": 0.0194,
|
| 226 |
+
"step": 3600
|
| 227 |
+
},
|
| 228 |
+
{
|
| 229 |
+
"grad_norm": 0.3329998552799225,
|
| 230 |
+
"learning_rate": 9.85382504811049e-05,
|
| 231 |
+
"loss": 0.0193,
|
| 232 |
+
"step": 3700
|
| 233 |
+
},
|
| 234 |
+
{
|
| 235 |
+
"grad_norm": 0.40920156240463257,
|
| 236 |
+
"learning_rate": 9.840300911439591e-05,
|
| 237 |
+
"loss": 0.0187,
|
| 238 |
+
"step": 3800
|
| 239 |
+
},
|
| 240 |
+
{
|
| 241 |
+
"grad_norm": 0.3399767577648163,
|
| 242 |
+
"learning_rate": 9.826188638462718e-05,
|
| 243 |
+
"loss": 0.0194,
|
| 244 |
+
"step": 3900
|
| 245 |
+
},
|
| 246 |
+
{
|
| 247 |
+
"grad_norm": 0.38177067041397095,
|
| 248 |
+
"learning_rate": 9.811489943936922e-05,
|
| 249 |
+
"loss": 0.018,
|
| 250 |
+
"step": 4000
|
| 251 |
+
},
|
| 252 |
+
{
|
| 253 |
+
"grad_norm": 0.46843287348747253,
|
| 254 |
+
"learning_rate": 9.796206613874283e-05,
|
| 255 |
+
"loss": 0.0178,
|
| 256 |
+
"step": 4100
|
| 257 |
+
},
|
| 258 |
+
{
|
| 259 |
+
"grad_norm": 0.3877963423728943,
|
| 260 |
+
"learning_rate": 9.780340505324901e-05,
|
| 261 |
+
"loss": 0.0178,
|
| 262 |
+
"step": 4200
|
| 263 |
+
},
|
| 264 |
+
{
|
| 265 |
+
"grad_norm": 0.42964568734169006,
|
| 266 |
+
"learning_rate": 9.763893546151244e-05,
|
| 267 |
+
"loss": 0.0183,
|
| 268 |
+
"step": 4300
|
| 269 |
+
},
|
| 270 |
+
{
|
| 271 |
+
"grad_norm": 0.32521942257881165,
|
| 272 |
+
"learning_rate": 9.746867734793903e-05,
|
| 273 |
+
"loss": 0.0179,
|
| 274 |
+
"step": 4400
|
| 275 |
+
},
|
| 276 |
+
{
|
| 277 |
+
"grad_norm": 0.36554285883903503,
|
| 278 |
+
"learning_rate": 9.729265140028762e-05,
|
| 279 |
+
"loss": 0.0179,
|
| 280 |
+
"step": 4500
|
| 281 |
+
},
|
| 282 |
+
{
|
| 283 |
+
"grad_norm": 0.41965073347091675,
|
| 284 |
+
"learning_rate": 9.711087900715627e-05,
|
| 285 |
+
"loss": 0.0171,
|
| 286 |
+
"step": 4600
|
| 287 |
+
},
|
| 288 |
+
{
|
| 289 |
+
"grad_norm": 0.3225506842136383,
|
| 290 |
+
"learning_rate": 9.692338225538333e-05,
|
| 291 |
+
"loss": 0.0171,
|
| 292 |
+
"step": 4700
|
| 293 |
+
},
|
| 294 |
+
{
|
| 295 |
+
"grad_norm": 0.38699615001678467,
|
| 296 |
+
"learning_rate": 9.673018392736374e-05,
|
| 297 |
+
"loss": 0.0159,
|
| 298 |
+
"step": 4800
|
| 299 |
+
},
|
| 300 |
+
{
|
| 301 |
+
"grad_norm": 0.2938508987426758,
|
| 302 |
+
"learning_rate": 9.653130749828075e-05,
|
| 303 |
+
"loss": 0.0167,
|
| 304 |
+
"step": 4900
|
| 305 |
+
},
|
| 306 |
+
{
|
| 307 |
+
"grad_norm": 0.30346158146858215,
|
| 308 |
+
"learning_rate": 9.632677713325353e-05,
|
| 309 |
+
"loss": 0.0165,
|
| 310 |
+
"step": 5000
|
| 311 |
+
},
|
| 312 |
+
{
|
| 313 |
+
"grad_norm": 0.3095133304595947,
|
| 314 |
+
"learning_rate": 9.611661768440093e-05,
|
| 315 |
+
"loss": 0.017,
|
| 316 |
+
"step": 5100
|
| 317 |
+
},
|
| 318 |
+
{
|
| 319 |
+
"grad_norm": 0.37841683626174927,
|
| 320 |
+
"learning_rate": 9.590085468782162e-05,
|
| 321 |
+
"loss": 0.0171,
|
| 322 |
+
"step": 5200
|
| 323 |
+
},
|
| 324 |
+
{
|
| 325 |
+
"grad_norm": 0.3431496322154999,
|
| 326 |
+
"learning_rate": 9.56795143604914e-05,
|
| 327 |
+
"loss": 0.0159,
|
| 328 |
+
"step": 5300
|
| 329 |
+
},
|
| 330 |
+
{
|
| 331 |
+
"grad_norm": 0.3221835196018219,
|
| 332 |
+
"learning_rate": 9.545262359707756e-05,
|
| 333 |
+
"loss": 0.0161,
|
| 334 |
+
"step": 5400
|
| 335 |
+
},
|
| 336 |
+
{
|
| 337 |
+
"grad_norm": 0.416565477848053,
|
| 338 |
+
"learning_rate": 9.522020996667092e-05,
|
| 339 |
+
"loss": 0.0153,
|
| 340 |
+
"step": 5500
|
| 341 |
+
},
|
| 342 |
+
{
|
| 343 |
+
"grad_norm": 0.35791927576065063,
|
| 344 |
+
"learning_rate": 9.498230170943596e-05,
|
| 345 |
+
"loss": 0.0159,
|
| 346 |
+
"step": 5600
|
| 347 |
+
},
|
| 348 |
+
{
|
| 349 |
+
"grad_norm": 0.27683424949645996,
|
| 350 |
+
"learning_rate": 9.473892773317952e-05,
|
| 351 |
+
"loss": 0.0153,
|
| 352 |
+
"step": 5700
|
| 353 |
+
},
|
| 354 |
+
{
|
| 355 |
+
"grad_norm": 0.323682576417923,
|
| 356 |
+
"learning_rate": 9.449011760983809e-05,
|
| 357 |
+
"loss": 0.0154,
|
| 358 |
+
"step": 5800
|
| 359 |
+
},
|
| 360 |
+
{
|
| 361 |
+
"grad_norm": 0.2249479591846466,
|
| 362 |
+
"learning_rate": 9.423590157188474e-05,
|
| 363 |
+
"loss": 0.0156,
|
| 364 |
+
"step": 5900
|
| 365 |
+
},
|
| 366 |
+
{
|
| 367 |
+
"grad_norm": 0.30651965737342834,
|
| 368 |
+
"learning_rate": 9.397631050865554e-05,
|
| 369 |
+
"loss": 0.016,
|
| 370 |
+
"step": 6000
|
| 371 |
+
},
|
| 372 |
+
{
|
| 373 |
+
"grad_norm": 0.292286217212677,
|
| 374 |
+
"learning_rate": 9.371137596259623e-05,
|
| 375 |
+
"loss": 0.0149,
|
| 376 |
+
"step": 6100
|
| 377 |
+
},
|
| 378 |
+
{
|
| 379 |
+
"grad_norm": 0.29317429661750793,
|
| 380 |
+
"learning_rate": 9.344113012542963e-05,
|
| 381 |
+
"loss": 0.015,
|
| 382 |
+
"step": 6200
|
| 383 |
+
},
|
| 384 |
+
{
|
| 385 |
+
"grad_norm": 0.28626301884651184,
|
| 386 |
+
"learning_rate": 9.316560583424403e-05,
|
| 387 |
+
"loss": 0.0147,
|
| 388 |
+
"step": 6300
|
| 389 |
+
},
|
| 390 |
+
{
|
| 391 |
+
"grad_norm": 0.2962157428264618,
|
| 392 |
+
"learning_rate": 9.288483656750322e-05,
|
| 393 |
+
"loss": 0.0147,
|
| 394 |
+
"step": 6400
|
| 395 |
+
},
|
| 396 |
+
{
|
| 397 |
+
"grad_norm": 0.3566649854183197,
|
| 398 |
+
"learning_rate": 9.259885644097861e-05,
|
| 399 |
+
"loss": 0.0147,
|
| 400 |
+
"step": 6500
|
| 401 |
+
},
|
| 402 |
+
{
|
| 403 |
+
"grad_norm": 0.2812446653842926,
|
| 404 |
+
"learning_rate": 9.230770020360382e-05,
|
| 405 |
+
"loss": 0.0141,
|
| 406 |
+
"step": 6600
|
| 407 |
+
},
|
| 408 |
+
{
|
| 409 |
+
"grad_norm": 0.29787158966064453,
|
| 410 |
+
"learning_rate": 9.201140323325248e-05,
|
| 411 |
+
"loss": 0.0145,
|
| 412 |
+
"step": 6700
|
| 413 |
+
},
|
| 414 |
+
{
|
| 415 |
+
"grad_norm": 0.3021422028541565,
|
| 416 |
+
"learning_rate": 9.171000153243947e-05,
|
| 417 |
+
"loss": 0.0148,
|
| 418 |
+
"step": 6800
|
| 419 |
+
},
|
| 420 |
+
{
|
| 421 |
+
"grad_norm": 0.34592023491859436,
|
| 422 |
+
"learning_rate": 9.140353172394636e-05,
|
| 423 |
+
"loss": 0.0141,
|
| 424 |
+
"step": 6900
|
| 425 |
+
},
|
| 426 |
+
{
|
| 427 |
+
"grad_norm": 0.26403725147247314,
|
| 428 |
+
"learning_rate": 9.109203104637137e-05,
|
| 429 |
+
"loss": 0.0133,
|
| 430 |
+
"step": 7000
|
| 431 |
+
},
|
| 432 |
+
{
|
| 433 |
+
"grad_norm": 0.27189886569976807,
|
| 434 |
+
"learning_rate": 9.077553734960468e-05,
|
| 435 |
+
"loss": 0.0139,
|
| 436 |
+
"step": 7100
|
| 437 |
+
},
|
| 438 |
+
{
|
| 439 |
+
"grad_norm": 0.2867370843887329,
|
| 440 |
+
"learning_rate": 9.045408909022928e-05,
|
| 441 |
+
"loss": 0.0136,
|
| 442 |
+
"step": 7200
|
| 443 |
+
},
|
| 444 |
+
{
|
| 445 |
+
"grad_norm": 0.2724402844905853,
|
| 446 |
+
"learning_rate": 9.012772532684818e-05,
|
| 447 |
+
"loss": 0.0136,
|
| 448 |
+
"step": 7300
|
| 449 |
+
},
|
| 450 |
+
{
|
| 451 |
+
"grad_norm": 0.3136213421821594,
|
| 452 |
+
"learning_rate": 8.979648571533852e-05,
|
| 453 |
+
"loss": 0.0141,
|
| 454 |
+
"step": 7400
|
| 455 |
+
},
|
| 456 |
+
{
|
| 457 |
+
"grad_norm": 0.2873416543006897,
|
| 458 |
+
"learning_rate": 8.9460410504033e-05,
|
| 459 |
+
"loss": 0.0134,
|
| 460 |
+
"step": 7500
|
| 461 |
+
},
|
| 462 |
+
{
|
| 463 |
+
"grad_norm": 0.328340619802475,
|
| 464 |
+
"learning_rate": 8.91195405288294e-05,
|
| 465 |
+
"loss": 0.0139,
|
| 466 |
+
"step": 7600
|
| 467 |
+
},
|
| 468 |
+
{
|
| 469 |
+
"grad_norm": 0.17781493067741394,
|
| 470 |
+
"learning_rate": 8.877391720822875e-05,
|
| 471 |
+
"loss": 0.014,
|
| 472 |
+
"step": 7700
|
| 473 |
+
},
|
| 474 |
+
{
|
| 475 |
+
"grad_norm": 0.2965923547744751,
|
| 476 |
+
"learning_rate": 8.842358253830246e-05,
|
| 477 |
+
"loss": 0.0139,
|
| 478 |
+
"step": 7800
|
| 479 |
+
},
|
| 480 |
+
{
|
| 481 |
+
"grad_norm": 0.2564098536968231,
|
| 482 |
+
"learning_rate": 8.806857908758967e-05,
|
| 483 |
+
"loss": 0.0138,
|
| 484 |
+
"step": 7900
|
| 485 |
+
},
|
| 486 |
+
{
|
| 487 |
+
"grad_norm": 0.308029443025589,
|
| 488 |
+
"learning_rate": 8.770894999192467e-05,
|
| 489 |
+
"loss": 0.013,
|
| 490 |
+
"step": 8000
|
| 491 |
+
},
|
| 492 |
+
{
|
| 493 |
+
"grad_norm": 0.24848896265029907,
|
| 494 |
+
"learning_rate": 8.734473894919564e-05,
|
| 495 |
+
"loss": 0.0129,
|
| 496 |
+
"step": 8100
|
| 497 |
+
},
|
| 498 |
+
{
|
| 499 |
+
"grad_norm": 0.2171054631471634,
|
| 500 |
+
"learning_rate": 8.697599021403489e-05,
|
| 501 |
+
"loss": 0.0129,
|
| 502 |
+
"step": 8200
|
| 503 |
+
},
|
| 504 |
+
{
|
| 505 |
+
"grad_norm": 0.2320365458726883,
|
| 506 |
+
"learning_rate": 8.660274859244168e-05,
|
| 507 |
+
"loss": 0.0124,
|
| 508 |
+
"step": 8300
|
| 509 |
+
},
|
| 510 |
+
{
|
| 511 |
+
"grad_norm": 0.30146345496177673,
|
| 512 |
+
"learning_rate": 8.62250594363378e-05,
|
| 513 |
+
"loss": 0.0126,
|
| 514 |
+
"step": 8400
|
| 515 |
+
},
|
| 516 |
+
{
|
| 517 |
+
"grad_norm": 0.2490146905183792,
|
| 518 |
+
"learning_rate": 8.584296863805709e-05,
|
| 519 |
+
"loss": 0.0127,
|
| 520 |
+
"step": 8500
|
| 521 |
+
},
|
| 522 |
+
{
|
| 523 |
+
"grad_norm": 0.28983908891677856,
|
| 524 |
+
"learning_rate": 8.545652262476897e-05,
|
| 525 |
+
"loss": 0.0129,
|
| 526 |
+
"step": 8600
|
| 527 |
+
},
|
| 528 |
+
{
|
| 529 |
+
"grad_norm": 0.24519403278827667,
|
| 530 |
+
"learning_rate": 8.50657683528373e-05,
|
| 531 |
+
"loss": 0.0125,
|
| 532 |
+
"step": 8700
|
| 533 |
+
},
|
| 534 |
+
{
|
| 535 |
+
"grad_norm": 0.3755444586277008,
|
| 536 |
+
"learning_rate": 8.467075330211474e-05,
|
| 537 |
+
"loss": 0.0123,
|
| 538 |
+
"step": 8800
|
| 539 |
+
},
|
| 540 |
+
{
|
| 541 |
+
"grad_norm": 0.2849225699901581,
|
| 542 |
+
"learning_rate": 8.42715254701735e-05,
|
| 543 |
+
"loss": 0.0119,
|
| 544 |
+
"step": 8900
|
| 545 |
+
},
|
| 546 |
+
{
|
| 547 |
+
"grad_norm": 0.30486589670181274,
|
| 548 |
+
"learning_rate": 8.386813336647339e-05,
|
| 549 |
+
"loss": 0.012,
|
| 550 |
+
"step": 9000
|
| 551 |
+
},
|
| 552 |
+
{
|
| 553 |
+
"grad_norm": 0.30513668060302734,
|
| 554 |
+
"learning_rate": 8.346062600646738e-05,
|
| 555 |
+
"loss": 0.0123,
|
| 556 |
+
"step": 9100
|
| 557 |
+
},
|
| 558 |
+
{
|
| 559 |
+
"grad_norm": 0.2400287687778473,
|
| 560 |
+
"learning_rate": 8.304905290564586e-05,
|
| 561 |
+
"loss": 0.0129,
|
| 562 |
+
"step": 9200
|
| 563 |
+
},
|
| 564 |
+
{
|
| 565 |
+
"grad_norm": 0.26750028133392334,
|
| 566 |
+
"learning_rate": 8.263346407352018e-05,
|
| 567 |
+
"loss": 0.0124,
|
| 568 |
+
"step": 9300
|
| 569 |
+
},
|
| 570 |
+
{
|
| 571 |
+
"grad_norm": 0.2039482742547989,
|
| 572 |
+
"learning_rate": 8.22139100075459e-05,
|
| 573 |
+
"loss": 0.0119,
|
| 574 |
+
"step": 9400
|
| 575 |
+
},
|
| 576 |
+
{
|
| 577 |
+
"grad_norm": 0.21343234181404114,
|
| 578 |
+
"learning_rate": 8.179044168698721e-05,
|
| 579 |
+
"loss": 0.0121,
|
| 580 |
+
"step": 9500
|
| 581 |
+
},
|
| 582 |
+
{
|
| 583 |
+
"grad_norm": 0.2371583878993988,
|
| 584 |
+
"learning_rate": 8.136311056672223e-05,
|
| 585 |
+
"loss": 0.0127,
|
| 586 |
+
"step": 9600
|
| 587 |
+
},
|
| 588 |
+
{
|
| 589 |
+
"grad_norm": 0.18970274925231934,
|
| 590 |
+
"learning_rate": 8.093196857099106e-05,
|
| 591 |
+
"loss": 0.0119,
|
| 592 |
+
"step": 9700
|
| 593 |
+
},
|
| 594 |
+
{
|
| 595 |
+
"grad_norm": 0.23267799615859985,
|
| 596 |
+
"learning_rate": 8.04970680870864e-05,
|
| 597 |
+
"loss": 0.0115,
|
| 598 |
+
"step": 9800
|
| 599 |
+
},
|
| 600 |
+
{
|
| 601 |
+
"grad_norm": 0.27450788021087646,
|
| 602 |
+
"learning_rate": 8.005846195898814e-05,
|
| 603 |
+
"loss": 0.0115,
|
| 604 |
+
"step": 9900
|
| 605 |
+
},
|
| 606 |
+
{
|
| 607 |
+
"grad_norm": 0.22092421352863312,
|
| 608 |
+
"learning_rate": 7.961620348094241e-05,
|
| 609 |
+
"loss": 0.0119,
|
| 610 |
+
"step": 10000
|
| 611 |
+
},
|
| 612 |
+
{
|
| 613 |
+
"grad_norm": 0.23360943794250488,
|
| 614 |
+
"learning_rate": 7.917034639098579e-05,
|
| 615 |
+
"loss": 0.0113,
|
| 616 |
+
"step": 10100
|
| 617 |
+
},
|
| 618 |
+
{
|
| 619 |
+
"grad_norm": 0.24877329170703888,
|
| 620 |
+
"learning_rate": 7.872094486441579e-05,
|
| 621 |
+
"loss": 0.0111,
|
| 622 |
+
"step": 10200
|
| 623 |
+
},
|
| 624 |
+
{
|
| 625 |
+
"grad_norm": 0.1870928406715393,
|
| 626 |
+
"learning_rate": 7.826805350720807e-05,
|
| 627 |
+
"loss": 0.0111,
|
| 628 |
+
"step": 10300
|
| 629 |
+
},
|
| 630 |
+
{
|
| 631 |
+
"grad_norm": 0.22153514623641968,
|
| 632 |
+
"learning_rate": 7.781172734938136e-05,
|
| 633 |
+
"loss": 0.0112,
|
| 634 |
+
"step": 10400
|
| 635 |
+
},
|
| 636 |
+
{
|
| 637 |
+
"grad_norm": 0.18555976450443268,
|
| 638 |
+
"learning_rate": 7.735202183831085e-05,
|
| 639 |
+
"loss": 0.0112,
|
| 640 |
+
"step": 10500
|
| 641 |
+
},
|
| 642 |
+
{
|
| 643 |
+
"grad_norm": 0.19689300656318665,
|
| 644 |
+
"learning_rate": 7.688899283199096e-05,
|
| 645 |
+
"loss": 0.0108,
|
| 646 |
+
"step": 10600
|
| 647 |
+
},
|
| 648 |
+
{
|
| 649 |
+
"grad_norm": 0.25766801834106445,
|
| 650 |
+
"learning_rate": 7.642269659224803e-05,
|
| 651 |
+
"loss": 0.0111,
|
| 652 |
+
"step": 10700
|
| 653 |
+
},
|
| 654 |
+
{
|
| 655 |
+
"grad_norm": 0.23566016554832458,
|
| 656 |
+
"learning_rate": 7.595318977790408e-05,
|
| 657 |
+
"loss": 0.0112,
|
| 658 |
+
"step": 10800
|
| 659 |
+
},
|
| 660 |
+
{
|
| 661 |
+
"grad_norm": 0.23820903897285461,
|
| 662 |
+
"learning_rate": 7.54805294378923e-05,
|
| 663 |
+
"loss": 0.011,
|
| 664 |
+
"step": 10900
|
| 665 |
+
},
|
| 666 |
+
{
|
| 667 |
+
"grad_norm": 0.2603837847709656,
|
| 668 |
+
"learning_rate": 7.50047730043252e-05,
|
| 669 |
+
"loss": 0.011,
|
| 670 |
+
"step": 11000
|
| 671 |
+
},
|
| 672 |
+
{
|
| 673 |
+
"grad_norm": 0.23611845076084137,
|
| 674 |
+
"learning_rate": 7.452597828551604e-05,
|
| 675 |
+
"loss": 0.0111,
|
| 676 |
+
"step": 11100
|
| 677 |
+
},
|
| 678 |
+
{
|
| 679 |
+
"grad_norm": 0.3005886375904083,
|
| 680 |
+
"learning_rate": 7.404420345895467e-05,
|
| 681 |
+
"loss": 0.011,
|
| 682 |
+
"step": 11200
|
| 683 |
+
},
|
| 684 |
+
{
|
| 685 |
+
"grad_norm": 0.20239433646202087,
|
| 686 |
+
"learning_rate": 7.355950706423853e-05,
|
| 687 |
+
"loss": 0.0111,
|
| 688 |
+
"step": 11300
|
| 689 |
+
},
|
| 690 |
+
{
|
| 691 |
+
"grad_norm": 0.21899692714214325,
|
| 692 |
+
"learning_rate": 7.307194799595958e-05,
|
| 693 |
+
"loss": 0.0115,
|
| 694 |
+
"step": 11400
|
| 695 |
+
},
|
| 696 |
+
{
|
| 697 |
+
"grad_norm": 0.17552578449249268,
|
| 698 |
+
"learning_rate": 7.25815854965481e-05,
|
| 699 |
+
"loss": 0.0102,
|
| 700 |
+
"step": 11500
|
| 701 |
+
},
|
| 702 |
+
{
|
| 703 |
+
"grad_norm": 0.22923225164413452,
|
| 704 |
+
"learning_rate": 7.20884791490743e-05,
|
| 705 |
+
"loss": 0.0107,
|
| 706 |
+
"step": 11600
|
| 707 |
+
},
|
| 708 |
+
{
|
| 709 |
+
"grad_norm": 0.30257314443588257,
|
| 710 |
+
"learning_rate": 7.159268887000846e-05,
|
| 711 |
+
"loss": 0.0104,
|
| 712 |
+
"step": 11700
|
| 713 |
+
},
|
| 714 |
+
{
|
| 715 |
+
"grad_norm": 0.27300789952278137,
|
| 716 |
+
"learning_rate": 7.109427490194056e-05,
|
| 717 |
+
"loss": 0.0113,
|
| 718 |
+
"step": 11800
|
| 719 |
+
},
|
| 720 |
+
{
|
| 721 |
+
"grad_norm": 0.14556348323822021,
|
| 722 |
+
"learning_rate": 7.059329780626034e-05,
|
| 723 |
+
"loss": 0.0101,
|
| 724 |
+
"step": 11900
|
| 725 |
+
},
|
| 726 |
+
{
|
| 727 |
+
"grad_norm": 0.22159963846206665,
|
| 728 |
+
"learning_rate": 7.008981845579864e-05,
|
| 729 |
+
"loss": 0.0102,
|
| 730 |
+
"step": 12000
|
| 731 |
+
},
|
| 732 |
+
{
|
| 733 |
+
"grad_norm": 0.21763812005519867,
|
| 734 |
+
"learning_rate": 6.958389802743077e-05,
|
| 735 |
+
"loss": 0.0104,
|
| 736 |
+
"step": 12100
|
| 737 |
+
},
|
| 738 |
+
{
|
| 739 |
+
"grad_norm": 0.25054824352264404,
|
| 740 |
+
"learning_rate": 6.9075597994643e-05,
|
| 741 |
+
"loss": 0.0105,
|
| 742 |
+
"step": 12200
|
| 743 |
+
},
|
| 744 |
+
{
|
| 745 |
+
"grad_norm": 0.16334953904151917,
|
| 746 |
+
"learning_rate": 6.856498012006318e-05,
|
| 747 |
+
"loss": 0.0106,
|
| 748 |
+
"step": 12300
|
| 749 |
+
},
|
| 750 |
+
{
|
| 751 |
+
"grad_norm": 0.16856291890144348,
|
| 752 |
+
"learning_rate": 6.805210644795588e-05,
|
| 753 |
+
"loss": 0.0099,
|
| 754 |
+
"step": 12400
|
| 755 |
+
},
|
| 756 |
+
{
|
| 757 |
+
"grad_norm": 0.17980563640594482,
|
| 758 |
+
"learning_rate": 6.753703929668362e-05,
|
| 759 |
+
"loss": 0.0106,
|
| 760 |
+
"step": 12500
|
| 761 |
+
},
|
| 762 |
+
{
|
| 763 |
+
"grad_norm": 0.1946818083524704,
|
| 764 |
+
"learning_rate": 6.70198412511347e-05,
|
| 765 |
+
"loss": 0.01,
|
| 766 |
+
"step": 12600
|
| 767 |
+
},
|
| 768 |
+
{
|
| 769 |
+
"grad_norm": 0.15612050890922546,
|
| 770 |
+
"learning_rate": 6.650057515511848e-05,
|
| 771 |
+
"loss": 0.0096,
|
| 772 |
+
"step": 12700
|
| 773 |
+
},
|
| 774 |
+
{
|
| 775 |
+
"grad_norm": 0.34721267223358154,
|
| 776 |
+
"learning_rate": 6.597930410372954e-05,
|
| 777 |
+
"loss": 0.01,
|
| 778 |
+
"step": 12800
|
| 779 |
+
},
|
| 780 |
+
{
|
| 781 |
+
"grad_norm": 0.22096088528633118,
|
| 782 |
+
"learning_rate": 6.54560914356809e-05,
|
| 783 |
+
"loss": 0.0099,
|
| 784 |
+
"step": 12900
|
| 785 |
+
},
|
| 786 |
+
{
|
| 787 |
+
"grad_norm": 0.18642286956310272,
|
| 788 |
+
"learning_rate": 6.493100072560799e-05,
|
| 789 |
+
"loss": 0.0098,
|
| 790 |
+
"step": 13000
|
| 791 |
+
},
|
| 792 |
+
{
|
| 793 |
+
"grad_norm": 0.2427489161491394,
|
| 794 |
+
"learning_rate": 6.44040957763438e-05,
|
| 795 |
+
"loss": 0.0098,
|
| 796 |
+
"step": 13100
|
| 797 |
+
},
|
| 798 |
+
{
|
| 799 |
+
"grad_norm": 0.2215547412633896,
|
| 800 |
+
"learning_rate": 6.387544061116621e-05,
|
| 801 |
+
"loss": 0.0101,
|
| 802 |
+
"step": 13200
|
| 803 |
+
},
|
| 804 |
+
{
|
| 805 |
+
"grad_norm": 0.1875009834766388,
|
| 806 |
+
"learning_rate": 6.334509946601879e-05,
|
| 807 |
+
"loss": 0.01,
|
| 808 |
+
"step": 13300
|
| 809 |
+
},
|
| 810 |
+
{
|
| 811 |
+
"grad_norm": 0.25244301557540894,
|
| 812 |
+
"learning_rate": 6.281313678170542e-05,
|
| 813 |
+
"loss": 0.0096,
|
| 814 |
+
"step": 13400
|
| 815 |
+
},
|
| 816 |
+
{
|
| 817 |
+
"grad_norm": 0.1568467915058136,
|
| 818 |
+
"learning_rate": 6.227961719606039e-05,
|
| 819 |
+
"loss": 0.0096,
|
| 820 |
+
"step": 13500
|
| 821 |
+
},
|
| 822 |
+
{
|
| 823 |
+
"grad_norm": 0.22288033366203308,
|
| 824 |
+
"learning_rate": 6.174460553609426e-05,
|
| 825 |
+
"loss": 0.0094,
|
| 826 |
+
"step": 13600
|
| 827 |
+
},
|
| 828 |
+
{
|
| 829 |
+
"grad_norm": 0.16578468680381775,
|
| 830 |
+
"learning_rate": 6.120816681011684e-05,
|
| 831 |
+
"loss": 0.0092,
|
| 832 |
+
"step": 13700
|
| 833 |
+
},
|
| 834 |
+
{
|
| 835 |
+
"grad_norm": 0.20067058503627777,
|
| 836 |
+
"learning_rate": 6.0670366199838215e-05,
|
| 837 |
+
"loss": 0.0095,
|
| 838 |
+
"step": 13800
|
| 839 |
+
},
|
| 840 |
+
{
|
| 841 |
+
"grad_norm": 0.16595327854156494,
|
| 842 |
+
"learning_rate": 6.013126905244858e-05,
|
| 843 |
+
"loss": 0.0099,
|
| 844 |
+
"step": 13900
|
| 845 |
+
},
|
| 846 |
+
{
|
| 847 |
+
"grad_norm": 0.18826332688331604,
|
| 848 |
+
"learning_rate": 5.959094087267804e-05,
|
| 849 |
+
"loss": 0.0096,
|
| 850 |
+
"step": 14000
|
| 851 |
+
},
|
| 852 |
+
{
|
| 853 |
+
"grad_norm": 0.17039699852466583,
|
| 854 |
+
"learning_rate": 5.904944731483723e-05,
|
| 855 |
+
"loss": 0.0098,
|
| 856 |
+
"step": 14100
|
| 857 |
+
},
|
| 858 |
+
{
|
| 859 |
+
"grad_norm": 0.17122985422611237,
|
| 860 |
+
"learning_rate": 5.850685417483983e-05,
|
| 861 |
+
"loss": 0.0093,
|
| 862 |
+
"step": 14200
|
| 863 |
+
},
|
| 864 |
+
{
|
| 865 |
+
"grad_norm": 0.21498365700244904,
|
| 866 |
+
"learning_rate": 5.796322738220774e-05,
|
| 867 |
+
"loss": 0.009,
|
| 868 |
+
"step": 14300
|
| 869 |
+
},
|
| 870 |
+
{
|
| 871 |
+
"grad_norm": 0.1524549275636673,
|
| 872 |
+
"learning_rate": 5.741863299206014e-05,
|
| 873 |
+
"loss": 0.0091,
|
| 874 |
+
"step": 14400
|
| 875 |
+
},
|
| 876 |
+
{
|
| 877 |
+
"grad_norm": 0.16607609391212463,
|
| 878 |
+
"learning_rate": 5.687313717708728e-05,
|
| 879 |
+
"loss": 0.0094,
|
| 880 |
+
"step": 14500
|
| 881 |
+
},
|
| 882 |
+
{
|
| 883 |
+
"grad_norm": 0.203726127743721,
|
| 884 |
+
"learning_rate": 5.632680621950992e-05,
|
| 885 |
+
"loss": 0.0092,
|
| 886 |
+
"step": 14600
|
| 887 |
+
},
|
| 888 |
+
{
|
| 889 |
+
"grad_norm": 0.15645408630371094,
|
| 890 |
+
"learning_rate": 5.57797065030255e-05,
|
| 891 |
+
"loss": 0.0088,
|
| 892 |
+
"step": 14700
|
| 893 |
+
},
|
| 894 |
+
{
|
| 895 |
+
"grad_norm": 0.1770261824131012,
|
| 896 |
+
"learning_rate": 5.523190450474197e-05,
|
| 897 |
+
"loss": 0.0088,
|
| 898 |
+
"step": 14800
|
| 899 |
+
},
|
| 900 |
+
{
|
| 901 |
+
"grad_norm": 0.2279656082391739,
|
| 902 |
+
"learning_rate": 5.4683466787100304e-05,
|
| 903 |
+
"loss": 0.009,
|
| 904 |
+
"step": 14900
|
| 905 |
+
},
|
| 906 |
+
{
|
| 907 |
+
"grad_norm": 0.1725756675004959,
|
| 908 |
+
"learning_rate": 5.413445998978658e-05,
|
| 909 |
+
"loss": 0.0089,
|
| 910 |
+
"step": 15000
|
| 911 |
+
},
|
| 912 |
+
{
|
| 913 |
+
"grad_norm": 0.17754536867141724,
|
| 914 |
+
"learning_rate": 5.358495082163476e-05,
|
| 915 |
+
"loss": 0.009,
|
| 916 |
+
"step": 15100
|
| 917 |
+
},
|
| 918 |
+
{
|
| 919 |
+
"grad_norm": 0.20454689860343933,
|
| 920 |
+
"learning_rate": 5.3035006052520955e-05,
|
| 921 |
+
"loss": 0.009,
|
| 922 |
+
"step": 15200
|
| 923 |
+
},
|
| 924 |
+
{
|
| 925 |
+
"grad_norm": 0.13121873140335083,
|
| 926 |
+
"learning_rate": 5.2484692505250376e-05,
|
| 927 |
+
"loss": 0.0085,
|
| 928 |
+
"step": 15300
|
| 929 |
+
},
|
| 930 |
+
{
|
| 931 |
+
"grad_norm": 0.20115773379802704,
|
| 932 |
+
"learning_rate": 5.193407704743782e-05,
|
| 933 |
+
"loss": 0.0081,
|
| 934 |
+
"step": 15400
|
| 935 |
+
},
|
| 936 |
+
{
|
| 937 |
+
"grad_norm": 0.20742738246917725,
|
| 938 |
+
"learning_rate": 5.1383226583382684e-05,
|
| 939 |
+
"loss": 0.0082,
|
| 940 |
+
"step": 15500
|
| 941 |
+
},
|
| 942 |
+
{
|
| 943 |
+
"grad_norm": 0.15646615624427795,
|
| 944 |
+
"learning_rate": 5.0832208045939556e-05,
|
| 945 |
+
"loss": 0.0088,
|
| 946 |
+
"step": 15600
|
| 947 |
+
},
|
| 948 |
+
{
|
| 949 |
+
"grad_norm": 0.2039646953344345,
|
| 950 |
+
"learning_rate": 5.028108838838533e-05,
|
| 951 |
+
"loss": 0.0082,
|
| 952 |
+
"step": 15700
|
| 953 |
+
},
|
| 954 |
+
{
|
| 955 |
+
"grad_norm": 0.1563130021095276,
|
| 956 |
+
"learning_rate": 4.972993457628381e-05,
|
| 957 |
+
"loss": 0.0085,
|
| 958 |
+
"step": 15800
|
| 959 |
+
},
|
| 960 |
+
{
|
| 961 |
+
"grad_norm": 0.1761900782585144,
|
| 962 |
+
"learning_rate": 4.9178813579348865e-05,
|
| 963 |
+
"loss": 0.0083,
|
| 964 |
+
"step": 15900
|
| 965 |
+
},
|
| 966 |
+
{
|
| 967 |
+
"grad_norm": 0.18039952218532562,
|
| 968 |
+
"learning_rate": 4.862779236330705e-05,
|
| 969 |
+
"loss": 0.0084,
|
| 970 |
+
"step": 16000
|
| 971 |
+
},
|
| 972 |
+
{
|
| 973 |
+
"grad_norm": 0.17899063229560852,
|
| 974 |
+
"learning_rate": 4.807693788176071e-05,
|
| 975 |
+
"loss": 0.0086,
|
| 976 |
+
"step": 16100
|
| 977 |
+
},
|
| 978 |
+
{
|
| 979 |
+
"grad_norm": 0.2654184401035309,
|
| 980 |
+
"learning_rate": 4.752631706805261e-05,
|
| 981 |
+
"loss": 0.0081,
|
| 982 |
+
"step": 16200
|
| 983 |
+
},
|
| 984 |
+
{
|
| 985 |
+
"grad_norm": 0.14795733988285065,
|
| 986 |
+
"learning_rate": 4.6975996827132926e-05,
|
| 987 |
+
"loss": 0.0084,
|
| 988 |
+
"step": 16300
|
| 989 |
+
},
|
| 990 |
+
{
|
| 991 |
+
"grad_norm": 0.12497590482234955,
|
| 992 |
+
"learning_rate": 4.6426044027429786e-05,
|
| 993 |
+
"loss": 0.0081,
|
| 994 |
+
"step": 16400
|
| 995 |
+
},
|
| 996 |
+
{
|
| 997 |
+
"grad_norm": 0.1409577578306198,
|
| 998 |
+
"learning_rate": 4.58765254927242e-05,
|
| 999 |
+
"loss": 0.0084,
|
| 1000 |
+
"step": 16500
|
| 1001 |
+
},
|
| 1002 |
+
{
|
| 1003 |
+
"grad_norm": 0.22204318642616272,
|
| 1004 |
+
"learning_rate": 4.5327507994030394e-05,
|
| 1005 |
+
"loss": 0.0079,
|
| 1006 |
+
"step": 16600
|
| 1007 |
+
},
|
| 1008 |
+
{
|
| 1009 |
+
"grad_norm": 0.13597023487091064,
|
| 1010 |
+
"learning_rate": 4.47790582414826e-05,
|
| 1011 |
+
"loss": 0.0079,
|
| 1012 |
+
"step": 16700
|
| 1013 |
+
},
|
| 1014 |
+
{
|
| 1015 |
+
"grad_norm": 0.20554523169994354,
|
| 1016 |
+
"learning_rate": 4.423124287622926e-05,
|
| 1017 |
+
"loss": 0.0082,
|
| 1018 |
+
"step": 16800
|
| 1019 |
+
},
|
| 1020 |
+
{
|
| 1021 |
+
"grad_norm": 0.16015072166919708,
|
| 1022 |
+
"learning_rate": 4.3684128462335533e-05,
|
| 1023 |
+
"loss": 0.0081,
|
| 1024 |
+
"step": 16900
|
| 1025 |
+
},
|
| 1026 |
+
{
|
| 1027 |
+
"grad_norm": 0.1688092052936554,
|
| 1028 |
+
"learning_rate": 4.3137781478695236e-05,
|
| 1029 |
+
"loss": 0.008,
|
| 1030 |
+
"step": 17000
|
| 1031 |
+
},
|
| 1032 |
+
{
|
| 1033 |
+
"grad_norm": 0.1572929322719574,
|
| 1034 |
+
"learning_rate": 4.25922683109531e-05,
|
| 1035 |
+
"loss": 0.0079,
|
| 1036 |
+
"step": 17100
|
| 1037 |
+
},
|
| 1038 |
+
{
|
| 1039 |
+
"grad_norm": 0.13499732315540314,
|
| 1040 |
+
"learning_rate": 4.204765524343841e-05,
|
| 1041 |
+
"loss": 0.0081,
|
| 1042 |
+
"step": 17200
|
| 1043 |
+
},
|
| 1044 |
+
{
|
| 1045 |
+
"grad_norm": 0.1706732213497162,
|
| 1046 |
+
"learning_rate": 4.1504008451110845e-05,
|
| 1047 |
+
"loss": 0.0076,
|
| 1048 |
+
"step": 17300
|
| 1049 |
+
},
|
| 1050 |
+
{
|
| 1051 |
+
"grad_norm": 0.18587198853492737,
|
| 1052 |
+
"learning_rate": 4.096139399151971e-05,
|
| 1053 |
+
"loss": 0.0075,
|
| 1054 |
+
"step": 17400
|
| 1055 |
+
},
|
| 1056 |
+
{
|
| 1057 |
+
"grad_norm": 0.18033955991268158,
|
| 1058 |
+
"learning_rate": 4.041987779677745e-05,
|
| 1059 |
+
"loss": 0.0076,
|
| 1060 |
+
"step": 17500
|
| 1061 |
+
},
|
| 1062 |
+
{
|
| 1063 |
+
"grad_norm": 0.17771348357200623,
|
| 1064 |
+
"learning_rate": 3.9879525665548276e-05,
|
| 1065 |
+
"loss": 0.0074,
|
| 1066 |
+
"step": 17600
|
| 1067 |
+
},
|
| 1068 |
+
{
|
| 1069 |
+
"grad_norm": 0.16985709965229034,
|
| 1070 |
+
"learning_rate": 3.9340403255053135e-05,
|
| 1071 |
+
"loss": 0.0077,
|
| 1072 |
+
"step": 17700
|
| 1073 |
+
},
|
| 1074 |
+
{
|
| 1075 |
+
"grad_norm": 0.14482390880584717,
|
| 1076 |
+
"learning_rate": 3.880257607309178e-05,
|
| 1077 |
+
"loss": 0.0076,
|
| 1078 |
+
"step": 17800
|
| 1079 |
+
},
|
| 1080 |
+
{
|
| 1081 |
+
"grad_norm": 0.14010176062583923,
|
| 1082 |
+
"learning_rate": 3.826610947008312e-05,
|
| 1083 |
+
"loss": 0.0076,
|
| 1084 |
+
"step": 17900
|
| 1085 |
+
},
|
| 1086 |
+
{
|
| 1087 |
+
"grad_norm": 0.1676109880208969,
|
| 1088 |
+
"learning_rate": 3.773106863112451e-05,
|
| 1089 |
+
"loss": 0.0076,
|
| 1090 |
+
"step": 18000
|
| 1091 |
+
},
|
| 1092 |
+
{
|
| 1093 |
+
"grad_norm": 0.20549866557121277,
|
| 1094 |
+
"learning_rate": 3.719751856807125e-05,
|
| 1095 |
+
"loss": 0.0071,
|
| 1096 |
+
"step": 18100
|
| 1097 |
+
},
|
| 1098 |
+
{
|
| 1099 |
+
"grad_norm": 0.13862955570220947,
|
| 1100 |
+
"learning_rate": 3.666552411163718e-05,
|
| 1101 |
+
"loss": 0.0074,
|
| 1102 |
+
"step": 18200
|
| 1103 |
+
},
|
| 1104 |
+
{
|
| 1105 |
+
"grad_norm": 0.1274552196264267,
|
| 1106 |
+
"learning_rate": 3.6135149903517115e-05,
|
| 1107 |
+
"loss": 0.0071,
|
| 1108 |
+
"step": 18300
|
| 1109 |
+
},
|
| 1110 |
+
{
|
| 1111 |
+
"grad_norm": 0.13789066672325134,
|
| 1112 |
+
"learning_rate": 3.56064603885324e-05,
|
| 1113 |
+
"loss": 0.0072,
|
| 1114 |
+
"step": 18400
|
| 1115 |
+
},
|
| 1116 |
+
{
|
| 1117 |
+
"grad_norm": 0.15795986354351044,
|
| 1118 |
+
"learning_rate": 3.5079519806800374e-05,
|
| 1119 |
+
"loss": 0.0072,
|
| 1120 |
+
"step": 18500
|
| 1121 |
+
},
|
| 1122 |
+
{
|
| 1123 |
+
"grad_norm": 0.17951546609401703,
|
| 1124 |
+
"learning_rate": 3.4554392185928564e-05,
|
| 1125 |
+
"loss": 0.0073,
|
| 1126 |
+
"step": 18600
|
| 1127 |
+
},
|
| 1128 |
+
{
|
| 1129 |
+
"grad_norm": 0.1273060142993927,
|
| 1130 |
+
"learning_rate": 3.40311413332349e-05,
|
| 1131 |
+
"loss": 0.0072,
|
| 1132 |
+
"step": 18700
|
| 1133 |
+
},
|
| 1134 |
+
{
|
| 1135 |
+
"grad_norm": 0.1507553607225418,
|
| 1136 |
+
"learning_rate": 3.350983082799456e-05,
|
| 1137 |
+
"loss": 0.0074,
|
| 1138 |
+
"step": 18800
|
| 1139 |
+
},
|
| 1140 |
+
{
|
| 1141 |
+
"grad_norm": 0.11984481662511826,
|
| 1142 |
+
"learning_rate": 3.299052401371456e-05,
|
| 1143 |
+
"loss": 0.0069,
|
| 1144 |
+
"step": 18900
|
| 1145 |
+
},
|
| 1146 |
+
{
|
| 1147 |
+
"grad_norm": 0.1150362491607666,
|
| 1148 |
+
"learning_rate": 3.247328399043706e-05,
|
| 1149 |
+
"loss": 0.0071,
|
| 1150 |
+
"step": 19000
|
| 1151 |
+
},
|
| 1152 |
+
{
|
| 1153 |
+
"grad_norm": 0.1713079959154129,
|
| 1154 |
+
"learning_rate": 3.195817360707207e-05,
|
| 1155 |
+
"loss": 0.0068,
|
| 1156 |
+
"step": 19100
|
| 1157 |
+
},
|
| 1158 |
+
{
|
| 1159 |
+
"grad_norm": 0.1618063896894455,
|
| 1160 |
+
"learning_rate": 3.144525545376095e-05,
|
| 1161 |
+
"loss": 0.0066,
|
| 1162 |
+
"step": 19200
|
| 1163 |
+
},
|
| 1164 |
+
{
|
| 1165 |
+
"grad_norm": 0.17947286367416382,
|
| 1166 |
+
"learning_rate": 3.093459185427102e-05,
|
| 1167 |
+
"loss": 0.0068,
|
| 1168 |
+
"step": 19300
|
| 1169 |
+
},
|
| 1170 |
+
{
|
| 1171 |
+
"grad_norm": 0.1826605647802353,
|
| 1172 |
+
"learning_rate": 3.0426244858422847e-05,
|
| 1173 |
+
"loss": 0.0069,
|
| 1174 |
+
"step": 19400
|
| 1175 |
+
},
|
| 1176 |
+
{
|
| 1177 |
+
"grad_norm": 0.18567310273647308,
|
| 1178 |
+
"learning_rate": 2.9920276234550636e-05,
|
| 1179 |
+
"loss": 0.007,
|
| 1180 |
+
"step": 19500
|
| 1181 |
+
},
|
| 1182 |
+
{
|
| 1183 |
+
"grad_norm": 0.21001702547073364,
|
| 1184 |
+
"learning_rate": 2.9416747461996853e-05,
|
| 1185 |
+
"loss": 0.0065,
|
| 1186 |
+
"step": 19600
|
| 1187 |
+
},
|
| 1188 |
+
{
|
| 1189 |
+
"grad_norm": 0.12665317952632904,
|
| 1190 |
+
"learning_rate": 2.8915719723641975e-05,
|
| 1191 |
+
"loss": 0.0069,
|
| 1192 |
+
"step": 19700
|
| 1193 |
+
},
|
| 1194 |
+
{
|
| 1195 |
+
"grad_norm": 0.14662450551986694,
|
| 1196 |
+
"learning_rate": 2.841725389847032e-05,
|
| 1197 |
+
"loss": 0.0065,
|
| 1198 |
+
"step": 19800
|
| 1199 |
+
},
|
| 1200 |
+
{
|
| 1201 |
+
"grad_norm": 0.20009121298789978,
|
| 1202 |
+
"learning_rate": 2.7921410554172723e-05,
|
| 1203 |
+
"loss": 0.0067,
|
| 1204 |
+
"step": 19900
|
| 1205 |
+
},
|
| 1206 |
+
{
|
| 1207 |
+
"grad_norm": 0.13614611327648163,
|
| 1208 |
+
"learning_rate": 2.7428249939787e-05,
|
| 1209 |
+
"loss": 0.0066,
|
| 1210 |
+
"step": 20000
|
| 1211 |
+
},
|
| 1212 |
+
{
|
| 1213 |
+
"grad_norm": 0.1399415284395218,
|
| 1214 |
+
"learning_rate": 2.6937831978377325e-05,
|
| 1215 |
+
"loss": 0.0069,
|
| 1216 |
+
"step": 20100
|
| 1217 |
+
},
|
| 1218 |
+
{
|
| 1219 |
+
"grad_norm": 0.18569815158843994,
|
| 1220 |
+
"learning_rate": 2.6450216259753e-05,
|
| 1221 |
+
"loss": 0.0067,
|
| 1222 |
+
"step": 20200
|
| 1223 |
+
},
|
| 1224 |
+
{
|
| 1225 |
+
"grad_norm": 0.15461581945419312,
|
| 1226 |
+
"learning_rate": 2.596546203322777e-05,
|
| 1227 |
+
"loss": 0.0063,
|
| 1228 |
+
"step": 20300
|
| 1229 |
+
},
|
| 1230 |
+
{
|
| 1231 |
+
"grad_norm": 0.1224842369556427,
|
| 1232 |
+
"learning_rate": 2.5483628200420644e-05,
|
| 1233 |
+
"loss": 0.0064,
|
| 1234 |
+
"step": 20400
|
| 1235 |
+
},
|
| 1236 |
+
{
|
| 1237 |
+
"grad_norm": 0.12034544348716736,
|
| 1238 |
+
"learning_rate": 2.500477330809881e-05,
|
| 1239 |
+
"loss": 0.0064,
|
| 1240 |
+
"step": 20500
|
| 1241 |
+
},
|
| 1242 |
+
{
|
| 1243 |
+
"grad_norm": 0.18553292751312256,
|
| 1244 |
+
"learning_rate": 2.452895554106368e-05,
|
| 1245 |
+
"loss": 0.0065,
|
| 1246 |
+
"step": 20600
|
| 1247 |
+
},
|
| 1248 |
+
{
|
| 1249 |
+
"grad_norm": 0.1320573389530182,
|
| 1250 |
+
"learning_rate": 2.405623271508108e-05,
|
| 1251 |
+
"loss": 0.0065,
|
| 1252 |
+
"step": 20700
|
| 1253 |
+
},
|
| 1254 |
+
{
|
| 1255 |
+
"grad_norm": 0.1200307309627533,
|
| 1256 |
+
"learning_rate": 2.358666226985599e-05,
|
| 1257 |
+
"loss": 0.0063,
|
| 1258 |
+
"step": 20800
|
| 1259 |
+
},
|
| 1260 |
+
{
|
| 1261 |
+
"grad_norm": 0.187001034617424,
|
| 1262 |
+
"learning_rate": 2.312030126205335e-05,
|
| 1263 |
+
"loss": 0.0063,
|
| 1264 |
+
"step": 20900
|
| 1265 |
+
},
|
| 1266 |
+
{
|
| 1267 |
+
"grad_norm": 0.11407079547643661,
|
| 1268 |
+
"learning_rate": 2.2657206358364997e-05,
|
| 1269 |
+
"loss": 0.0063,
|
| 1270 |
+
"step": 21000
|
| 1271 |
+
},
|
| 1272 |
+
{
|
| 1273 |
+
"grad_norm": 0.12954096496105194,
|
| 1274 |
+
"learning_rate": 2.219743382862437e-05,
|
| 1275 |
+
"loss": 0.0062,
|
| 1276 |
+
"step": 21100
|
| 1277 |
+
},
|
| 1278 |
+
{
|
| 1279 |
+
"grad_norm": 0.1469729244709015,
|
| 1280 |
+
"learning_rate": 2.1741039538969183e-05,
|
| 1281 |
+
"loss": 0.0063,
|
| 1282 |
+
"step": 21200
|
| 1283 |
+
},
|
| 1284 |
+
{
|
| 1285 |
+
"grad_norm": 0.14081987738609314,
|
| 1286 |
+
"learning_rate": 2.1288078945053193e-05,
|
| 1287 |
+
"loss": 0.0065,
|
| 1288 |
+
"step": 21300
|
| 1289 |
+
},
|
| 1290 |
+
{
|
| 1291 |
+
"grad_norm": 0.10564892739057541,
|
| 1292 |
+
"learning_rate": 2.083860708530798e-05,
|
| 1293 |
+
"loss": 0.0063,
|
| 1294 |
+
"step": 21400
|
| 1295 |
+
},
|
| 1296 |
+
{
|
| 1297 |
+
"grad_norm": 0.14359961450099945,
|
| 1298 |
+
"learning_rate": 2.0392678574255282e-05,
|
| 1299 |
+
"loss": 0.006,
|
| 1300 |
+
"step": 21500
|
| 1301 |
+
},
|
| 1302 |
+
{
|
| 1303 |
+
"grad_norm": 0.1474694311618805,
|
| 1304 |
+
"learning_rate": 1.99503475958708e-05,
|
| 1305 |
+
"loss": 0.0061,
|
| 1306 |
+
"step": 21600
|
| 1307 |
+
},
|
| 1308 |
+
{
|
| 1309 |
+
"grad_norm": 0.12522459030151367,
|
| 1310 |
+
"learning_rate": 1.9511667897000574e-05,
|
| 1311 |
+
"loss": 0.0062,
|
| 1312 |
+
"step": 21700
|
| 1313 |
+
},
|
| 1314 |
+
{
|
| 1315 |
+
"grad_norm": 0.12364368885755539,
|
| 1316 |
+
"learning_rate": 1.9076692780830114e-05,
|
| 1317 |
+
"loss": 0.0059,
|
| 1318 |
+
"step": 21800
|
| 1319 |
+
},
|
| 1320 |
+
{
|
| 1321 |
+
"grad_norm": 0.1437225043773651,
|
| 1322 |
+
"learning_rate": 1.8645475100407788e-05,
|
| 1323 |
+
"loss": 0.0058,
|
| 1324 |
+
"step": 21900
|
| 1325 |
+
},
|
| 1326 |
+
{
|
| 1327 |
+
"grad_norm": 0.09141670167446136,
|
| 1328 |
+
"learning_rate": 1.8218067252222597e-05,
|
| 1329 |
+
"loss": 0.0058,
|
| 1330 |
+
"step": 22000
|
| 1331 |
+
},
|
| 1332 |
+
{
|
| 1333 |
+
"grad_norm": 0.12311328947544098,
|
| 1334 |
+
"learning_rate": 1.779452116983769e-05,
|
| 1335 |
+
"loss": 0.0058,
|
| 1336 |
+
"step": 22100
|
| 1337 |
+
},
|
| 1338 |
+
{
|
| 1339 |
+
"grad_norm": 0.12757642567157745,
|
| 1340 |
+
"learning_rate": 1.737488831757997e-05,
|
| 1341 |
+
"loss": 0.0058,
|
| 1342 |
+
"step": 22200
|
| 1343 |
+
},
|
| 1344 |
+
{
|
| 1345 |
+
"grad_norm": 0.13844671845436096,
|
| 1346 |
+
"learning_rate": 1.695921968428668e-05,
|
| 1347 |
+
"loss": 0.0057,
|
| 1348 |
+
"step": 22300
|
| 1349 |
+
},
|
| 1350 |
+
{
|
| 1351 |
+
"grad_norm": 0.10866222530603409,
|
| 1352 |
+
"learning_rate": 1.6547565777109976e-05,
|
| 1353 |
+
"loss": 0.0056,
|
| 1354 |
+
"step": 22400
|
| 1355 |
+
},
|
| 1356 |
+
{
|
| 1357 |
+
"grad_norm": 0.10315632075071335,
|
| 1358 |
+
"learning_rate": 1.613997661537981e-05,
|
| 1359 |
+
"loss": 0.0058,
|
| 1360 |
+
"step": 22500
|
| 1361 |
+
},
|
| 1362 |
+
{
|
| 1363 |
+
"grad_norm": 0.14259645342826843,
|
| 1364 |
+
"learning_rate": 1.573650172452615e-05,
|
| 1365 |
+
"loss": 0.0059,
|
| 1366 |
+
"step": 22600
|
| 1367 |
+
},
|
| 1368 |
+
{
|
| 1369 |
+
"grad_norm": 0.0896521657705307,
|
| 1370 |
+
"learning_rate": 1.5337190130061318e-05,
|
| 1371 |
+
"loss": 0.0058,
|
| 1372 |
+
"step": 22700
|
| 1373 |
+
},
|
| 1374 |
+
{
|
| 1375 |
+
"grad_norm": 0.12061295658349991,
|
| 1376 |
+
"learning_rate": 1.4942090351622883e-05,
|
| 1377 |
+
"loss": 0.0057,
|
| 1378 |
+
"step": 22800
|
| 1379 |
+
},
|
| 1380 |
+
{
|
| 1381 |
+
"grad_norm": 0.12187090516090393,
|
| 1382 |
+
"learning_rate": 1.4551250397078253e-05,
|
| 1383 |
+
"loss": 0.0056,
|
| 1384 |
+
"step": 22900
|
| 1385 |
+
},
|
| 1386 |
+
{
|
| 1387 |
+
"grad_norm": 0.10126439481973648,
|
| 1388 |
+
"learning_rate": 1.4164717756691176e-05,
|
| 1389 |
+
"loss": 0.0054,
|
| 1390 |
+
"step": 23000
|
| 1391 |
+
},
|
| 1392 |
+
{
|
| 1393 |
+
"grad_norm": 0.08180845528841019,
|
| 1394 |
+
"learning_rate": 1.3782539397351418e-05,
|
| 1395 |
+
"loss": 0.0056,
|
| 1396 |
+
"step": 23100
|
| 1397 |
+
},
|
| 1398 |
+
{
|
| 1399 |
+
"grad_norm": 0.13818106055259705,
|
| 1400 |
+
"learning_rate": 1.340476175686784e-05,
|
| 1401 |
+
"loss": 0.0054,
|
| 1402 |
+
"step": 23200
|
| 1403 |
+
},
|
| 1404 |
+
{
|
| 1405 |
+
"grad_norm": 0.1303163766860962,
|
| 1406 |
+
"learning_rate": 1.3031430738325778e-05,
|
| 1407 |
+
"loss": 0.0057,
|
| 1408 |
+
"step": 23300
|
| 1409 |
+
},
|
| 1410 |
+
{
|
| 1411 |
+
"grad_norm": 0.11419674754142761,
|
| 1412 |
+
"learning_rate": 1.2662591704509547e-05,
|
| 1413 |
+
"loss": 0.0057,
|
| 1414 |
+
"step": 23400
|
| 1415 |
+
},
|
| 1416 |
+
{
|
| 1417 |
+
"grad_norm": 0.18105146288871765,
|
| 1418 |
+
"learning_rate": 1.2298289472390417e-05,
|
| 1419 |
+
"loss": 0.0055,
|
| 1420 |
+
"step": 23500
|
| 1421 |
+
},
|
| 1422 |
+
{
|
| 1423 |
+
"grad_norm": 0.08884292095899582,
|
| 1424 |
+
"learning_rate": 1.1938568307680964e-05,
|
| 1425 |
+
"loss": 0.0054,
|
| 1426 |
+
"step": 23600
|
| 1427 |
+
},
|
| 1428 |
+
{
|
| 1429 |
+
"grad_norm": 0.10794202238321304,
|
| 1430 |
+
"learning_rate": 1.1583471919456506e-05,
|
| 1431 |
+
"loss": 0.0054,
|
| 1432 |
+
"step": 23700
|
| 1433 |
+
},
|
| 1434 |
+
{
|
| 1435 |
+
"grad_norm": 0.1046849936246872,
|
| 1436 |
+
"learning_rate": 1.1233043454844016e-05,
|
| 1437 |
+
"loss": 0.0056,
|
| 1438 |
+
"step": 23800
|
| 1439 |
+
},
|
| 1440 |
+
{
|
| 1441 |
+
"grad_norm": 0.09406965225934982,
|
| 1442 |
+
"learning_rate": 1.0887325493779405e-05,
|
| 1443 |
+
"loss": 0.0052,
|
| 1444 |
+
"step": 23900
|
| 1445 |
+
},
|
| 1446 |
+
{
|
| 1447 |
+
"grad_norm": 0.10608553141355515,
|
| 1448 |
+
"learning_rate": 1.0546360043833747e-05,
|
| 1449 |
+
"loss": 0.0055,
|
| 1450 |
+
"step": 24000
|
| 1451 |
+
},
|
| 1452 |
+
{
|
| 1453 |
+
"grad_norm": 0.09856823086738586,
|
| 1454 |
+
"learning_rate": 1.0210188535108995e-05,
|
| 1455 |
+
"loss": 0.0054,
|
| 1456 |
+
"step": 24100
|
| 1457 |
+
},
|
| 1458 |
+
{
|
| 1459 |
+
"grad_norm": 0.0984424352645874,
|
| 1460 |
+
"learning_rate": 9.878851815203882e-06,
|
| 1461 |
+
"loss": 0.0054,
|
| 1462 |
+
"step": 24200
|
| 1463 |
+
},
|
| 1464 |
+
{
|
| 1465 |
+
"grad_norm": 0.1498441994190216,
|
| 1466 |
+
"learning_rate": 9.55239014425055e-06,
|
| 1467 |
+
"loss": 0.0054,
|
| 1468 |
+
"step": 24300
|
| 1469 |
+
},
|
| 1470 |
+
{
|
| 1471 |
+
"grad_norm": 0.14669004082679749,
|
| 1472 |
+
"learning_rate": 9.230843190022725e-06,
|
| 1473 |
+
"loss": 0.0053,
|
| 1474 |
+
"step": 24400
|
| 1475 |
+
},
|
| 1476 |
+
{
|
| 1477 |
+
"grad_norm": 0.09039339423179626,
|
| 1478 |
+
"learning_rate": 8.914250023115672e-06,
|
| 1479 |
+
"loss": 0.005,
|
| 1480 |
+
"step": 24500
|
| 1481 |
+
},
|
| 1482 |
+
{
|
| 1483 |
+
"grad_norm": 0.09510180354118347,
|
| 1484 |
+
"learning_rate": 8.602649112198796e-06,
|
| 1485 |
+
"loss": 0.0052,
|
| 1486 |
+
"step": 24600
|
| 1487 |
+
},
|
| 1488 |
+
{
|
| 1489 |
+
"grad_norm": 0.09959646314382553,
|
| 1490 |
+
"learning_rate": 8.296078319341443e-06,
|
| 1491 |
+
"loss": 0.0052,
|
| 1492 |
+
"step": 24700
|
| 1493 |
+
},
|
| 1494 |
+
{
|
| 1495 |
+
"grad_norm": 0.10918956249952316,
|
| 1496 |
+
"learning_rate": 7.994574895412293e-06,
|
| 1497 |
+
"loss": 0.0051,
|
| 1498 |
+
"step": 24800
|
| 1499 |
+
},
|
| 1500 |
+
{
|
| 1501 |
+
"grad_norm": 0.09922792762517929,
|
| 1502 |
+
"learning_rate": 7.698175475553077e-06,
|
| 1503 |
+
"loss": 0.005,
|
| 1504 |
+
"step": 24900
|
| 1505 |
+
},
|
| 1506 |
+
{
|
| 1507 |
+
"grad_norm": 0.08747448772192001,
|
| 1508 |
+
"learning_rate": 7.40691607472715e-06,
|
| 1509 |
+
"loss": 0.0052,
|
| 1510 |
+
"step": 25000
|
| 1511 |
+
},
|
| 1512 |
+
{
|
| 1513 |
+
"grad_norm": 0.11222247034311295,
|
| 1514 |
+
"learning_rate": 7.120832083343337e-06,
|
| 1515 |
+
"loss": 0.0052,
|
| 1516 |
+
"step": 25100
|
| 1517 |
+
},
|
| 1518 |
+
{
|
| 1519 |
+
"grad_norm": 0.08529548346996307,
|
| 1520 |
+
"learning_rate": 6.839958262955709e-06,
|
| 1521 |
+
"loss": 0.005,
|
| 1522 |
+
"step": 25200
|
| 1523 |
+
},
|
| 1524 |
+
{
|
| 1525 |
+
"grad_norm": 0.14458157122135162,
|
| 1526 |
+
"learning_rate": 6.564328742039783e-06,
|
| 1527 |
+
"loss": 0.0051,
|
| 1528 |
+
"step": 25300
|
| 1529 |
+
},
|
| 1530 |
+
{
|
| 1531 |
+
"grad_norm": 0.08075789362192154,
|
| 1532 |
+
"learning_rate": 6.293977011845648e-06,
|
| 1533 |
+
"loss": 0.0052,
|
| 1534 |
+
"step": 25400
|
| 1535 |
+
},
|
| 1536 |
+
{
|
| 1537 |
+
"grad_norm": 0.06792178004980087,
|
| 1538 |
+
"learning_rate": 6.0289359223284434e-06,
|
| 1539 |
+
"loss": 0.0051,
|
| 1540 |
+
"step": 25500
|
| 1541 |
+
},
|
| 1542 |
+
{
|
| 1543 |
+
"grad_norm": 0.09026592969894409,
|
| 1544 |
+
"learning_rate": 5.769237678156897e-06,
|
| 1545 |
+
"loss": 0.0049,
|
| 1546 |
+
"step": 25600
|
| 1547 |
+
},
|
| 1548 |
+
{
|
| 1549 |
+
"grad_norm": 0.08753223717212677,
|
| 1550 |
+
"learning_rate": 5.514913834800134e-06,
|
| 1551 |
+
"loss": 0.0049,
|
| 1552 |
+
"step": 25700
|
| 1553 |
+
},
|
| 1554 |
+
{
|
| 1555 |
+
"grad_norm": 0.06475996971130371,
|
| 1556 |
+
"learning_rate": 5.265995294693471e-06,
|
| 1557 |
+
"loss": 0.0048,
|
| 1558 |
+
"step": 25800
|
| 1559 |
+
},
|
| 1560 |
+
{
|
| 1561 |
+
"grad_norm": 0.07268524169921875,
|
| 1562 |
+
"learning_rate": 5.022512303483451e-06,
|
| 1563 |
+
"loss": 0.005,
|
| 1564 |
+
"step": 25900
|
| 1565 |
+
},
|
| 1566 |
+
{
|
| 1567 |
+
"grad_norm": 0.1075899675488472,
|
| 1568 |
+
"learning_rate": 4.784494446352833e-06,
|
| 1569 |
+
"loss": 0.0048,
|
| 1570 |
+
"step": 26000
|
| 1571 |
+
},
|
| 1572 |
+
{
|
| 1573 |
+
"grad_norm": 0.08950436115264893,
|
| 1574 |
+
"learning_rate": 4.551970644425707e-06,
|
| 1575 |
+
"loss": 0.0046,
|
| 1576 |
+
"step": 26100
|
| 1577 |
+
},
|
| 1578 |
+
{
|
| 1579 |
+
"grad_norm": 0.10498484969139099,
|
| 1580 |
+
"learning_rate": 4.324969151253317e-06,
|
| 1581 |
+
"loss": 0.0047,
|
| 1582 |
+
"step": 26200
|
| 1583 |
+
},
|
| 1584 |
+
{
|
| 1585 |
+
"grad_norm": 0.09521999210119247,
|
| 1586 |
+
"learning_rate": 4.10351754938107e-06,
|
| 1587 |
+
"loss": 0.0048,
|
| 1588 |
+
"step": 26300
|
| 1589 |
+
},
|
| 1590 |
+
{
|
| 1591 |
+
"grad_norm": 0.10910005867481232,
|
| 1592 |
+
"learning_rate": 3.887642746997017e-06,
|
| 1593 |
+
"loss": 0.0049,
|
| 1594 |
+
"step": 26400
|
| 1595 |
+
},
|
| 1596 |
+
{
|
| 1597 |
+
"grad_norm": 0.06993323564529419,
|
| 1598 |
+
"learning_rate": 3.6773709746622363e-06,
|
| 1599 |
+
"loss": 0.0048,
|
| 1600 |
+
"step": 26500
|
| 1601 |
+
},
|
| 1602 |
+
{
|
| 1603 |
+
"grad_norm": 0.08529256284236908,
|
| 1604 |
+
"learning_rate": 3.472727782123697e-06,
|
| 1605 |
+
"loss": 0.0048,
|
| 1606 |
+
"step": 26600
|
| 1607 |
+
},
|
| 1608 |
+
{
|
| 1609 |
+
"grad_norm": 0.06200827285647392,
|
| 1610 |
+
"learning_rate": 3.273738035209678e-06,
|
| 1611 |
+
"loss": 0.0048,
|
| 1612 |
+
"step": 26700
|
| 1613 |
+
},
|
| 1614 |
+
{
|
| 1615 |
+
"grad_norm": 0.06921591609716415,
|
| 1616 |
+
"learning_rate": 3.0804259128083936e-06,
|
| 1617 |
+
"loss": 0.0045,
|
| 1618 |
+
"step": 26800
|
| 1619 |
+
},
|
| 1620 |
+
{
|
| 1621 |
+
"grad_norm": 0.07834340631961823,
|
| 1622 |
+
"learning_rate": 2.8928149039300524e-06,
|
| 1623 |
+
"loss": 0.0049,
|
| 1624 |
+
"step": 26900
|
| 1625 |
+
},
|
| 1626 |
+
{
|
| 1627 |
+
"grad_norm": 0.07856951653957367,
|
| 1628 |
+
"learning_rate": 2.7109278048527752e-06,
|
| 1629 |
+
"loss": 0.0047,
|
| 1630 |
+
"step": 27000
|
| 1631 |
+
},
|
| 1632 |
+
{
|
| 1633 |
+
"grad_norm": 0.04707355797290802,
|
| 1634 |
+
"learning_rate": 2.5347867163526384e-06,
|
| 1635 |
+
"loss": 0.0048,
|
| 1636 |
+
"step": 27100
|
| 1637 |
+
},
|
| 1638 |
+
{
|
| 1639 |
+
"grad_norm": 0.08059140294790268,
|
| 1640 |
+
"learning_rate": 2.364413041018232e-06,
|
| 1641 |
+
"loss": 0.0047,
|
| 1642 |
+
"step": 27200
|
| 1643 |
+
},
|
| 1644 |
+
{
|
| 1645 |
+
"grad_norm": 0.07572878897190094,
|
| 1646 |
+
"learning_rate": 2.1998274806501074e-06,
|
| 1647 |
+
"loss": 0.0047,
|
| 1648 |
+
"step": 27300
|
| 1649 |
+
},
|
| 1650 |
+
{
|
| 1651 |
+
"grad_norm": 0.05332242697477341,
|
| 1652 |
+
"learning_rate": 2.0410500337453176e-06,
|
| 1653 |
+
"loss": 0.0048,
|
| 1654 |
+
"step": 27400
|
| 1655 |
+
},
|
| 1656 |
+
{
|
| 1657 |
+
"grad_norm": 0.0765126571059227,
|
| 1658 |
+
"learning_rate": 1.8880999930674215e-06,
|
| 1659 |
+
"loss": 0.0046,
|
| 1660 |
+
"step": 27500
|
| 1661 |
+
},
|
| 1662 |
+
{
|
| 1663 |
+
"grad_norm": 0.06838490813970566,
|
| 1664 |
+
"learning_rate": 1.7409959433022871e-06,
|
| 1665 |
+
"loss": 0.0047,
|
| 1666 |
+
"step": 27600
|
| 1667 |
+
},
|
| 1668 |
+
{
|
| 1669 |
+
"grad_norm": 0.08280456066131592,
|
| 1670 |
+
"learning_rate": 1.5997557587998868e-06,
|
| 1671 |
+
"loss": 0.0047,
|
| 1672 |
+
"step": 27700
|
| 1673 |
+
},
|
| 1674 |
+
{
|
| 1675 |
+
"grad_norm": 0.08387008309364319,
|
| 1676 |
+
"learning_rate": 1.4643966014023957e-06,
|
| 1677 |
+
"loss": 0.0046,
|
| 1678 |
+
"step": 27800
|
| 1679 |
+
},
|
| 1680 |
+
{
|
| 1681 |
+
"grad_norm": 0.047321420162916183,
|
| 1682 |
+
"learning_rate": 1.3349349183589155e-06,
|
| 1683 |
+
"loss": 0.0049,
|
| 1684 |
+
"step": 27900
|
| 1685 |
+
},
|
| 1686 |
+
{
|
| 1687 |
+
"grad_norm": 0.05912323296070099,
|
| 1688 |
+
"learning_rate": 1.2113864403269836e-06,
|
| 1689 |
+
"loss": 0.0045,
|
| 1690 |
+
"step": 28000
|
| 1691 |
+
},
|
| 1692 |
+
{
|
| 1693 |
+
"grad_norm": 0.04512840136885643,
|
| 1694 |
+
"learning_rate": 1.0937661794611864e-06,
|
| 1695 |
+
"loss": 0.0046,
|
| 1696 |
+
"step": 28100
|
| 1697 |
+
},
|
| 1698 |
+
{
|
| 1699 |
+
"grad_norm": 0.0754234790802002,
|
| 1700 |
+
"learning_rate": 9.820884275890286e-07,
|
| 1701 |
+
"loss": 0.005,
|
| 1702 |
+
"step": 28200
|
| 1703 |
+
},
|
| 1704 |
+
{
|
| 1705 |
+
"grad_norm": 0.03836540877819061,
|
| 1706 |
+
"learning_rate": 8.763667544743836e-07,
|
| 1707 |
+
"loss": 0.0046,
|
| 1708 |
+
"step": 28300
|
| 1709 |
+
},
|
| 1710 |
+
{
|
| 1711 |
+
"grad_norm": 0.05754861980676651,
|
| 1712 |
+
"learning_rate": 7.766140061686522e-07,
|
| 1713 |
+
"loss": 0.0047,
|
| 1714 |
+
"step": 28400
|
| 1715 |
+
},
|
| 1716 |
+
{
|
| 1717 |
+
"grad_norm": 0.04895985126495361,
|
| 1718 |
+
"learning_rate": 6.828423034498488e-07,
|
| 1719 |
+
"loss": 0.0045,
|
| 1720 |
+
"step": 28500
|
| 1721 |
+
},
|
| 1722 |
+
{
|
| 1723 |
+
"grad_norm": 0.08934417366981506,
|
| 1724 |
+
"learning_rate": 5.950630403498469e-07,
|
| 1725 |
+
"loss": 0.0044,
|
| 1726 |
+
"step": 28600
|
| 1727 |
+
},
|
| 1728 |
+
{
|
| 1729 |
+
"grad_norm": 0.05324174836277962,
|
| 1730 |
+
"learning_rate": 5.132868827698978e-07,
|
| 1731 |
+
"loss": 0.0046,
|
| 1732 |
+
"step": 28700
|
| 1733 |
+
},
|
| 1734 |
+
{
|
| 1735 |
+
"grad_norm": 0.06453988701105118,
|
| 1736 |
+
"learning_rate": 4.375237671846333e-07,
|
| 1737 |
+
"loss": 0.0044,
|
| 1738 |
+
"step": 28800
|
| 1739 |
+
},
|
| 1740 |
+
{
|
| 1741 |
+
"grad_norm": 0.04550166428089142,
|
| 1742 |
+
"learning_rate": 3.677828994347154e-07,
|
| 1743 |
+
"loss": 0.0044,
|
| 1744 |
+
"step": 28900
|
| 1745 |
+
},
|
| 1746 |
+
{
|
| 1747 |
+
"grad_norm": 0.05266235023736954,
|
| 1748 |
+
"learning_rate": 3.040727536082366e-07,
|
| 1749 |
+
"loss": 0.0043,
|
| 1750 |
+
"step": 29000
|
| 1751 |
+
},
|
| 1752 |
+
{
|
| 1753 |
+
"grad_norm": 0.061544593423604965,
|
| 1754 |
+
"learning_rate": 2.464010710110598e-07,
|
| 1755 |
+
"loss": 0.0047,
|
| 1756 |
+
"step": 29100
|
| 1757 |
+
},
|
| 1758 |
+
{
|
| 1759 |
+
"grad_norm": 0.050445180386304855,
|
| 1760 |
+
"learning_rate": 1.9477485922618222e-07,
|
| 1761 |
+
"loss": 0.0045,
|
| 1762 |
+
"step": 29200
|
| 1763 |
+
},
|
| 1764 |
+
{
|
| 1765 |
+
"grad_norm": 0.05489252880215645,
|
| 1766 |
+
"learning_rate": 1.4920039126225527e-07,
|
| 1767 |
+
"loss": 0.0046,
|
| 1768 |
+
"step": 29300
|
| 1769 |
+
},
|
| 1770 |
+
{
|
| 1771 |
+
"grad_norm": 0.04267454892396927,
|
| 1772 |
+
"learning_rate": 1.0968320479136095e-07,
|
| 1773 |
+
"loss": 0.0044,
|
| 1774 |
+
"step": 29400
|
| 1775 |
+
},
|
| 1776 |
+
{
|
| 1777 |
+
"grad_norm": 0.03720274195075035,
|
| 1778 |
+
"learning_rate": 7.622810147614456e-08,
|
| 1779 |
+
"loss": 0.0045,
|
| 1780 |
+
"step": 29500
|
| 1781 |
+
},
|
| 1782 |
+
{
|
| 1783 |
+
"grad_norm": 0.05552380904555321,
|
| 1784 |
+
"learning_rate": 4.883914638636467e-08,
|
| 1785 |
+
"loss": 0.0046,
|
| 1786 |
+
"step": 29600
|
| 1787 |
+
},
|
| 1788 |
+
{
|
| 1789 |
+
"grad_norm": 0.05537592992186546,
|
| 1790 |
+
"learning_rate": 2.7519667504971593e-08,
|
| 1791 |
+
"loss": 0.0046,
|
| 1792 |
+
"step": 29700
|
| 1793 |
+
},
|
| 1794 |
+
{
|
| 1795 |
+
"grad_norm": 0.05703074857592583,
|
| 1796 |
+
"learning_rate": 1.2272255323708681e-08,
|
| 1797 |
+
"loss": 0.0045,
|
| 1798 |
+
"step": 29800
|
| 1799 |
+
},
|
| 1800 |
+
{
|
| 1801 |
+
"grad_norm": 0.06217626854777336,
|
| 1802 |
+
"learning_rate": 3.0987625283473987e-09,
|
| 1803 |
+
"loss": 0.0044,
|
| 1804 |
+
"step": 29900
|
| 1805 |
+
},
|
| 1806 |
+
{
|
| 1807 |
+
"grad_norm": 0.047798920422792435,
|
| 1808 |
+
"learning_rate": 3.0377360626943076e-13,
|
| 1809 |
+
"loss": 0.0045,
|
| 1810 |
+
"step": 30000
|
| 1811 |
+
}
|
| 1812 |
+
],
|
| 1813 |
+
"logging_steps": 100,
|
| 1814 |
+
"max_steps": 30000,
|
| 1815 |
+
"num_input_tokens_seen": 0,
|
| 1816 |
+
"num_train_epochs": 9223372036854775807,
|
| 1817 |
+
"save_steps": 10000,
|
| 1818 |
+
"stateful_callbacks": {
|
| 1819 |
+
"TrainerControl": {
|
| 1820 |
+
"args": {
|
| 1821 |
+
"should_epoch_stop": false,
|
| 1822 |
+
"should_evaluate": false,
|
| 1823 |
+
"should_log": false,
|
| 1824 |
+
"should_save": true,
|
| 1825 |
+
"should_training_stop": true
|
| 1826 |
+
},
|
| 1827 |
+
"attributes": {}
|
| 1828 |
+
}
|
| 1829 |
+
},
|
| 1830 |
+
"total_flos": 0.0,
|
| 1831 |
+
"train_batch_size": 128,
|
| 1832 |
+
"trial_name": null,
|
| 1833 |
+
"trial_params": null
|
| 1834 |
+
}
|
autostep_v2_bs128_30k/gr00t_n17/checkpoints/checkpoint-30K/training_args.bin
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:124274f0bf033204ac3ba81ccf3ee9dcbadadbb56a17aa12548b2e6bcdbb37b8
|
| 3 |
+
size 5905
|
autostep_v2_bs128_30k/gr00t_n17/checkpoints/checkpoint-30K/wandb_config.json
ADDED
|
@@ -0,0 +1 @@
|
|
|
|
|
|
|
| 1 |
+
{"project": "kuavo", "run_id": "gr00t_n17"}
|