Hugging Face's logo

youliangtan
/
gr00t-n1.5-bridge-posttrain

Safetensors
gr00t_n1_5
Model card Files
xet
Community
youliangtan commited on
Commit
942c1af
·
verified ·
1 Parent(s): 56dcb24

Upload folder using huggingface_hub

Browse files
Files changed (6) hide show
  1. config.json +64 -0
  2. experiment_cfg/metadata.json +447 -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. 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": "4.51.1"
64
+ }
experiment_cfg/metadata.json ADDED
@@ -0,0 +1,447 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "new_embodiment": {
3
+ "statistics": {
4
+ "state": {
5
+ "x": {
6
+ "max": [
7
+ 0.5862360596656799
8
+ ],
9
+ "min": [
10
+ -0.04167502000927925
11
+ ],
12
+ "mean": [
13
+ 0.3094348609447479
14
+ ],
15
+ "std": [
16
+ 0.06056898087263107
17
+ ],
18
+ "q01": [
19
+ 0.17102622985839844
20
+ ],
21
+ "q99": [
22
+ 0.45322076976299286
23
+ ]
24
+ },
25
+ "y": {
26
+ "max": [
27
+ 0.4034728705883026
28
+ ],
29
+ "min": [
30
+ -0.3563207685947418
31
+ ],
32
+ "mean": [
33
+ 0.03072510100901127
34
+ ],
35
+ "std": [
36
+ 0.09190694242715836
37
+ ],
38
+ "q01": [
39
+ -0.1698118858039379
40
+ ],
41
+ "q99": [
42
+ 0.2354845181107521
43
+ ]
44
+ },
45
+ "z": {
46
+ "max": [
47
+ 0.3568263053894043
48
+ ],
49
+ "min": [
50
+ -0.15537554025650024
51
+ ],
52
+ "mean": [
53
+ 0.06444056332111359
54
+ ],
55
+ "std": [
56
+ 0.05156300589442253
57
+ ],
58
+ "q01": [
59
+ -0.055632163770496845
60
+ ],
61
+ "q99": [
62
+ 0.19489620998501778
63
+ ]
64
+ },
65
+ "roll": {
66
+ "max": [
67
+ 1.3517684936523438
68
+ ],
69
+ "min": [
70
+ -3.141592502593994
71
+ ],
72
+ "mean": [
73
+ 0.006490640342235565
74
+ ],
75
+ "std": [
76
+ 0.13109275698661804
77
+ ],
78
+ "q01": [
79
+ -0.36493172496557236
80
+ ],
81
+ "q99": [
82
+ 0.3780156075954437
83
+ ]
84
+ },
85
+ "pitch": {
86
+ "max": [
87
+ 1.570796251296997
88
+ ],
89
+ "min": [
90
+ -1.4992541074752808
91
+ ],
92
+ "mean": [
93
+ -0.07720029354095459
94
+ ],
95
+ "std": [
96
+ 0.16913272440433502
97
+ ],
98
+ "q01": [
99
+ -0.541871190071106
100
+ ],
101
+ "q99": [
102
+ 0.27568644285202026
103
+ ]
104
+ },
105
+ "yaw": {
106
+ "max": [
107
+ 3.141204357147217
108
+ ],
109
+ "min": [
110
+ -3.14153790473938
111
+ ],
112
+ "mean": [
113
+ 0.10765893757343292
114
+ ],
115
+ "std": [
116
+ 0.5779858231544495
117
+ ],
118
+ "q01": [
119
+ -1.354275494813919
120
+ ],
121
+ "q99": [
122
+ 1.8500566184520721
123
+ ]
124
+ },
125
+ "pad": {
126
+ "max": [
127
+ 0.0
128
+ ],
129
+ "min": [
130
+ 0.0
131
+ ],
132
+ "mean": [
133
+ 0.0
134
+ ],
135
+ "std": [
136
+ 0.0
137
+ ],
138
+ "q01": [
139
+ 0.0
140
+ ],
141
+ "q99": [
142
+ 0.0
143
+ ]
144
+ },
145
+ "gripper": {
146
+ "max": [
147
+ 1.1121242046356201
148
+ ],
149
+ "min": [
150
+ 0.04637829214334488
151
+ ],
152
+ "mean": [
153
+ 0.7081330418586731
154
+ ],
155
+ "std": [
156
+ 0.35254859924316406
157
+ ],
158
+ "q01": [
159
+ 0.052190229296684265
160
+ ],
161
+ "q99": [
162
+ 1.0105689764022827
163
+ ]
164
+ }
165
+ },
166
+ "action": {
167
+ "x": {
168
+ "max": [
169
+ 0.41691166162490845
170
+ ],
171
+ "min": [
172
+ -0.4007510244846344
173
+ ],
174
+ "mean": [
175
+ 0.00022731871285941452
176
+ ],
177
+ "std": [
178
+ 0.009770569391548634
179
+ ],
180
+ "q01": [
181
+ -0.02875255048274994
182
+ ],
183
+ "q99": [
184
+ 0.028306663036346436
185
+ ]
186
+ },
187
+ "y": {
188
+ "max": [
189
+ 0.25864794850349426
190
+ ],
191
+ "min": [
192
+ -0.13874775171279907
193
+ ],
194
+ "mean": [
195
+ 0.00013112221495248377
196
+ ],
197
+ "std": [
198
+ 0.013695062138140202
199
+ ],
200
+ "q01": [
201
+ -0.04170213546603918
202
+ ],
203
+ "q99": [
204
+ 0.04089853074401617
205
+ ]
206
+ },
207
+ "z": {
208
+ "max": [
209
+ 0.21218234300613403
210
+ ],
211
+ "min": [
212
+ -0.22553899884223938
213
+ ],
214
+ "mean": [
215
+ -0.00012641931243706495
216
+ ],
217
+ "std": [
218
+ 0.012675146572291851
219
+ ],
220
+ "q01": [
221
+ -0.026096721179783344
222
+ ],
223
+ "q99": [
224
+ 0.0401805154979229
225
+ ]
226
+ },
227
+ "roll": {
228
+ "max": [
229
+ 3.122201919555664
230
+ ],
231
+ "min": [
232
+ -3.2010786533355713
233
+ ],
234
+ "mean": [
235
+ -0.00014410706353373826
236
+ ],
237
+ "std": [
238
+ 0.028455283492803574
239
+ ],
240
+ "q01": [
241
+ -0.08052874729037285
242
+ ],
243
+ "q99": [
244
+ 0.08173403143882751
245
+ ]
246
+ },
247
+ "pitch": {
248
+ "max": [
249
+ 1.8618112802505493
250
+ ],
251
+ "min": [
252
+ -1.8618112802505493
253
+ ],
254
+ "mean": [
255
+ -0.00039030605694279075
256
+ ],
257
+ "std": [
258
+ 0.03052123636007309
259
+ ],
260
+ "q01": [
261
+ -0.09249906800687313
262
+ ],
263
+ "q99": [
264
+ 0.07760760560631752
265
+ ]
266
+ },
267
+ "yaw": {
268
+ "max": [
269
+ 6.272472858428955
270
+ ],
271
+ "min": [
272
+ -6.279075622558594
273
+ ],
274
+ "mean": [
275
+ 0.0002406332059763372
276
+ ],
277
+ "std": [
278
+ 0.07739030569791794
279
+ ],
280
+ "q01": [
281
+ -0.20738555490970612
282
+ ],
283
+ "q99": [
284
+ 0.2038465365767479
285
+ ]
286
+ },
287
+ "gripper": {
288
+ "max": [
289
+ 1.0
290
+ ],
291
+ "min": [
292
+ 0.0
293
+ ],
294
+ "mean": [
295
+ 0.5765891671180725
296
+ ],
297
+ "std": [
298
+ 0.4966523349285126
299
+ ],
300
+ "q01": [
301
+ 0.0
302
+ ],
303
+ "q99": [
304
+ 1.0
305
+ ]
306
+ }
307
+ }
308
+ },
309
+ "modalities": {
310
+ "video": {
311
+ "image_0": {
312
+ "resolution": [
313
+ 256,
314
+ 256
315
+ ],
316
+ "channels": 3,
317
+ "fps": 5.0
318
+ }
319
+ },
320
+ "state": {
321
+ "x": {
322
+ "absolute": true,
323
+ "rotation_type": null,
324
+ "shape": [
325
+ 1
326
+ ],
327
+ "continuous": true
328
+ },
329
+ "y": {
330
+ "absolute": true,
331
+ "rotation_type": null,
332
+ "shape": [
333
+ 1
334
+ ],
335
+ "continuous": true
336
+ },
337
+ "z": {
338
+ "absolute": true,
339
+ "rotation_type": null,
340
+ "shape": [
341
+ 1
342
+ ],
343
+ "continuous": true
344
+ },
345
+ "roll": {
346
+ "absolute": true,
347
+ "rotation_type": null,
348
+ "shape": [
349
+ 1
350
+ ],
351
+ "continuous": true
352
+ },
353
+ "pitch": {
354
+ "absolute": true,
355
+ "rotation_type": null,
356
+ "shape": [
357
+ 1
358
+ ],
359
+ "continuous": true
360
+ },
361
+ "yaw": {
362
+ "absolute": true,
363
+ "rotation_type": null,
364
+ "shape": [
365
+ 1
366
+ ],
367
+ "continuous": true
368
+ },
369
+ "pad": {
370
+ "absolute": true,
371
+ "rotation_type": null,
372
+ "shape": [
373
+ 1
374
+ ],
375
+ "continuous": true
376
+ },
377
+ "gripper": {
378
+ "absolute": true,
379
+ "rotation_type": null,
380
+ "shape": [
381
+ 1
382
+ ],
383
+ "continuous": true
384
+ }
385
+ },
386
+ "action": {
387
+ "x": {
388
+ "absolute": true,
389
+ "rotation_type": null,
390
+ "shape": [
391
+ 1
392
+ ],
393
+ "continuous": true
394
+ },
395
+ "y": {
396
+ "absolute": true,
397
+ "rotation_type": null,
398
+ "shape": [
399
+ 1
400
+ ],
401
+ "continuous": true
402
+ },
403
+ "z": {
404
+ "absolute": true,
405
+ "rotation_type": null,
406
+ "shape": [
407
+ 1
408
+ ],
409
+ "continuous": true
410
+ },
411
+ "roll": {
412
+ "absolute": true,
413
+ "rotation_type": null,
414
+ "shape": [
415
+ 1
416
+ ],
417
+ "continuous": true
418
+ },
419
+ "pitch": {
420
+ "absolute": true,
421
+ "rotation_type": null,
422
+ "shape": [
423
+ 1
424
+ ],
425
+ "continuous": true
426
+ },
427
+ "yaw": {
428
+ "absolute": true,
429
+ "rotation_type": null,
430
+ "shape": [
431
+ 1
432
+ ],
433
+ "continuous": true
434
+ },
435
+ "gripper": {
436
+ "absolute": true,
437
+ "rotation_type": null,
438
+ "shape": [
439
+ 1
440
+ ],
441
+ "continuous": true
442
+ }
443
+ }
444
+ },
445
+ "embodiment_tag": "new_embodiment"
446
+ }
447
+ }
model-00001-of-00002.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:dd814c70340b5d074205930f7618995fc35ba834baece140ef88e8a25294a989
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:ceb0c5db84ed4a0feb3aa1cf9fdf38d05612acf4b2f14e1dcf575ef85dde3972
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