Upload folder using huggingface_hub
Browse files- config.json +64 -0
- experiment_cfg/metadata.json +307 -0
- model-00001-of-00002.safetensors +3 -0
- model-00002-of-00002.safetensors +3 -0
- model.safetensors.index.json +0 -0
- trainer_state.json +0 -0
config.json
ADDED
|
@@ -0,0 +1,64 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"action_dim": 32,
|
| 3 |
+
"action_head_cfg": {
|
| 4 |
+
"action_dim": 32,
|
| 5 |
+
"action_horizon": 16,
|
| 6 |
+
"add_pos_embed": true,
|
| 7 |
+
"backbone_embedding_dim": 2048,
|
| 8 |
+
"diffusion_model_cfg": {
|
| 9 |
+
"attention_head_dim": 48,
|
| 10 |
+
"cross_attention_dim": 2048,
|
| 11 |
+
"dropout": 0.2,
|
| 12 |
+
"final_dropout": true,
|
| 13 |
+
"interleave_self_attention": true,
|
| 14 |
+
"norm_type": "ada_norm",
|
| 15 |
+
"num_attention_heads": 32,
|
| 16 |
+
"num_layers": 16,
|
| 17 |
+
"output_dim": 1024,
|
| 18 |
+
"positional_embeddings": null
|
| 19 |
+
},
|
| 20 |
+
"hidden_size": 1024,
|
| 21 |
+
"input_embedding_dim": 1536,
|
| 22 |
+
"max_action_dim": 32,
|
| 23 |
+
"max_state_dim": 64,
|
| 24 |
+
"model_dtype": "float32",
|
| 25 |
+
"noise_beta_alpha": 1.5,
|
| 26 |
+
"noise_beta_beta": 1.0,
|
| 27 |
+
"noise_s": 0.999,
|
| 28 |
+
"num_inference_timesteps": 4,
|
| 29 |
+
"num_target_vision_tokens": 32,
|
| 30 |
+
"num_timestep_buckets": 1000,
|
| 31 |
+
"tune_diffusion_model": true,
|
| 32 |
+
"tune_projector": true,
|
| 33 |
+
"use_vlln": true,
|
| 34 |
+
"vl_self_attention_cfg": {
|
| 35 |
+
"attention_head_dim": 64,
|
| 36 |
+
"dropout": 0.2,
|
| 37 |
+
"final_dropout": true,
|
| 38 |
+
"num_attention_heads": 32,
|
| 39 |
+
"num_layers": 4,
|
| 40 |
+
"positional_embeddings": null
|
| 41 |
+
}
|
| 42 |
+
},
|
| 43 |
+
"action_horizon": 16,
|
| 44 |
+
"architectures": [
|
| 45 |
+
"GR00T_N1_5"
|
| 46 |
+
],
|
| 47 |
+
"attn_implementation": null,
|
| 48 |
+
"backbone_cfg": {
|
| 49 |
+
"eagle_path": "NVEagle/eagle_er-qwen3_1_7B-Siglip2_400M_stage1_5_128gpu_er_v7_1mlp_nops",
|
| 50 |
+
"load_bf16": false,
|
| 51 |
+
"project_to_dim": null,
|
| 52 |
+
"reproject_vision": false,
|
| 53 |
+
"select_layer": 12,
|
| 54 |
+
"tune_llm": false,
|
| 55 |
+
"tune_visual": true,
|
| 56 |
+
"use_flash_attention": true
|
| 57 |
+
},
|
| 58 |
+
"compute_dtype": "bfloat16",
|
| 59 |
+
"hidden_size": 2048,
|
| 60 |
+
"model_dtype": "float32",
|
| 61 |
+
"model_type": "gr00t_n1_5",
|
| 62 |
+
"torch_dtype": "bfloat16",
|
| 63 |
+
"transformers_version": "4.51.3"
|
| 64 |
+
}
|
experiment_cfg/metadata.json
ADDED
|
@@ -0,0 +1,307 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"new_embodiment": {
|
| 3 |
+
"statistics": {
|
| 4 |
+
"state": {
|
| 5 |
+
"left_arm": {
|
| 6 |
+
"max": [
|
| 7 |
+
0.5445871353149414,
|
| 8 |
+
0.8112002015113831,
|
| 9 |
+
0.6374648809432983,
|
| 10 |
+
-0.8231125473976135,
|
| 11 |
+
1.1653939485549927,
|
| 12 |
+
1.045743465423584,
|
| 13 |
+
0.2361546903848648,
|
| 14 |
+
1.2059297561645508
|
| 15 |
+
],
|
| 16 |
+
"min": [
|
| 17 |
+
-1.0513520240783691,
|
| 18 |
+
-0.000107858024421148,
|
| 19 |
+
-1.257924199104309,
|
| 20 |
+
-2.296549081802368,
|
| 21 |
+
-0.07028748095035553,
|
| 22 |
+
-0.5859926342964172,
|
| 23 |
+
-1.569265604019165,
|
| 24 |
+
0.0
|
| 25 |
+
],
|
| 26 |
+
"mean": [
|
| 27 |
+
0.04702014476060867,
|
| 28 |
+
0.2067810744047165,
|
| 29 |
+
-0.15320426225662231,
|
| 30 |
+
-1.547008991241455,
|
| 31 |
+
0.26597100496292114,
|
| 32 |
+
-0.07297467440366745,
|
| 33 |
+
-0.36368048191070557,
|
| 34 |
+
0.2825794816017151
|
| 35 |
+
],
|
| 36 |
+
"std": [
|
| 37 |
+
0.35919681191444397,
|
| 38 |
+
0.15744692087173462,
|
| 39 |
+
0.22181986272335052,
|
| 40 |
+
0.21383564174175262,
|
| 41 |
+
0.20425817370414734,
|
| 42 |
+
0.3615856468677521,
|
| 43 |
+
0.49840205907821655,
|
| 44 |
+
0.32313093543052673
|
| 45 |
+
],
|
| 46 |
+
"q01": [
|
| 47 |
+
-0.8543237709999084,
|
| 48 |
+
0.0,
|
| 49 |
+
-0.8280413961410522,
|
| 50 |
+
-2.0186822986602784,
|
| 51 |
+
0.003032008884474635,
|
| 52 |
+
-0.5468531346321106,
|
| 53 |
+
-1.3865834426879884,
|
| 54 |
+
0.038613513112068176
|
| 55 |
+
],
|
| 56 |
+
"q99": [
|
| 57 |
+
0.42267353534698493,
|
| 58 |
+
0.7178459906578064,
|
| 59 |
+
0.3515457284450534,
|
| 60 |
+
-1.0023030662536618,
|
| 61 |
+
0.8920035696029665,
|
| 62 |
+
0.7357211589813234,
|
| 63 |
+
0.18714067339897156,
|
| 64 |
+
1.0440499782562256
|
| 65 |
+
]
|
| 66 |
+
},
|
| 67 |
+
"right_arm": {
|
| 68 |
+
"max": [
|
| 69 |
+
0.6417672038078308,
|
| 70 |
+
5.9921123465755954e-05,
|
| 71 |
+
1.2177530527114868,
|
| 72 |
+
-1.0384330749511719,
|
| 73 |
+
0.7533283829689026,
|
| 74 |
+
0.9873323440551758,
|
| 75 |
+
1.363911509513855,
|
| 76 |
+
1.1524648666381836
|
| 77 |
+
],
|
| 78 |
+
"min": [
|
| 79 |
+
-0.9047250747680664,
|
| 80 |
+
-0.8373377919197083,
|
| 81 |
+
-0.6043045520782471,
|
| 82 |
+
-2.3454806804656982,
|
| 83 |
+
-0.7402056455612183,
|
| 84 |
+
-0.7786630392074585,
|
| 85 |
+
-0.36711385846138,
|
| 86 |
+
0.0
|
| 87 |
+
],
|
| 88 |
+
"mean": [
|
| 89 |
+
0.1410096287727356,
|
| 90 |
+
-0.17599809169769287,
|
| 91 |
+
0.14405515789985657,
|
| 92 |
+
-1.6743669509887695,
|
| 93 |
+
-0.13138747215270996,
|
| 94 |
+
-0.04509587958455086,
|
| 95 |
+
0.31651967763900757,
|
| 96 |
+
0.3027613162994385
|
| 97 |
+
],
|
| 98 |
+
"std": [
|
| 99 |
+
0.29741761088371277,
|
| 100 |
+
0.151063472032547,
|
| 101 |
+
0.20714345574378967,
|
| 102 |
+
0.19010624289512634,
|
| 103 |
+
0.17496848106384277,
|
| 104 |
+
0.4267149567604065,
|
| 105 |
+
0.46089720726013184,
|
| 106 |
+
0.3306356966495514
|
| 107 |
+
],
|
| 108 |
+
"q01": [
|
| 109 |
+
-0.6769078516960144,
|
| 110 |
+
-0.6911599922180176,
|
| 111 |
+
-0.3895208740234375,
|
| 112 |
+
-2.1717092990875244,
|
| 113 |
+
-0.6094356870651245,
|
| 114 |
+
-0.6896490097045899,
|
| 115 |
+
-0.2534691643714905,
|
| 116 |
+
0.0044554052874445915
|
| 117 |
+
],
|
| 118 |
+
"q99": [
|
| 119 |
+
0.527884025573731,
|
| 120 |
+
0.0,
|
| 121 |
+
0.7277799344062805,
|
| 122 |
+
-1.2496684312820432,
|
| 123 |
+
0.45819765686988895,
|
| 124 |
+
0.7940996932983398,
|
| 125 |
+
1.3328254985809327,
|
| 126 |
+
1.058901309967041
|
| 127 |
+
]
|
| 128 |
+
}
|
| 129 |
+
},
|
| 130 |
+
"action": {
|
| 131 |
+
"left_arm": {
|
| 132 |
+
"max": [
|
| 133 |
+
0.5414952039718628,
|
| 134 |
+
0.8114758133888245,
|
| 135 |
+
0.6427379250526428,
|
| 136 |
+
-0.8222137093544006,
|
| 137 |
+
1.165825366973877,
|
| 138 |
+
1.0492428541183472,
|
| 139 |
+
0.23623304069042206,
|
| 140 |
+
1.2065826654434204
|
| 141 |
+
],
|
| 142 |
+
"min": [
|
| 143 |
+
-1.052310824394226,
|
| 144 |
+
-0.08130098134279251,
|
| 145 |
+
-1.2593982219696045,
|
| 146 |
+
-2.297903299331665,
|
| 147 |
+
-0.07056311517953873,
|
| 148 |
+
-0.5859806537628174,
|
| 149 |
+
-1.5815341472625732,
|
| 150 |
+
-0.11570879071950912
|
| 151 |
+
],
|
| 152 |
+
"mean": [
|
| 153 |
+
0.04731864482164383,
|
| 154 |
+
0.20576359331607819,
|
| 155 |
+
-0.1530255377292633,
|
| 156 |
+
-1.5469857454299927,
|
| 157 |
+
0.2660427391529083,
|
| 158 |
+
-0.07302618026733398,
|
| 159 |
+
-0.3637376129627228,
|
| 160 |
+
0.29268550872802734
|
| 161 |
+
],
|
| 162 |
+
"std": [
|
| 163 |
+
0.35969358682632446,
|
| 164 |
+
0.15887367725372314,
|
| 165 |
+
0.22229817509651184,
|
| 166 |
+
0.2142370641231537,
|
| 167 |
+
0.20440088212490082,
|
| 168 |
+
0.36180412769317627,
|
| 169 |
+
0.49843642115592957,
|
| 170 |
+
0.3544231057167053
|
| 171 |
+
],
|
| 172 |
+
"q01": [
|
| 173 |
+
-0.855961263179779,
|
| 174 |
+
-0.026077672839164734,
|
| 175 |
+
-0.8284723687171937,
|
| 176 |
+
-2.0202527046203613,
|
| 177 |
+
0.003067961661145091,
|
| 178 |
+
-0.5460971593856812,
|
| 179 |
+
-1.3867186307907104,
|
| 180 |
+
0.02848540060222149
|
| 181 |
+
],
|
| 182 |
+
"q99": [
|
| 183 |
+
0.424912691116333,
|
| 184 |
+
0.7179030179977417,
|
| 185 |
+
0.35281556844711304,
|
| 186 |
+
-1.0016894340515137,
|
| 187 |
+
0.8927768468856812,
|
| 188 |
+
0.7378447651863098,
|
| 189 |
+
0.1871456503868103,
|
| 190 |
+
1.1714238452911403
|
| 191 |
+
]
|
| 192 |
+
},
|
| 193 |
+
"right_arm": {
|
| 194 |
+
"max": [
|
| 195 |
+
0.6442719101905823,
|
| 196 |
+
0.05368932709097862,
|
| 197 |
+
1.2179807424545288,
|
| 198 |
+
-1.0369709730148315,
|
| 199 |
+
0.754718542098999,
|
| 200 |
+
0.9894176125526428,
|
| 201 |
+
1.5769321918487549,
|
| 202 |
+
1.3139612674713135
|
| 203 |
+
],
|
| 204 |
+
"min": [
|
| 205 |
+
-0.905048668384552,
|
| 206 |
+
-0.8375535011291504,
|
| 207 |
+
-0.6059224009513855,
|
| 208 |
+
-2.3469905853271484,
|
| 209 |
+
-0.7409127354621887,
|
| 210 |
+
-0.7792622447013855,
|
| 211 |
+
-0.3666214048862457,
|
| 212 |
+
-0.14638841152191162
|
| 213 |
+
],
|
| 214 |
+
"mean": [
|
| 215 |
+
0.14115391671657562,
|
| 216 |
+
-0.17551517486572266,
|
| 217 |
+
0.14400339126586914,
|
| 218 |
+
-1.6742379665374756,
|
| 219 |
+
-0.1316625475883484,
|
| 220 |
+
-0.045258209109306335,
|
| 221 |
+
0.316999614238739,
|
| 222 |
+
0.3188447654247284
|
| 223 |
+
],
|
| 224 |
+
"std": [
|
| 225 |
+
0.2977924644947052,
|
| 226 |
+
0.15177558362483978,
|
| 227 |
+
0.20768530666828156,
|
| 228 |
+
0.19047634303569794,
|
| 229 |
+
0.17517884075641632,
|
| 230 |
+
0.42713773250579834,
|
| 231 |
+
0.4618784189224243,
|
| 232 |
+
0.37785831093788147
|
| 233 |
+
],
|
| 234 |
+
"q01": [
|
| 235 |
+
-0.6780195236206055,
|
| 236 |
+
-0.6918253302574158,
|
| 237 |
+
-0.3911651074886322,
|
| 238 |
+
-2.172116756439209,
|
| 239 |
+
-0.6105243563652039,
|
| 240 |
+
-0.6902913451194763,
|
| 241 |
+
-0.25310683250427246,
|
| 242 |
+
-0.0313398502767086
|
| 243 |
+
],
|
| 244 |
+
"q99": [
|
| 245 |
+
0.5292233824729919,
|
| 246 |
+
0.019941750913858414,
|
| 247 |
+
0.7286408543586731,
|
| 248 |
+
-1.2486603260040283,
|
| 249 |
+
0.4601942300796509,
|
| 250 |
+
0.7961360216140747,
|
| 251 |
+
1.3330292701721191,
|
| 252 |
+
1.2740777730941772
|
| 253 |
+
]
|
| 254 |
+
}
|
| 255 |
+
}
|
| 256 |
+
},
|
| 257 |
+
"modalities": {
|
| 258 |
+
"video": {
|
| 259 |
+
"cam_head": {
|
| 260 |
+
"resolution": [
|
| 261 |
+
672,
|
| 262 |
+
376
|
| 263 |
+
],
|
| 264 |
+
"channels": 3,
|
| 265 |
+
"fps": 10.0
|
| 266 |
+
}
|
| 267 |
+
},
|
| 268 |
+
"state": {
|
| 269 |
+
"left_arm": {
|
| 270 |
+
"absolute": true,
|
| 271 |
+
"rotation_type": null,
|
| 272 |
+
"shape": [
|
| 273 |
+
8
|
| 274 |
+
],
|
| 275 |
+
"continuous": true
|
| 276 |
+
},
|
| 277 |
+
"right_arm": {
|
| 278 |
+
"absolute": true,
|
| 279 |
+
"rotation_type": null,
|
| 280 |
+
"shape": [
|
| 281 |
+
8
|
| 282 |
+
],
|
| 283 |
+
"continuous": true
|
| 284 |
+
}
|
| 285 |
+
},
|
| 286 |
+
"action": {
|
| 287 |
+
"left_arm": {
|
| 288 |
+
"absolute": true,
|
| 289 |
+
"rotation_type": null,
|
| 290 |
+
"shape": [
|
| 291 |
+
8
|
| 292 |
+
],
|
| 293 |
+
"continuous": true
|
| 294 |
+
},
|
| 295 |
+
"right_arm": {
|
| 296 |
+
"absolute": true,
|
| 297 |
+
"rotation_type": null,
|
| 298 |
+
"shape": [
|
| 299 |
+
8
|
| 300 |
+
],
|
| 301 |
+
"continuous": true
|
| 302 |
+
}
|
| 303 |
+
}
|
| 304 |
+
},
|
| 305 |
+
"embodiment_tag": "new_embodiment"
|
| 306 |
+
}
|
| 307 |
+
}
|
model-00001-of-00002.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:8506da0ddeafcef536c0f16dfe1ad11ffa4f802ba6ccf261fefdb76a95b82aee
|
| 3 |
+
size 4999367032
|
model-00002-of-00002.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:4162e4845d89f0a94d6ecd29c023ba1ec074115a372317fde52838437565304f
|
| 3 |
+
size 2586705312
|
model.safetensors.index.json
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
trainer_state.json
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|