Hugging Face's logo

RLinf
/
RLinf-Gr00t-RL-Spatial-Step400

Safetensors
gr00t_n1_5
Model card Files
xet
 Community
lixiang-95 commited on
Commit
a0a9a6b
·
verified ·
1 Parent(s): 57909f5

Initial model upload

Browse files
Files changed (5) hide show
  1. config.json +64 -0
  2. experiment_cfg/metadata.json +433 -0
  3. model.safetensors +3 -0
  4. scheduler.pt +3 -0
  5. trainer_state.json +90 -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
+ 0.1759040206670761
8
+ ],
9
+ "min": [
10
+ -0.3095473051071167
11
+ ],
12
+ "mean": [
13
+ -0.024462558329105377
14
+ ],
15
+ "std": [
16
+ 0.1101478561758995
17
+ ],
18
+ "q01": [
19
+ -0.2727657300233841
20
+ ],
21
+ "q99": [
22
+ 0.13529365032911292
23
+ ]
24
+ },
25
+ "y": {
26
+ "max": [
27
+ 0.3904820382595062
28
+ ],
29
+ "min": [
30
+ -0.29250794649124146
31
+ ],
32
+ "mean": [
33
+ 0.106529600918293
34
+ ],
35
+ "std": [
36
+ 0.13784688711166382
37
+ ],
38
+ "q01": [
39
+ -0.23721413239836692
40
+ ],
41
+ "q99": [
42
+ 0.3629165390133857
43
+ ]
44
+ },
45
+ "z": {
46
+ "max": [
47
+ 1.3290715217590332
48
+ ],
49
+ "min": [
50
+ 0.9095591306686401
51
+ ],
52
+ "mean": [
53
+ 1.0580483675003052
54
+ ],
55
+ "std": [
56
+ 0.1044282391667366
57
+ ],
58
+ "q01": [
59
+ 0.9160063165426254
60
+ ],
61
+ "q99": [
62
+ 1.2862326657772063
63
+ ]
64
+ },
65
+ "roll": {
66
+ "max": [
67
+ 3.4566118717193604
68
+ ],
69
+ "min": [
70
+ 2.497488260269165
71
+ ],
72
+ "mean": [
73
+ 3.0628468990325928
74
+ ],
75
+ "std": [
76
+ 0.10451053828001022
77
+ ],
78
+ "q01": [
79
+ 2.77949666261673
80
+ ],
81
+ "q99": [
82
+ 3.2829698753356933
83
+ ]
84
+ },
85
+ "pitch": {
86
+ "max": [
87
+ 1.2268599271774292
88
+ ],
89
+ "min": [
90
+ -1.8006486892700195
91
+ ],
92
+ "mean": [
93
+ -0.10464039444923401
94
+ ],
95
+ "std": [
96
+ 0.4112098217010498
97
+ ],
98
+ "q01": [
99
+ -1.3187511622905732
100
+ ],
101
+ "q99": [
102
+ 0.9332760351896285
103
+ ]
104
+ },
105
+ "yaw": {
106
+ "max": [
107
+ 1.0429412126541138
108
+ ],
109
+ "min": [
110
+ -0.7207611203193665
111
+ ],
112
+ "mean": [
113
+ 0.08307311683893204
114
+ ],
115
+ "std": [
116
+ 0.2176690548658371
117
+ ],
118
+ "q01": [
119
+ -0.41989982962608335
120
+ ],
121
+ "q99": [
122
+ 0.6325724506378171
123
+ ]
124
+ },
125
+ "gripper": {
126
+ "max": [
127
+ 0.041053611785173416,
128
+ 0.000775813648942858
129
+ ],
130
+ "min": [
131
+ -0.0004703797458205372,
132
+ -0.041536275297403336
133
+ ],
134
+ "mean": [
135
+ 0.01995457336306572,
136
+ -0.020162804052233696
137
+ ],
138
+ "std": [
139
+ 0.017260896041989326,
140
+ 0.0171116404235363
141
+ ],
142
+ "q01": [
143
+ 0.001503719249740243,
144
+ -0.03989770736545324
145
+ ],
146
+ "q99": [
147
+ 0.039933966137468815,
148
+ -0.001671919699292631
149
+ ]
150
+ }
151
+ },
152
+ "action": {
153
+ "x": {
154
+ "max": [
155
+ 0.9375
156
+ ],
157
+ "min": [
158
+ -0.9375
159
+ ],
160
+ "mean": [
161
+ 0.15312479436397552
162
+ ],
163
+ "std": [
164
+ 0.41272708773612976
165
+ ],
166
+ "q01": [
167
+ -0.7454732114076613
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
+ 0.13707277178764343
182
+ ],
183
+ "std": [
184
+ 0.34724321961402893
185
+ ],
186
+ "q01": [
187
+ -0.6616071462631226
188
+ ],
189
+ "q99": [
190
+ 0.8758928775787354
191
+ ]
192
+ },
193
+ "z": {
194
+ "max": [
195
+ 0.9375
196
+ ],
197
+ "min": [
198
+ -0.9375
199
+ ],
200
+ "mean": [
201
+ -0.15526802837848663
202
+ ],
203
+ "std": [
204
+ 0.50869220495224
205
+ ],
206
+ "q01": [
207
+ -0.9375
208
+ ],
209
+ "q99": [
210
+ 0.9321428537368774
211
+ ]
212
+ },
213
+ "roll": {
214
+ "max": [
215
+ 0.1971428543329239
216
+ ],
217
+ "min": [
218
+ -0.1875
219
+ ],
220
+ "mean": [
221
+ -0.005176450591534376
222
+ ],
223
+ "std": [
224
+ 0.037266165018081665
225
+ ],
226
+ "q01": [
227
+ -0.1071428582072258
228
+ ],
229
+ "q99": [
230
+ 0.1039285734295845
231
+ ]
232
+ },
233
+ "pitch": {
234
+ "max": [
235
+ 0.33642858266830444
236
+ ],
237
+ "min": [
238
+ -0.3675000071525574
239
+ ],
240
+ "mean": [
241
+ -0.01120874285697937
242
+ ],
243
+ "std": [
244
+ 0.07244449853897095
245
+ ],
246
+ "q01": [
247
+ -0.20678570866584778
248
+ ],
249
+ "q99": [
250
+ 0.17678570747375488
251
+ ]
252
+ },
253
+ "yaw": {
254
+ "max": [
255
+ 0.375
256
+ ],
257
+ "min": [
258
+ -0.36000001430511475
259
+ ],
260
+ "mean": [
261
+ -0.020194264128804207
262
+ ],
263
+ "std": [
264
+ 0.05762382969260216
265
+ ],
266
+ "q01": [
267
+ -0.1842857152223587
268
+ ],
269
+ "q99": [
270
+ 0.14571428298950195
271
+ ]
272
+ },
273
+ "gripper": {
274
+ "max": [
275
+ 1.0
276
+ ],
277
+ "min": [
278
+ 0.0
279
+ ],
280
+ "mean": [
281
+ 0.4578818082809448
282
+ ],
283
+ "std": [
284
+ 0.49827873706817627
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.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:83984aabcad10167d417285c1718c1d5a3807cf2f8b991c6815dcd781756aec6
3
+ size 5449839514
scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:da226c44306b9d44cb710a3e43a9625a10e7d1c86f899d409c15432084a7060b
3
+ size 1064
trainer_state.json ADDED
@@ -0,0 +1,90 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_global_step": null,
3
+ "best_metric": null,
4
+ "best_model_checkpoint": null,
5
+ "epoch": 1.5384615384615383,
6
+ "eval_steps": 500,
7
+ "global_step": 80,
8
+ "is_hyper_param_search": false,
9
+ "is_local_process_zero": true,
10
+ "is_world_process_zero": true,
11
+ "log_history": [
12
+ {
13
+ "epoch": 0.19230769230769232,
14
+ "grad_norm": 1.0270750522613525,
15
+ "learning_rate": 9.956320346634876e-05,
16
+ "loss": 0.5994,
17
+ "step": 10
18
+ },
19
+ {
20
+ "epoch": 0.38461538461538464,
21
+ "grad_norm": 1.0943247079849243,
22
+ "learning_rate": 9.473646649103818e-05,
23
+ "loss": 0.306,
24
+ "step": 20
25
+ },
26
+ {
27
+ "epoch": 0.5769230769230769,
28
+ "grad_norm": 0.8214321732521057,
29
+ "learning_rate": 8.506183921362443e-05,
30
+ "loss": 0.2207,
31
+ "step": 30
32
+ },
33
+ {
34
+ "epoch": 0.7692307692307693,
35
+ "grad_norm": 0.7468852400779724,
36
+ "learning_rate": 7.158771761692464e-05,
37
+ "loss": 0.1827,
38
+ "step": 40
39
+ },
40
+ {
41
+ "epoch": 0.9615384615384616,
42
+ "grad_norm": 0.5866342782974243,
43
+ "learning_rate": 5.577423184847932e-05,
44
+ "loss": 0.1552,
45
+ "step": 50
46
+ },
47
+ {
48
+ "epoch": 1.1538461538461537,
49
+ "grad_norm": 0.3692125678062439,
50
+ "learning_rate": 3.933501846281267e-05,
51
+ "loss": 0.1458,
52
+ "step": 60
53
+ },
54
+ {
55
+ "epoch": 1.3461538461538463,
56
+ "grad_norm": 0.33734169602394104,
57
+ "learning_rate": 2.405152131093926e-05,
58
+ "loss": 0.1356,
59
+ "step": 70
60
+ },
61
+ {
62
+ "epoch": 1.5384615384615383,
63
+ "grad_norm": 0.29717421531677246,
64
+ "learning_rate": 1.157994445715706e-05,
65
+ "loss": 0.1356,
66
+ "step": 80
67
+ }
68
+ ],
69
+ "logging_steps": 10,
70
+ "max_steps": 100,
71
+ "num_input_tokens_seen": 0,
72
+ "num_train_epochs": 2,
73
+ "save_steps": 20,
74
+ "stateful_callbacks": {
75
+ "TrainerControl": {
76
+ "args": {
77
+ "should_epoch_stop": false,
78
+ "should_evaluate": false,
79
+ "should_log": false,
80
+ "should_save": true,
81
+ "should_training_stop": false
82
+ },
83
+ "attributes": {}
84
+ }
85
+ },
86
+ "total_flos": 0.0,
87
+ "train_batch_size": 128,
88
+ "trial_name": null,
89
+ "trial_params": null
90
+ }