Hugging Face's logo

nextbig
/
gr00t-3b-sl

Safetensors
gr00t_n1_5
Model card Files
xet
 Community
nextbig commited on
Commit
3e93247
·
verified ·
1 Parent(s): ce90d25

Upload folder using huggingface_hub

Browse files
Files changed (7) hide show
  1. config.json +64 -0
  2. experiment_cfg/metadata.json +387 -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
  7. training_args.bin +3 -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": "4.52.1"
64
+ }
experiment_cfg/metadata.json ADDED
@@ -0,0 +1,387 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "new_embodiment": {
3
+ "statistics": {
4
+ "state": {
5
+ "left_arm": {
6
+ "max": [
7
+ 0.23685863614082336,
8
+ 2.63502836227417,
9
+ 0.03330088034272194,
10
+ 1.4575419425964355,
11
+ 0.8281587362289429,
12
+ 1.9787147045135498
13
+ ],
14
+ "min": [
15
+ -0.797720193862915,
16
+ -0.012758356519043446,
17
+ -2.2290821075439453,
18
+ -1.7361313104629517,
19
+ -0.532290518283844,
20
+ -1.4025291204452515
21
+ ],
22
+ "mean": [
23
+ -0.09509609639644623,
24
+ 0.3003239929676056,
25
+ -0.17810438573360443,
26
+ -0.001026484533213079,
27
+ 0.07963123917579651,
28
+ 0.00482672406360507
29
+ ],
30
+ "std": [
31
+ 0.11718923598527908,
32
+ 0.6811562180519104,
33
+ 0.49466368556022644,
34
+ 0.22098089754581451,
35
+ 0.08493965864181519,
36
+ 0.23760953545570374
37
+ ],
38
+ "q01": [
39
+ -0.6195136725902557,
40
+ -0.012758356519043446,
41
+ -1.828329119682312,
42
+ -0.8194373416900635,
43
+ -0.26055216789245605,
44
+ -0.6677455306053162
45
+ ],
46
+ "q99": [
47
+ 0.0558854416012764,
48
+ 2.302248034477236,
49
+ 0.03330088034272194,
50
+ 0.8961595630645758,
51
+ 0.3783116203546525,
52
+ 0.9678304481506353
53
+ ]
54
+ },
55
+ "left_gripper": {
56
+ "max": [
57
+ 0.054955385625362396
58
+ ],
59
+ "min": [
60
+ -0.0037799999117851257
61
+ ],
62
+ "mean": [
63
+ 0.0005705303628928959
64
+ ],
65
+ "std": [
66
+ 0.00873688142746687
67
+ ],
68
+ "q01": [
69
+ -0.0037799999117851257
70
+ ],
71
+ "q99": [
72
+ 0.04826769232749939
73
+ ]
74
+ },
75
+ "right_arm": {
76
+ "max": [
77
+ 0.8744099736213684,
78
+ 2.716465473175049,
79
+ 33088.39453125,
80
+ 1.7382432222366333,
81
+ 1.2609829902648926,
82
+ 0.7371398210525513
83
+ ],
84
+ "min": [
85
+ -0.30108675360679626,
86
+ 0.0,
87
+ -2.7580041885375977,
88
+ -1.2717341184616089,
89
+ -0.7101046442985535,
90
+ -2.1718528270721436
91
+ ],
92
+ "mean": [
93
+ 0.13213729858398438,
94
+ 1.0804740190505981,
95
+ -0.37861087918281555,
96
+ 0.08227251470088959,
97
+ 0.3457416296005249,
98
+ -0.3394346833229065
99
+ ],
100
+ "std": [
101
+ 0.23602594435214996,
102
+ 0.9582993984222412,
103
+ 120.51954650878906,
104
+ 0.26776427030563354,
105
+ 0.4478372037410736,
106
+ 0.4248640537261963
107
+ ],
108
+ "q01": [
109
+ -0.1656310448050499,
110
+ 0.0,
111
+ -2.3988981008529664,
112
+ -0.7335520935058594,
113
+ -0.20960182219743728,
114
+ -1.8287385702133179
115
+ ],
116
+ "q99": [
117
+ 0.7345469546318054,
118
+ 2.519456400871277,
119
+ 0.009965830482542515,
120
+ 1.0597094678878796,
121
+ 1.2147316932678223,
122
+ 0.37196457386016846
123
+ ]
124
+ },
125
+ "right_gripper": {
126
+ "max": [
127
+ 0.08273077011108398
128
+ ],
129
+ "min": [
130
+ -0.0005769230774603784
131
+ ],
132
+ "mean": [
133
+ 0.017376570031046867
134
+ ],
135
+ "std": [
136
+ 0.024361947551369667
137
+ ],
138
+ "q01": [
139
+ -0.0004615384677890688
140
+ ],
141
+ "q99": [
142
+ 0.07165384292602539
143
+ ]
144
+ }
145
+ },
146
+ "action": {
147
+ "left_arm": {
148
+ "max": [
149
+ 0.23685863614082336,
150
+ 2.63502836227417,
151
+ 0.03330088034272194,
152
+ 1.4575419425964355,
153
+ 0.8281587362289429,
154
+ 1.9787147045135498
155
+ ],
156
+ "min": [
157
+ -0.797720193862915,
158
+ -0.012758356519043446,
159
+ -2.2290821075439453,
160
+ -1.7361313104629517,
161
+ -0.532290518283844,
162
+ -1.4025291204452515
163
+ ],
164
+ "mean": [
165
+ -0.09509586542844772,
166
+ 0.30032211542129517,
167
+ -0.17810463905334473,
168
+ -0.001011453103274107,
169
+ 0.07963286340236664,
170
+ 0.004815105814486742
171
+ ],
172
+ "std": [
173
+ 0.11718949675559998,
174
+ 0.6811572313308716,
175
+ 0.4946635067462921,
176
+ 0.22098401188850403,
177
+ 0.08493973314762115,
178
+ 0.2376156598329544
179
+ ],
180
+ "q01": [
181
+ -0.6195136725902557,
182
+ -0.012758356519043446,
183
+ -1.828329119682312,
184
+ -0.8194373416900635,
185
+ -0.26055216789245605,
186
+ -0.6677455306053162
187
+ ],
188
+ "q99": [
189
+ 0.0558854416012764,
190
+ 2.302248034477236,
191
+ 0.03330088034272194,
192
+ 0.8961595630645758,
193
+ 0.3783116203546525,
194
+ 0.9678304481506353
195
+ ]
196
+ },
197
+ "left_gripper": {
198
+ "max": [
199
+ 0.054955385625362396
200
+ ],
201
+ "min": [
202
+ -0.0037799999117851257
203
+ ],
204
+ "mean": [
205
+ 0.0005705304211005569
206
+ ],
207
+ "std": [
208
+ 0.00873688142746687
209
+ ],
210
+ "q01": [
211
+ -0.0037799999117851257
212
+ ],
213
+ "q99": [
214
+ 0.04826769232749939
215
+ ]
216
+ },
217
+ "right_arm": {
218
+ "max": [
219
+ 0.8744099736213684,
220
+ 2.716465473175049,
221
+ 33088.39453125,
222
+ 1.7382432222366333,
223
+ 1.2609829902648926,
224
+ 0.7371398210525513
225
+ ],
226
+ "min": [
227
+ -0.30108675360679626,
228
+ 0.0,
229
+ -2.7580041885375977,
230
+ -1.2717341184616089,
231
+ -0.7101046442985535,
232
+ -2.1718528270721436
233
+ ],
234
+ "mean": [
235
+ 0.13213366270065308,
236
+ 1.0804693698883057,
237
+ -0.378610223531723,
238
+ 0.08227236568927765,
239
+ 0.34574177861213684,
240
+ -0.33944571018218994
241
+ ],
242
+ "std": [
243
+ 0.23602980375289917,
244
+ 0.9583073854446411,
245
+ 120.51954650878906,
246
+ 0.2677631676197052,
247
+ 0.4478361904621124,
248
+ 0.4248557686805725
249
+ ],
250
+ "q01": [
251
+ -0.1658991301059723,
252
+ 0.0,
253
+ -2.3988981008529664,
254
+ -0.7335520935058594,
255
+ -0.20960182219743728,
256
+ -1.8287385702133179
257
+ ],
258
+ "q99": [
259
+ 0.7345469546318054,
260
+ 2.519456400871277,
261
+ 0.009965830482542515,
262
+ 1.0597094678878796,
263
+ 1.2147316932678223,
264
+ 0.37196457386016846
265
+ ]
266
+ },
267
+ "right_gripper": {
268
+ "max": [
269
+ 0.08273077011108398
270
+ ],
271
+ "min": [
272
+ -0.0005769230774603784
273
+ ],
274
+ "mean": [
275
+ 0.01737651415169239
276
+ ],
277
+ "std": [
278
+ 0.02436199225485325
279
+ ],
280
+ "q01": [
281
+ -0.0004615384677890688
282
+ ],
283
+ "q99": [
284
+ 0.07165384292602539
285
+ ]
286
+ }
287
+ }
288
+ },
289
+ "modalities": {
290
+ "video": {
291
+ "cam_high": {
292
+ "resolution": [
293
+ 640,
294
+ 480
295
+ ],
296
+ "channels": 3,
297
+ "fps": 30.0
298
+ },
299
+ "cam_left_wrist": {
300
+ "resolution": [
301
+ 640,
302
+ 480
303
+ ],
304
+ "channels": 3,
305
+ "fps": 30.0
306
+ },
307
+ "cam_right_wrist": {
308
+ "resolution": [
309
+ 640,
310
+ 480
311
+ ],
312
+ "channels": 3,
313
+ "fps": 30.0
314
+ }
315
+ },
316
+ "state": {
317
+ "left_arm": {
318
+ "absolute": true,
319
+ "rotation_type": null,
320
+ "shape": [
321
+ 6
322
+ ],
323
+ "continuous": true
324
+ },
325
+ "left_gripper": {
326
+ "absolute": true,
327
+ "rotation_type": null,
328
+ "shape": [
329
+ 1
330
+ ],
331
+ "continuous": true
332
+ },
333
+ "right_arm": {
334
+ "absolute": true,
335
+ "rotation_type": null,
336
+ "shape": [
337
+ 6
338
+ ],
339
+ "continuous": true
340
+ },
341
+ "right_gripper": {
342
+ "absolute": true,
343
+ "rotation_type": null,
344
+ "shape": [
345
+ 1
346
+ ],
347
+ "continuous": true
348
+ }
349
+ },
350
+ "action": {
351
+ "left_arm": {
352
+ "absolute": true,
353
+ "rotation_type": null,
354
+ "shape": [
355
+ 6
356
+ ],
357
+ "continuous": true
358
+ },
359
+ "left_gripper": {
360
+ "absolute": true,
361
+ "rotation_type": null,
362
+ "shape": [
363
+ 1
364
+ ],
365
+ "continuous": true
366
+ },
367
+ "right_arm": {
368
+ "absolute": true,
369
+ "rotation_type": null,
370
+ "shape": [
371
+ 6
372
+ ],
373
+ "continuous": true
374
+ },
375
+ "right_gripper": {
376
+ "absolute": true,
377
+ "rotation_type": null,
378
+ "shape": [
379
+ 1
380
+ ],
381
+ "continuous": true
382
+ }
383
+ }
384
+ },
385
+ "embodiment_tag": "new_embodiment"
386
+ }
387
+ }
model-00001-of-00002.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6099cfa3cec340ac30ce75da741c4217ba2b10fe5f4bf9e311a53e84b68d4b19
3
+ size 4999367032
model-00002-of-00002.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2d742567b7d02be0045789ee39eb8699e802481ba80b588c6178b82083531baf
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
 
training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f8b682d7afcd7cf8578efb61b306051cfc2d8b7a1732d63ee799eb905a576d72
3
+ size 5713