Add files using upload-large-folder tool
Browse files- config.json +64 -0
- experiment_cfg/metadata.json +219 -0
- model-00001-of-00002.safetensors +3 -0
- model-00002-of-00002.safetensors +3 -0
- model.safetensors.index.json +0 -0
- optimizer.pt +3 -0
- rng_state_0.pth +3 -0
- rng_state_1.pth +3 -0
- scheduler.pt +3 -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": 64,
|
| 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 +1,219 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"oxe": {
|
| 3 |
+
"statistics": {
|
| 4 |
+
"state": {
|
| 5 |
+
"joint_pos_abs": {
|
| 6 |
+
"max": [
|
| 7 |
+
1.4700034856796265,
|
| 8 |
+
1.4340834617614746,
|
| 9 |
+
0.8115426898002625,
|
| 10 |
+
-0.37125325202941895,
|
| 11 |
+
1.0951130390167236,
|
| 12 |
+
3.3310701847076416,
|
| 13 |
+
0.9769507646560669
|
| 14 |
+
],
|
| 15 |
+
"min": [
|
| 16 |
+
-0.8163706660270691,
|
| 17 |
+
-0.5892691612243652,
|
| 18 |
+
-0.9640398621559143,
|
| 19 |
+
-2.803720474243164,
|
| 20 |
+
-0.683822512626648,
|
| 21 |
+
1.2045881748199463,
|
| 22 |
+
-0.8131589889526367
|
| 23 |
+
],
|
| 24 |
+
"mean": [
|
| 25 |
+
-0.027940594168493794,
|
| 26 |
+
0.6596433834598774,
|
| 27 |
+
-0.031911526851368485,
|
| 28 |
+
-1.6408408484011878,
|
| 29 |
+
0.037508157039820744,
|
| 30 |
+
1.9986611685105218,
|
| 31 |
+
0.04911849992521497
|
| 32 |
+
],
|
| 33 |
+
"std": [
|
| 34 |
+
0.0894720809500577,
|
| 35 |
+
0.4762244043735688,
|
| 36 |
+
0.06392021830146798,
|
| 37 |
+
0.6111356214569408,
|
| 38 |
+
0.09202056944650079,
|
| 39 |
+
0.317005110088692,
|
| 40 |
+
0.18209302855943732
|
| 41 |
+
],
|
| 42 |
+
"q01": [
|
| 43 |
+
-0.1312229024890012,
|
| 44 |
+
-0.555360202134788,
|
| 45 |
+
-0.14311420377155637,
|
| 46 |
+
-2.6676338773962915,
|
| 47 |
+
-0.09753329225679075,
|
| 48 |
+
1.6491974640294542,
|
| 49 |
+
-0.1537885292215507
|
| 50 |
+
],
|
| 51 |
+
"q99": [
|
| 52 |
+
0.11126110298403323,
|
| 53 |
+
1.1427102520179837,
|
| 54 |
+
0.06764181893031622,
|
| 55 |
+
-0.950655298467393,
|
| 56 |
+
0.1851135870516939,
|
| 57 |
+
2.5313520646833525,
|
| 58 |
+
0.2979468408337512
|
| 59 |
+
]
|
| 60 |
+
},
|
| 61 |
+
"gripper_close": {
|
| 62 |
+
"max": [
|
| 63 |
+
0.8634361028671265
|
| 64 |
+
],
|
| 65 |
+
"min": [
|
| 66 |
+
0.0
|
| 67 |
+
],
|
| 68 |
+
"mean": [
|
| 69 |
+
0.23236840231418943
|
| 70 |
+
],
|
| 71 |
+
"std": [
|
| 72 |
+
0.22722538294108865
|
| 73 |
+
],
|
| 74 |
+
"q01": [
|
| 75 |
+
-1.000000013351432e-10
|
| 76 |
+
],
|
| 77 |
+
"q99": [
|
| 78 |
+
0.5155544382599047
|
| 79 |
+
]
|
| 80 |
+
}
|
| 81 |
+
},
|
| 82 |
+
"action": {
|
| 83 |
+
"joint_pos_abs": {
|
| 84 |
+
"max": [
|
| 85 |
+
1.500838041305542,
|
| 86 |
+
1.450729250907898,
|
| 87 |
+
0.8541431427001953,
|
| 88 |
+
-0.33947330713272095,
|
| 89 |
+
1.1617882251739502,
|
| 90 |
+
3.3409955501556396,
|
| 91 |
+
0.9983596205711365
|
| 92 |
+
],
|
| 93 |
+
"min": [
|
| 94 |
+
-0.8593479990959167,
|
| 95 |
+
-0.5840162038803101,
|
| 96 |
+
-1.0138425827026367,
|
| 97 |
+
-2.8810651302337646,
|
| 98 |
+
-0.7228302955627441,
|
| 99 |
+
1.145877718925476,
|
| 100 |
+
-0.8519567251205444
|
| 101 |
+
],
|
| 102 |
+
"mean": [
|
| 103 |
+
-0.033636524315637434,
|
| 104 |
+
0.6610624029742379,
|
| 105 |
+
-0.027039617710828998,
|
| 106 |
+
-1.60411367943591,
|
| 107 |
+
0.040267942887403076,
|
| 108 |
+
1.9767596899447475,
|
| 109 |
+
0.04430793626427893
|
| 110 |
+
],
|
| 111 |
+
"std": [
|
| 112 |
+
0.09693623941367845,
|
| 113 |
+
0.4571853752394202,
|
| 114 |
+
0.06754284766194828,
|
| 115 |
+
0.6131280073483107,
|
| 116 |
+
0.09923421454101686,
|
| 117 |
+
0.3281667624377318,
|
| 118 |
+
0.19031844948755644
|
| 119 |
+
],
|
| 120 |
+
"q01": [
|
| 121 |
+
-0.15580597237434826,
|
| 122 |
+
-0.5482164328556879,
|
| 123 |
+
-0.148363991241337,
|
| 124 |
+
-2.6469299193499722,
|
| 125 |
+
-0.11701282571570755,
|
| 126 |
+
1.61094377607778,
|
| 127 |
+
-0.17864016467989705
|
| 128 |
+
],
|
| 129 |
+
"q99": [
|
| 130 |
+
0.1192240869492337,
|
| 131 |
+
1.150366396519556,
|
| 132 |
+
0.07592797434800115,
|
| 133 |
+
-0.9070544020680874,
|
| 134 |
+
0.20428732634092767,
|
| 135 |
+
2.527201384560406,
|
| 136 |
+
0.31139192295501966
|
| 137 |
+
]
|
| 138 |
+
},
|
| 139 |
+
"gripper_close": {
|
| 140 |
+
"max": [
|
| 141 |
+
1.0
|
| 142 |
+
],
|
| 143 |
+
"min": [
|
| 144 |
+
0.0
|
| 145 |
+
],
|
| 146 |
+
"mean": [
|
| 147 |
+
0.34991235082714245
|
| 148 |
+
],
|
| 149 |
+
"std": [
|
| 150 |
+
0.3468042078152479
|
| 151 |
+
],
|
| 152 |
+
"q01": [
|
| 153 |
+
-1.000000013351432e-10
|
| 154 |
+
],
|
| 155 |
+
"q99": [
|
| 156 |
+
0.7396587963132695
|
| 157 |
+
]
|
| 158 |
+
}
|
| 159 |
+
}
|
| 160 |
+
},
|
| 161 |
+
"modalities": {
|
| 162 |
+
"video": {
|
| 163 |
+
"exterior_image_1_left": {
|
| 164 |
+
"resolution": [
|
| 165 |
+
1280,
|
| 166 |
+
720
|
| 167 |
+
],
|
| 168 |
+
"channels": 3,
|
| 169 |
+
"fps": 10.0
|
| 170 |
+
},
|
| 171 |
+
"wrist_image_left": {
|
| 172 |
+
"resolution": [
|
| 173 |
+
1280,
|
| 174 |
+
720
|
| 175 |
+
],
|
| 176 |
+
"channels": 3,
|
| 177 |
+
"fps": 10.0
|
| 178 |
+
}
|
| 179 |
+
},
|
| 180 |
+
"state": {
|
| 181 |
+
"joint_pos_abs": {
|
| 182 |
+
"absolute": true,
|
| 183 |
+
"rotation_type": null,
|
| 184 |
+
"shape": [
|
| 185 |
+
7
|
| 186 |
+
],
|
| 187 |
+
"continuous": true
|
| 188 |
+
},
|
| 189 |
+
"gripper_close": {
|
| 190 |
+
"absolute": true,
|
| 191 |
+
"rotation_type": null,
|
| 192 |
+
"shape": [
|
| 193 |
+
1
|
| 194 |
+
],
|
| 195 |
+
"continuous": true
|
| 196 |
+
}
|
| 197 |
+
},
|
| 198 |
+
"action": {
|
| 199 |
+
"joint_pos_abs": {
|
| 200 |
+
"absolute": true,
|
| 201 |
+
"rotation_type": null,
|
| 202 |
+
"shape": [
|
| 203 |
+
7
|
| 204 |
+
],
|
| 205 |
+
"continuous": true
|
| 206 |
+
},
|
| 207 |
+
"gripper_close": {
|
| 208 |
+
"absolute": true,
|
| 209 |
+
"rotation_type": null,
|
| 210 |
+
"shape": [
|
| 211 |
+
1
|
| 212 |
+
],
|
| 213 |
+
"continuous": true
|
| 214 |
+
}
|
| 215 |
+
}
|
| 216 |
+
},
|
| 217 |
+
"embodiment_tag": "oxe"
|
| 218 |
+
}
|
| 219 |
+
}
|
model-00001-of-00002.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:3ec19c5758b0a2842e3367a3f30f15693707499da4e5369f3aaee1393fb59bd2
|
| 3 |
+
size 4999367032
|
model-00002-of-00002.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:442197e23150879c4fb0fd3d4f35be66e5720275bdc0317d65af8aba974f82e2
|
| 3 |
+
size 2604611128
|
model.safetensors.index.json
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
optimizer.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:e5d5f0a750e6c8f1df63069addc3f6b832f42d5df967a0a63306702fe40cdb68
|
| 3 |
+
size 8586540054
|
rng_state_0.pth
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:8ea77c3c8d7380d70b31844e7cb96c36fcda61d4a3e35fdfbd39da7028c0fab9
|
| 3 |
+
size 14512
|
rng_state_1.pth
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:49a5f1310a1b7db8ac5a3acab1cf37e770fb8eb0052d6c4ad504e745f74dc271
|
| 3 |
+
size 14512
|
scheduler.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:ac42b4efebda871e7035b931b3f1fd2282cf31a9d4249ad956a15227d00969cb
|
| 3 |
+
size 1064
|
trainer_state.json
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|