Hugging Face's logo

RLinf
/
RLinf-Gr00t-SFT-Goal

Safetensors
gr00t_n1_5
Model card Files
xet
 Community
lixiang-95 commited on
Commit
71412ec
·
verified ·
1 Parent(s): 8b2f88c

Initial model upload

Browse files
Files changed (7) hide show
  1. config.json +64 -0
  2. experiment_cfg/metadata.json +433 -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. scheduler.pt +3 -0
  7. trainer_state.json +118 -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 +1,433 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "libero_franka": {
3
+ "statistics": {
4
+ "state": {
5
+ "x": {
6
+ "max": [
7
+ 0.13579000532627106
8
+ ],
9
+ "min": [
10
+ -0.46141114830970764
11
+ ],
12
+ "mean": [
13
+ -0.09923473745584488
14
+ ],
15
+ "std": [
16
+ 0.11653962731361389
17
+ ],
18
+ "q01": [
19
+ -0.42401049643754957
20
+ ],
21
+ "q99": [
22
+ 0.08990443304181095
23
+ ]
24
+ },
25
+ "y": {
26
+ "max": [
27
+ 0.33316105604171753
28
+ ],
29
+ "min": [
30
+ -0.30129560828208923
31
+ ],
32
+ "mean": [
33
+ 0.013597904704511166
34
+ ],
35
+ "std": [
36
+ 0.11478105187416077
37
+ ],
38
+ "q01": [
39
+ -0.27338370531797407
40
+ ],
41
+ "q99": [
42
+ 0.26473945528268716
43
+ ]
44
+ },
45
+ "z": {
46
+ "max": [
47
+ 1.3660105466842651
48
+ ],
49
+ "min": [
50
+ 0.9083037972450256
51
+ ],
52
+ "mean": [
53
+ 1.0694637298583984
54
+ ],
55
+ "std": [
56
+ 0.10487838834524155
57
+ ],
58
+ "q01": [
59
+ 0.911226047873497
60
+ ],
61
+ "q99": [
62
+ 1.2910678112506866
63
+ ]
64
+ },
65
+ "roll": {
66
+ "max": [
67
+ 3.473310708999634
68
+ ],
69
+ "min": [
70
+ 0.35277295112609863
71
+ ],
72
+ "mean": [
73
+ 2.82898211479187
74
+ ],
75
+ "std": [
76
+ 0.5570293664932251
77
+ ],
78
+ "q01": [
79
+ 1.3085840785503386
80
+ ],
81
+ "q99": [
82
+ 3.2425890421867365
83
+ ]
84
+ },
85
+ "pitch": {
86
+ "max": [
87
+ 2.6688623428344727
88
+ ],
89
+ "min": [
90
+ -1.4858465194702148
91
+ ],
92
+ "mean": [
93
+ 0.30799180269241333
94
+ ],
95
+ "std": [
96
+ 0.7221656441688538
97
+ ],
98
+ "q01": [
99
+ -0.691297555565834
100
+ ],
101
+ "q99": [
102
+ 2.3376442337036116
103
+ ]
104
+ },
105
+ "yaw": {
106
+ "max": [
107
+ 0.8255361318588257
108
+ ],
109
+ "min": [
110
+ -1.5227035284042358
111
+ ],
112
+ "mean": [
113
+ -0.274286687374115
114
+ ],
115
+ "std": [
116
+ 0.36479514837265015
117
+ ],
118
+ "q01": [
119
+ -1.130668159723282
120
+ ],
121
+ "q99": [
122
+ 0.4659483411908149
123
+ ]
124
+ },
125
+ "gripper": {
126
+ "max": [
127
+ 0.04233968257904053,
128
+ 0.0010111660230904818
129
+ ],
130
+ "min": [
131
+ -0.0013586411951109767,
132
+ -0.042040832340717316
133
+ ],
134
+ "mean": [
135
+ 0.028092455118894577,
136
+ -0.027339335530996323
137
+ ],
138
+ "std": [
139
+ 0.01507475133985281,
140
+ 0.014990941621363163
141
+ ],
142
+ "q01": [
143
+ 0.0016738151130266487,
144
+ -0.040336399003863335
145
+ ],
146
+ "q99": [
147
+ 0.040610933862626555,
148
+ -0.0015016929572448147
149
+ ]
150
+ }
151
+ },
152
+ "action": {
153
+ "x": {
154
+ "max": [
155
+ 0.9375
156
+ ],
157
+ "min": [
158
+ -0.9375
159
+ ],
160
+ "mean": [
161
+ 0.04721052572131157
162
+ ],
163
+ "std": [
164
+ 0.3968801498413086
165
+ ],
166
+ "q01": [
167
+ -0.8785714507102966
168
+ ],
169
+ "q99": [
170
+ 0.9375
171
+ ]
172
+ },
173
+ "y": {
174
+ "max": [
175
+ 0.9375
176
+ ],
177
+ "min": [
178
+ -0.9375
179
+ ],
180
+ "mean": [
181
+ 0.028835246339440346
182
+ ],
183
+ "std": [
184
+ 0.3473387360572815
185
+ ],
186
+ "q01": [
187
+ -0.7553571462631226
188
+ ],
189
+ "q99": [
190
+ 0.9107142686843872
191
+ ]
192
+ },
193
+ "z": {
194
+ "max": [
195
+ 0.9375
196
+ ],
197
+ "min": [
198
+ -0.9375
199
+ ],
200
+ "mean": [
201
+ -0.1485840231180191
202
+ ],
203
+ "std": [
204
+ 0.49239858984947205
205
+ ],
206
+ "q01": [
207
+ -0.9375
208
+ ],
209
+ "q99": [
210
+ 0.9375
211
+ ]
212
+ },
213
+ "roll": {
214
+ "max": [
215
+ 0.3557142913341522
216
+ ],
217
+ "min": [
218
+ -0.2582142949104309
219
+ ],
220
+ "mean": [
221
+ -0.0025010062381625175
222
+ ],
223
+ "std": [
224
+ 0.055331431329250336
225
+ ],
226
+ "q01": [
227
+ -0.1510714292526245
228
+ ],
229
+ "q99": [
230
+ 0.20357142388820648
231
+ ]
232
+ },
233
+ "pitch": {
234
+ "max": [
235
+ 0.375
236
+ ],
237
+ "min": [
238
+ -0.375
239
+ ],
240
+ "mean": [
241
+ 0.026408178731799126
242
+ ],
243
+ "std": [
244
+ 0.07844757288694382
245
+ ],
246
+ "q01": [
247
+ -0.1639285683631897
248
+ ],
249
+ "q99": [
250
+ 0.26357144117355347
251
+ ]
252
+ },
253
+ "yaw": {
254
+ "max": [
255
+ 0.375
256
+ ],
257
+ "min": [
258
+ -0.2871428430080414
259
+ ],
260
+ "mean": [
261
+ 0.027379808947443962
262
+ ],
263
+ "std": [
264
+ 0.10008802264928818
265
+ ],
266
+ "q01": [
267
+ -0.13777500048279764
268
+ ],
269
+ "q99": [
270
+ 0.375
271
+ ]
272
+ },
273
+ "gripper": {
274
+ "max": [
275
+ 1.0
276
+ ],
277
+ "min": [
278
+ 0.0
279
+ ],
280
+ "mean": [
281
+ 0.6299911737442017
282
+ ],
283
+ "std": [
284
+ 0.48270025849342346
285
+ ],
286
+ "q01": [
287
+ 0.0
288
+ ],
289
+ "q99": [
290
+ 1.0
291
+ ]
292
+ }
293
+ }
294
+ },
295
+ "modalities": {
296
+ "video": {
297
+ "image": {
298
+ "resolution": [
299
+ 256,
300
+ 256
301
+ ],
302
+ "channels": 3,
303
+ "fps": 20.0
304
+ },
305
+ "wrist_image": {
306
+ "resolution": [
307
+ 256,
308
+ 256
309
+ ],
310
+ "channels": 3,
311
+ "fps": 20.0
312
+ }
313
+ },
314
+ "state": {
315
+ "x": {
316
+ "absolute": true,
317
+ "rotation_type": null,
318
+ "shape": [
319
+ 1
320
+ ],
321
+ "continuous": true
322
+ },
323
+ "y": {
324
+ "absolute": true,
325
+ "rotation_type": null,
326
+ "shape": [
327
+ 1
328
+ ],
329
+ "continuous": true
330
+ },
331
+ "z": {
332
+ "absolute": true,
333
+ "rotation_type": null,
334
+ "shape": [
335
+ 1
336
+ ],
337
+ "continuous": true
338
+ },
339
+ "roll": {
340
+ "absolute": true,
341
+ "rotation_type": null,
342
+ "shape": [
343
+ 1
344
+ ],
345
+ "continuous": true
346
+ },
347
+ "pitch": {
348
+ "absolute": true,
349
+ "rotation_type": null,
350
+ "shape": [
351
+ 1
352
+ ],
353
+ "continuous": true
354
+ },
355
+ "yaw": {
356
+ "absolute": true,
357
+ "rotation_type": null,
358
+ "shape": [
359
+ 1
360
+ ],
361
+ "continuous": true
362
+ },
363
+ "gripper": {
364
+ "absolute": true,
365
+ "rotation_type": null,
366
+ "shape": [
367
+ 2
368
+ ],
369
+ "continuous": true
370
+ }
371
+ },
372
+ "action": {
373
+ "x": {
374
+ "absolute": true,
375
+ "rotation_type": null,
376
+ "shape": [
377
+ 1
378
+ ],
379
+ "continuous": true
380
+ },
381
+ "y": {
382
+ "absolute": true,
383
+ "rotation_type": null,
384
+ "shape": [
385
+ 1
386
+ ],
387
+ "continuous": true
388
+ },
389
+ "z": {
390
+ "absolute": true,
391
+ "rotation_type": null,
392
+ "shape": [
393
+ 1
394
+ ],
395
+ "continuous": true
396
+ },
397
+ "roll": {
398
+ "absolute": true,
399
+ "rotation_type": null,
400
+ "shape": [
401
+ 1
402
+ ],
403
+ "continuous": true
404
+ },
405
+ "pitch": {
406
+ "absolute": true,
407
+ "rotation_type": null,
408
+ "shape": [
409
+ 1
410
+ ],
411
+ "continuous": true
412
+ },
413
+ "yaw": {
414
+ "absolute": true,
415
+ "rotation_type": null,
416
+ "shape": [
417
+ 1
418
+ ],
419
+ "continuous": true
420
+ },
421
+ "gripper": {
422
+ "absolute": true,
423
+ "rotation_type": null,
424
+ "shape": [
425
+ 1
426
+ ],
427
+ "continuous": true
428
+ }
429
+ }
430
+ },
431
+ "embodiment_tag": "libero_franka"
432
+ }
433
+ }
model-00001-of-00002.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:55779e30362f6e34ba39c0ce9d2ac0919423c8047840ae160fd85122c4596b9f
3
+ size 4999367032
model-00002-of-00002.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:85ed0529a6209d2e77a6016b65e41ac871469afef53bb1e7a1d571225345e48c
3
+ size 2586705312
model.safetensors.index.json ADDED
The diff for this file is too large to render. See raw diff
 
scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:478e83c7632ed85eeb43c4adcffb356f6787ad4b114eba1103e1cde5f0e935e9
3
+ size 1064
trainer_state.json ADDED
@@ -0,0 +1,118 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_global_step": null,
3
+ "best_metric": null,
4
+ "best_model_checkpoint": null,
5
+ "epoch": 2.3529411764705883,
6
+ "eval_steps": 500,
7
+ "global_step": 120,
8
+ "is_hyper_param_search": false,
9
+ "is_local_process_zero": true,
10
+ "is_world_process_zero": true,
11
+ "log_history": [
12
+ {
13
+ "epoch": 0.19607843137254902,
14
+ "grad_norm": 2.844466209411621,
15
+ "learning_rate": 9e-05,
16
+ "loss": 0.5843,
17
+ "step": 10
18
+ },
19
+ {
20
+ "epoch": 0.39215686274509803,
21
+ "grad_norm": 1.278521180152893,
22
+ "learning_rate": 9.944739353007344e-05,
23
+ "loss": 0.2849,
24
+ "step": 20
25
+ },
26
+ {
27
+ "epoch": 0.5882352941176471,
28
+ "grad_norm": 0.7505949139595032,
29
+ "learning_rate": 9.755282581475769e-05,
30
+ "loss": 0.2106,
31
+ "step": 30
32
+ },
33
+ {
34
+ "epoch": 0.7843137254901961,
35
+ "grad_norm": 0.5767074227333069,
36
+ "learning_rate": 9.43611409721806e-05,
37
+ "loss": 0.169,
38
+ "step": 40
39
+ },
40
+ {
41
+ "epoch": 0.9803921568627451,
42
+ "grad_norm": 0.4589107632637024,
43
+ "learning_rate": 8.995939984474624e-05,
44
+ "loss": 0.1408,
45
+ "step": 50
46
+ },
47
+ {
48
+ "epoch": 1.1764705882352942,
49
+ "grad_norm": 0.5532191395759583,
50
+ "learning_rate": 8.44676704559283e-05,
51
+ "loss": 0.1278,
52
+ "step": 60
53
+ },
54
+ {
55
+ "epoch": 1.3725490196078431,
56
+ "grad_norm": 0.38863489031791687,
57
+ "learning_rate": 7.803575286758364e-05,
58
+ "loss": 0.1163,
59
+ "step": 70
60
+ },
61
+ {
62
+ "epoch": 1.5686274509803921,
63
+ "grad_norm": 0.43584027886390686,
64
+ "learning_rate": 7.083909302476453e-05,
65
+ "loss": 0.1131,
66
+ "step": 80
67
+ },
68
+ {
69
+ "epoch": 1.7647058823529411,
70
+ "grad_norm": 0.4503878951072693,
71
+ "learning_rate": 6.307399704769099e-05,
72
+ "loss": 0.1024,
73
+ "step": 90
74
+ },
75
+ {
76
+ "epoch": 1.9607843137254903,
77
+ "grad_norm": 0.4599202871322632,
78
+ "learning_rate": 5.495227651252315e-05,
79
+ "loss": 0.1015,
80
+ "step": 100
81
+ },
82
+ {
83
+ "epoch": 2.156862745098039,
84
+ "grad_norm": 0.15444983541965485,
85
+ "learning_rate": 4.669547078371504e-05,
86
+ "loss": 0.097,
87
+ "step": 110
88
+ },
89
+ {
90
+ "epoch": 2.3529411764705883,
91
+ "grad_norm": 0.2130330502986908,
92
+ "learning_rate": 3.852880399766243e-05,
93
+ "loss": 0.0958,
94
+ "step": 120
95
+ }
96
+ ],
97
+ "logging_steps": 10,
98
+ "max_steps": 200,
99
+ "num_input_tokens_seen": 0,
100
+ "num_train_epochs": 4,
101
+ "save_steps": 40,
102
+ "stateful_callbacks": {
103
+ "TrainerControl": {
104
+ "args": {
105
+ "should_epoch_stop": false,
106
+ "should_evaluate": false,
107
+ "should_log": false,
108
+ "should_save": true,
109
+ "should_training_stop": false
110
+ },
111
+ "attributes": {}
112
+ }
113
+ },
114
+ "total_flos": 0.0,
115
+ "train_batch_size": 128,
116
+ "trial_name": null,
117
+ "trial_params": null
118
+ }