Hugging Face's logo

Tacoin
/
GR00T-N1.5-3B-LIBERO-LONG-8K

Robotics
Safetensors
gr00t
gr00t_n1_5
robot-learning
diffusion-policy
libero
long-horizon
Model card Files
xet
 Community
Tacoin commited on
Commit
ffa492d
·
verified ·
1 Parent(s): c8696d0

Upload libero_long_8k checkpoint

Browse files
Files changed (9) hide show
  1. config.json +64 -0
  2. experiment_cfg/metadata.json +455 -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. rng_state.pth +3 -0
  8. scheduler.pt +3 -0
  9. 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 +1,455 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "new_embodiment": {
3
+ "statistics": {
4
+ "state": {
5
+ "x": {
6
+ "max": [
7
+ 0.21031762659549713
8
+ ],
9
+ "min": [
10
+ -0.4828203022480011
11
+ ],
12
+ "mean": [
13
+ -0.04190658777952194
14
+ ],
15
+ "std": [
16
+ 0.10743364691734314
17
+ ],
18
+ "q01": [
19
+ -0.3899900782108307
20
+ ],
21
+ "q99": [
22
+ 0.1530261474847791
23
+ ]
24
+ },
25
+ "y": {
26
+ "max": [
27
+ 0.39128610491752625
28
+ ],
29
+ "min": [
30
+ -0.3255046010017395
31
+ ],
32
+ "mean": [
33
+ 0.03539430722594261
34
+ ],
35
+ "std": [
36
+ 0.14424669742584229
37
+ ],
38
+ "q01": [
39
+ -0.2838300323486328
40
+ ],
41
+ "q99": [
42
+ 0.32915401458740223
43
+ ]
44
+ },
45
+ "z": {
46
+ "max": [
47
+ 1.3332009315490723
48
+ ],
49
+ "min": [
50
+ 0.445506751537323
51
+ ],
52
+ "mean": [
53
+ 0.8257141709327698
54
+ ],
55
+ "std": [
56
+ 0.2572328448295593
57
+ ],
58
+ "q01": [
59
+ 0.44795057058334353
60
+ ],
61
+ "q99": [
62
+ 1.2546923208236693
63
+ ]
64
+ },
65
+ "roll": {
66
+ "max": [
67
+ 3.6714255809783936
68
+ ],
69
+ "min": [
70
+ 1.1321442127227783
71
+ ],
72
+ "mean": [
73
+ 2.908308267593384
74
+ ],
75
+ "std": [
76
+ 0.3441362977027893
77
+ ],
78
+ "q01": [
79
+ 1.8810229921340942
80
+ ],
81
+ "q99": [
82
+ 3.303542451858519
83
+ ]
84
+ },
85
+ "pitch": {
86
+ "max": [
87
+ 3.560650587081909
88
+ ],
89
+ "min": [
90
+ -3.641430377960205
91
+ ],
92
+ "mean": [
93
+ -0.5562185049057007
94
+ ],
95
+ "std": [
96
+ 1.234421730041504
97
+ ],
98
+ "q01": [
99
+ -2.886677579879761
100
+ ],
101
+ "q99": [
102
+ 2.7496529006957933
103
+ ]
104
+ },
105
+ "yaw": {
106
+ "max": [
107
+ 1.386339545249939
108
+ ],
109
+ "min": [
110
+ -1.842738389968872
111
+ ],
112
+ "mean": [
113
+ -0.16649018228054047
114
+ ],
115
+ "std": [
116
+ 0.3579835891723633
117
+ ],
118
+ "q01": [
119
+ -1.1599004411697387
120
+ ],
121
+ "q99": [
122
+ 0.6893712210655194
123
+ ]
124
+ },
125
+ "gripper": {
126
+ "max": [
127
+ 0.04160946607589722
128
+ ],
129
+ "min": [
130
+ -0.0010040868073701859
131
+ ],
132
+ "mean": [
133
+ 0.028316624462604523
134
+ ],
135
+ "std": [
136
+ 0.013308707624673843
137
+ ],
138
+ "q01": [
139
+ 0.002066459748893976
140
+ ],
141
+ "q99": [
142
+ 0.040048558115959164
143
+ ]
144
+ },
145
+ "extra": {
146
+ "max": [
147
+ 0.0013633022317662835
148
+ ],
149
+ "min": [
150
+ -0.04111652821302414
151
+ ],
152
+ "mean": [
153
+ -0.028561657294631004
154
+ ],
155
+ "std": [
156
+ 0.013174631632864475
157
+ ],
158
+ "q01": [
159
+ -0.04001387819647789
160
+ ],
161
+ "q99": [
162
+ -0.0017598449345678235
163
+ ]
164
+ }
165
+ },
166
+ "action": {
167
+ "x": {
168
+ "max": [
169
+ 0.9375
170
+ ],
171
+ "min": [
172
+ -0.9375
173
+ ],
174
+ "mean": [
175
+ 0.01820324920117855
176
+ ],
177
+ "std": [
178
+ 0.2825464606285095
179
+ ],
180
+ "q01": [
181
+ -0.6348214149475098
182
+ ],
183
+ "q99": [
184
+ 0.7714285850524902
185
+ ]
186
+ },
187
+ "y": {
188
+ "max": [
189
+ 0.9375
190
+ ],
191
+ "min": [
192
+ -0.9375
193
+ ],
194
+ "mean": [
195
+ 0.05858374014496803
196
+ ],
197
+ "std": [
198
+ 0.35904666781425476
199
+ ],
200
+ "q01": [
201
+ -0.7741071581840515
202
+ ],
203
+ "q99": [
204
+ 0.8464285731315613
205
+ ]
206
+ },
207
+ "z": {
208
+ "max": [
209
+ 0.9375
210
+ ],
211
+ "min": [
212
+ -0.9375
213
+ ],
214
+ "mean": [
215
+ -0.05592384561896324
216
+ ],
217
+ "std": [
218
+ 0.3673802614212036
219
+ ],
220
+ "q01": [
221
+ -0.7633928656578064
222
+ ],
223
+ "q99": [
224
+ 0.9375
225
+ ]
226
+ },
227
+ "roll": {
228
+ "max": [
229
+ 0.30000001192092896
230
+ ],
231
+ "min": [
232
+ -0.23642857372760773
233
+ ],
234
+ "mean": [
235
+ 0.004626928828656673
236
+ ],
237
+ "std": [
238
+ 0.03770702704787254
239
+ ],
240
+ "q01": [
241
+ -0.09749999642372131
242
+ ],
243
+ "q99": [
244
+ 0.13928571343421936
245
+ ]
246
+ },
247
+ "pitch": {
248
+ "max": [
249
+ 0.29357144236564636
250
+ ],
251
+ "min": [
252
+ -0.3053571283817291
253
+ ],
254
+ "mean": [
255
+ 0.00289608770981431
256
+ ],
257
+ "std": [
258
+ 0.05429719388484955
259
+ ],
260
+ "q01": [
261
+ -0.14819999992847435
262
+ ],
263
+ "q99": [
264
+ 0.15964286029338837
265
+ ]
266
+ },
267
+ "yaw": {
268
+ "max": [
269
+ 0.375
270
+ ],
271
+ "min": [
272
+ -0.3675000071525574
273
+ ],
274
+ "mean": [
275
+ -0.007673131301999092
276
+ ],
277
+ "std": [
278
+ 0.08725254982709885
279
+ ],
280
+ "q01": [
281
+ -0.2742857038974762
282
+ ],
283
+ "q99": [
284
+ 0.3246428668498993
285
+ ]
286
+ },
287
+ "gripper": {
288
+ "max": [
289
+ 1.0
290
+ ],
291
+ "min": [
292
+ -1.0
293
+ ],
294
+ "mean": [
295
+ -0.09156490862369537
296
+ ],
297
+ "std": [
298
+ 0.9963046312332153
299
+ ],
300
+ "q01": [
301
+ -1.0
302
+ ],
303
+ "q99": [
304
+ 1.0
305
+ ]
306
+ }
307
+ }
308
+ },
309
+ "modalities": {
310
+ "video": {
311
+ "image": {
312
+ "resolution": [
313
+ 256,
314
+ 256
315
+ ],
316
+ "channels": 3,
317
+ "fps": 10.0
318
+ },
319
+ "wrist_image": {
320
+ "resolution": [
321
+ 256,
322
+ 256
323
+ ],
324
+ "channels": 3,
325
+ "fps": 10.0
326
+ }
327
+ },
328
+ "state": {
329
+ "x": {
330
+ "absolute": true,
331
+ "rotation_type": null,
332
+ "shape": [
333
+ 1
334
+ ],
335
+ "continuous": true
336
+ },
337
+ "y": {
338
+ "absolute": true,
339
+ "rotation_type": null,
340
+ "shape": [
341
+ 1
342
+ ],
343
+ "continuous": true
344
+ },
345
+ "z": {
346
+ "absolute": true,
347
+ "rotation_type": null,
348
+ "shape": [
349
+ 1
350
+ ],
351
+ "continuous": true
352
+ },
353
+ "roll": {
354
+ "absolute": true,
355
+ "rotation_type": null,
356
+ "shape": [
357
+ 1
358
+ ],
359
+ "continuous": true
360
+ },
361
+ "pitch": {
362
+ "absolute": true,
363
+ "rotation_type": null,
364
+ "shape": [
365
+ 1
366
+ ],
367
+ "continuous": true
368
+ },
369
+ "yaw": {
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
+ "extra": {
386
+ "absolute": true,
387
+ "rotation_type": null,
388
+ "shape": [
389
+ 1
390
+ ],
391
+ "continuous": true
392
+ }
393
+ },
394
+ "action": {
395
+ "x": {
396
+ "absolute": true,
397
+ "rotation_type": null,
398
+ "shape": [
399
+ 1
400
+ ],
401
+ "continuous": true
402
+ },
403
+ "y": {
404
+ "absolute": true,
405
+ "rotation_type": null,
406
+ "shape": [
407
+ 1
408
+ ],
409
+ "continuous": true
410
+ },
411
+ "z": {
412
+ "absolute": true,
413
+ "rotation_type": null,
414
+ "shape": [
415
+ 1
416
+ ],
417
+ "continuous": true
418
+ },
419
+ "roll": {
420
+ "absolute": true,
421
+ "rotation_type": null,
422
+ "shape": [
423
+ 1
424
+ ],
425
+ "continuous": true
426
+ },
427
+ "pitch": {
428
+ "absolute": true,
429
+ "rotation_type": null,
430
+ "shape": [
431
+ 1
432
+ ],
433
+ "continuous": true
434
+ },
435
+ "yaw": {
436
+ "absolute": true,
437
+ "rotation_type": null,
438
+ "shape": [
439
+ 1
440
+ ],
441
+ "continuous": true
442
+ },
443
+ "gripper": {
444
+ "absolute": true,
445
+ "rotation_type": null,
446
+ "shape": [
447
+ 1
448
+ ],
449
+ "continuous": true
450
+ }
451
+ }
452
+ },
453
+ "embodiment_tag": "new_embodiment"
454
+ }
455
+ }
model-00001-of-00002.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ee04adae923c690c932aa0f29adca39f19b991301409215932a2d27e771435b0
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:f4eab1ed6f6899514a492acfdeeebff4a2091154773d45458a5241cec373ad1e
3
+ size 2586705312
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:d730dc843badc927d3d864d775e6f1c68910456d183e0e3ee2f77b1e89f04fa2
3
+ size 8550720062
rng_state.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5a3cc906ec4a33e3520869943b72d9489cbcd345713ab99ab985b6293f5db344
3
+ size 14244
scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:edbc3981e072f4299031928e30087defd79eb3c63120a25505e116ee1f43c897
3
+ size 1064
trainer_state.json ADDED
The diff for this file is too large to render. See raw diff