Hugging Face's logo

InternRobotics
/
Gr00t-n1-5_Genmanip_IROS

Safetensors
gr00t_n1_5
Model card Files
xet
Community
JiantongChen commited on
Commit
d2eda09
·
verified ·
1 Parent(s): d020fe2

Upload folder using huggingface_hub

Browse files
Files changed (8) hide show
  1. config.json +64 -0
  2. experiment_cfg/metadata.json +399 -0
  3. model-00001-of-00002.safetensors +3 -0
  4. model-00002-of-00002.safetensors +3 -0
  5. model.safetensors.index.json +0 -0
  6. optimizer.pt +3 -0
  7. scheduler.pt +3 -0
  8. 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 +1,399 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "new_embodiment": {
3
+ "statistics": {
4
+ "state": {
5
+ "left_arm_qpos": {
6
+ "max": [
7
+ 0.9817636609077454,
8
+ 2.906359910964966,
9
+ 0.0004537280183285475,
10
+ 1.62539541721344,
11
+ 1.2199844121932983,
12
+ 1.2372479438781738
13
+ ],
14
+ "min": [
15
+ -1.5137280225753784,
16
+ -0.0004175282665528357,
17
+ -2.1384406089782715,
18
+ -1.8323923349380493,
19
+ -0.9081650376319885,
20
+ -2.397031545639038
21
+ ],
22
+ "mean": [
23
+ -0.1269621112350358,
24
+ 0.9319823582745339,
25
+ -0.4433606048262477,
26
+ -0.032110080556062154,
27
+ 0.2017761602008392,
28
+ -0.09985949073608533
29
+ ],
30
+ "std": [
31
+ 0.377008045487094,
32
+ 1.0305285652120173,
33
+ 0.5506187195898458,
34
+ 0.24978242466781378,
35
+ 0.3390767382656945,
36
+ 0.4371293925859658
37
+ ],
38
+ "q01": [
39
+ -0.9669479273124257,
40
+ 0.0023987626535762726,
41
+ -1.647388130082713,
42
+ -0.3212952227599285,
43
+ -0.257721101516588,
44
+ -1.184693200826756
45
+ ],
46
+ "q99": [
47
+ 0.4481088505177243,
48
+ 2.621667064048485,
49
+ -0.0012410853982822464,
50
+ 0.212910183930935,
51
+ 1.0136083454216869,
52
+ 0.4868166104189366
53
+ ]
54
+ },
55
+ "left_gripper_qpos_state": {
56
+ "max": [
57
+ 0.05000000074505806,
58
+ 0.0
59
+ ],
60
+ "min": [
61
+ 0.0,
62
+ -0.05000000074505806
63
+ ],
64
+ "mean": [
65
+ 0.04510546892183573,
66
+ -0.04513937599222491
67
+ ],
68
+ "std": [
69
+ 0.010394870452274656,
70
+ 0.010331194536151915
71
+ ],
72
+ "q01": [
73
+ 0.020937581471998583,
74
+ -0.05000000074505806
75
+ ],
76
+ "q99": [
77
+ 0.05000000074505806,
78
+ -0.021065237085878364
79
+ ]
80
+ },
81
+ "right_arm_qpos": {
82
+ "max": [
83
+ 1.48847496509552,
84
+ 2.892174243927002,
85
+ 1.352920207864372e-06,
86
+ 1.8329473733901978,
87
+ 1.2205382585525513,
88
+ 2.641817092895508
89
+ ],
90
+ "min": [
91
+ -0.9657754898071289,
92
+ -0.00018431521311867982,
93
+ -1.892966628074646,
94
+ -1.644637942314148,
95
+ -0.9023932814598083,
96
+ -1.1317652463912964
97
+ ],
98
+ "mean": [
99
+ 0.12632652331730418,
100
+ 0.8622424060454625,
101
+ -0.40988293050855507,
102
+ -0.005330470451327176,
103
+ 0.18023183290270667,
104
+ 0.13498211124956155
105
+ ],
106
+ "std": [
107
+ 0.3537567895522101,
108
+ 1.0054424500955776,
109
+ 0.5274209876633692,
110
+ 0.22274318100343346,
111
+ 0.3337768562289164,
112
+ 0.4237858376009564
113
+ ],
114
+ "q01": [
115
+ -0.4325246776494839,
116
+ 0.010349992344482523,
117
+ -1.5294983993623417,
118
+ -0.2701776041601203,
119
+ -0.24024883585539544,
120
+ -0.43958222498311184
121
+ ],
122
+ "q99": [
123
+ 0.9112513194214247,
124
+ 2.499901761404769,
125
+ -0.004218170142023899,
126
+ 0.29908044623713415,
127
+ 0.9595966412693012,
128
+ 1.1493524852754158
129
+ ]
130
+ },
131
+ "right_gripper_qpos_state": {
132
+ "max": [
133
+ 0.05000000074505806,
134
+ 0.0
135
+ ],
136
+ "min": [
137
+ 0.0,
138
+ -0.05000000074505806
139
+ ],
140
+ "mean": [
141
+ 0.04527297065813722,
142
+ -0.04531057170799728
143
+ ],
144
+ "std": [
145
+ 0.01060762645669795,
146
+ 0.010548578468195874
147
+ ],
148
+ "q01": [
149
+ 0.021579730762894184,
150
+ -0.04999995875952528
151
+ ],
152
+ "q99": [
153
+ 0.05000000074505806,
154
+ -0.021709689309794392
155
+ ]
156
+ }
157
+ },
158
+ "action": {
159
+ "left_arm_delta_qpos": {
160
+ "max": [
161
+ 0.029992341995239258,
162
+ 0.06384532898664474,
163
+ 0.030702590942382812,
164
+ 0.038200508803129196,
165
+ 0.028134286403656006,
166
+ 0.030016690492630005
167
+ ],
168
+ "min": [
169
+ -0.030527114868164062,
170
+ -0.029552221298217773,
171
+ -0.04401838034391403,
172
+ -0.04167499765753746,
173
+ -0.02967977523803711,
174
+ -0.06006285548210144
175
+ ],
176
+ "mean": [
177
+ 3.821466444864929e-05,
178
+ 0.00042845109823860084,
179
+ -0.00014026803222598545,
180
+ -2.8946628297871644e-06,
181
+ 2.817179753351224e-05,
182
+ 4.053795714823315e-05
183
+ ],
184
+ "std": [
185
+ 0.005530756034543259,
186
+ 0.011541299616515956,
187
+ 0.006498215104251148,
188
+ 0.003414693862834562,
189
+ 0.0053835734243439515,
190
+ 0.005841196531387051
191
+ ],
192
+ "q01": [
193
+ -0.021947088290789014,
194
+ -0.022308066073713374,
195
+ -0.021296351682562116,
196
+ -0.004892901405840323,
197
+ -0.013754208016569158,
198
+ -0.022779726505762037
199
+ ],
200
+ "q99": [
201
+ 0.011816893430933539,
202
+ 0.02992206951781668,
203
+ 0.015795769668274826,
204
+ 0.002640353098191997,
205
+ 0.01583988195319716,
206
+ 0.012934410068475436
207
+ ]
208
+ },
209
+ "right_arm_delta_qpos": {
210
+ "max": [
211
+ 0.026935935020446777,
212
+ 0.039149198681116104,
213
+ 0.030003011226654053,
214
+ 0.06768814474344254,
215
+ 0.050354115664958954,
216
+ 0.10000000149011612
217
+ ],
218
+ "min": [
219
+ -0.043110620230436325,
220
+ -0.04292352870106697,
221
+ -0.029925629496574402,
222
+ -0.026737341657280922,
223
+ -0.05000000074505806,
224
+ -0.10000000149011612
225
+ ],
226
+ "mean": [
227
+ -0.00010897532197173145,
228
+ -4.699495506415459e-07,
229
+ 3.356031398538973e-05,
230
+ 4.889761850156645e-05,
231
+ 5.05030139675559e-05,
232
+ -1.5763375481563124e-05
233
+ ],
234
+ "std": [
235
+ 0.005992895893009324,
236
+ 0.0029073012451196986,
237
+ 0.005445022217853271,
238
+ 0.005851184941486254,
239
+ 0.002859905935108947,
240
+ 0.00359616601056668
241
+ ],
242
+ "q01": [
243
+ -0.019669866170712546,
244
+ -0.0025259607834505725,
245
+ -0.014140649346943488,
246
+ -0.011932997287101507,
247
+ 0.0,
248
+ 0.0
249
+ ],
250
+ "q99": [
251
+ 0.014196241835007837,
252
+ 0.004579077300663446,
253
+ 0.016790677413677842,
254
+ 0.022628532019240522,
255
+ 0.0,
256
+ 0.0
257
+ ]
258
+ },
259
+ "left_gripper_close": {
260
+ "max": [
261
+ 1.0
262
+ ],
263
+ "min": [
264
+ -1.0
265
+ ],
266
+ "mean": [
267
+ -0.5990138348418236
268
+ ],
269
+ "std": [
270
+ 0.8005982900647917
271
+ ],
272
+ "q01": [
273
+ -1.0
274
+ ],
275
+ "q99": [
276
+ 1.0
277
+ ]
278
+ },
279
+ "right_gripper_close": {
280
+ "max": [
281
+ 1.0
282
+ ],
283
+ "min": [
284
+ -1.0
285
+ ],
286
+ "mean": [
287
+ -0.6291168778750481
288
+ ],
289
+ "std": [
290
+ 0.7773056224616752
291
+ ],
292
+ "q01": [
293
+ -1.0
294
+ ],
295
+ "q99": [
296
+ 0.92006804513259
297
+ ]
298
+ }
299
+ }
300
+ },
301
+ "modalities": {
302
+ "video": {
303
+ "left_view": {
304
+ "resolution": [
305
+ 640,
306
+ 480
307
+ ],
308
+ "channels": 3,
309
+ "fps": 30.0
310
+ },
311
+ "right_view": {
312
+ "resolution": [
313
+ 640,
314
+ 480
315
+ ],
316
+ "channels": 3,
317
+ "fps": 30.0
318
+ },
319
+ "top_view": {
320
+ "resolution": [
321
+ 640,
322
+ 480
323
+ ],
324
+ "channels": 3,
325
+ "fps": 30.0
326
+ }
327
+ },
328
+ "state": {
329
+ "left_arm_qpos": {
330
+ "absolute": true,
331
+ "rotation_type": null,
332
+ "shape": [
333
+ 6
334
+ ],
335
+ "continuous": true
336
+ },
337
+ "left_gripper_qpos_state": {
338
+ "absolute": true,
339
+ "rotation_type": null,
340
+ "shape": [
341
+ 2
342
+ ],
343
+ "continuous": true
344
+ },
345
+ "right_arm_qpos": {
346
+ "absolute": true,
347
+ "rotation_type": null,
348
+ "shape": [
349
+ 6
350
+ ],
351
+ "continuous": true
352
+ },
353
+ "right_gripper_qpos_state": {
354
+ "absolute": true,
355
+ "rotation_type": null,
356
+ "shape": [
357
+ 2
358
+ ],
359
+ "continuous": true
360
+ }
361
+ },
362
+ "action": {
363
+ "left_arm_delta_qpos": {
364
+ "absolute": true,
365
+ "rotation_type": null,
366
+ "shape": [
367
+ 6
368
+ ],
369
+ "continuous": true
370
+ },
371
+ "right_arm_delta_qpos": {
372
+ "absolute": true,
373
+ "rotation_type": null,
374
+ "shape": [
375
+ 6
376
+ ],
377
+ "continuous": true
378
+ },
379
+ "left_gripper_close": {
380
+ "absolute": true,
381
+ "rotation_type": null,
382
+ "shape": [
383
+ 1
384
+ ],
385
+ "continuous": true
386
+ },
387
+ "right_gripper_close": {
388
+ "absolute": true,
389
+ "rotation_type": null,
390
+ "shape": [
391
+ 1
392
+ ],
393
+ "continuous": true
394
+ }
395
+ }
396
+ },
397
+ "embodiment_tag": "new_embodiment"
398
+ }
399
+ }
model-00001-of-00002.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1237b6c130b7a9154dce0248ec34002038af28d8056eb7bac6e92316f60d8802
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:f67acc20693f0543fce2dd446b5f943b32c5b1469189efd1f89d364cc2b3df0d
3
+ size 2586508600
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:a71c3a1960b5f4b0e5ab92764f80a222f714f59b4746deb744e17170ef780ff2
3
+ size 8550325978
scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a6d34e46986ba4d0396b6e4487c1812a556c234dafb0e774d2716a4a97b41158
3
+ size 1064
trainer_state.json ADDED
The diff for this file is too large to render. See raw diff