Hugging Face's logo

nextbig
/
gr00t-3b-sl-5

Safetensors
gr00t_n1_5
Model card Files
xet
Community
nextbig commited on
Commit
5c34ec3
·
verified ·
1 Parent(s): 9980def

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
+ 0.0017627825727686286,
8
+ -0.009878563694655895,
9
+ 0.030438542366027832,
10
+ 0.08234463632106781,
11
+ 0.2016378939151764,
12
+ 0.13041099905967712
13
+ ],
14
+ "min": [
15
+ -0.12000884115695953,
16
+ -0.01612684316933155,
17
+ 0.024260075762867928,
18
+ -0.01539380382746458,
19
+ 0.031782444566488266,
20
+ -0.08897688239812851
21
+ ],
22
+ "mean": [
23
+ -0.01910265552887121,
24
+ -0.010410111039515967,
25
+ 0.029360902905809384,
26
+ 0.03840650130351472,
27
+ 0.13294711556213576,
28
+ -0.018013461938842318
29
+ ],
30
+ "std": [
31
+ 0.0311541949923402,
32
+ 0.001746231916101937,
33
+ 0.0017539935406906222,
34
+ 0.027976515323805115,
35
+ 0.0594243818602827,
36
+ 0.07420757231226441
37
+ ],
38
+ "q01": [
39
+ -0.0669675100838659,
40
+ -0.012851783063538121,
41
+ 0.027498543179217995,
42
+ 0.01679428606029558,
43
+ 0.03301159795065989,
44
+ -0.0700492343361025
45
+ ],
46
+ "q99": [
47
+ -0.009022958832457574,
48
+ -0.009878563694655895,
49
+ 0.03043854236602783,
50
+ 0.061614787198399024,
51
+ 0.1952524415383192,
52
+ 0.0838942406851405
53
+ ]
54
+ },
55
+ "left_gripper": {
56
+ "max": [
57
+ 0.00048461538972333074
58
+ ],
59
+ "min": [
60
+ 0.0
61
+ ],
62
+ "mean": [
63
+ 4.138201219854646e-05
64
+ ],
65
+ "std": [
66
+ 0.0001354398809075508
67
+ ],
68
+ "q01": [
69
+ 0.0
70
+ ],
71
+ "q99": [
72
+ 0.00023060233925573722
73
+ ]
74
+ },
75
+ "right_arm": {
76
+ "max": [
77
+ 1.209041953086853,
78
+ 2.5278475284576416,
79
+ 0.01567305624485016,
80
+ 0.699929416179657,
81
+ 1.2358676195144653,
82
+ 0.8392939567565918
83
+ ],
84
+ "min": [
85
+ -0.3023608326911926,
86
+ 0.002687806962057948,
87
+ -2.456254243850708,
88
+ -0.8461880087852478,
89
+ -0.441201776266098,
90
+ -1.7722247838974
91
+ ],
92
+ "mean": [
93
+ 0.47053953643579044,
94
+ 1.417530677092354,
95
+ -1.1486868317152683,
96
+ 0.0013949999148885383,
97
+ 0.6223417124783771,
98
+ -0.07632650141521694
99
+ ],
100
+ "std": [
101
+ 0.2868581278952159,
102
+ 0.8087564008773853,
103
+ 0.7159159222223449,
104
+ 0.12086897986061979,
105
+ 0.41072920043937744,
106
+ 0.24381936792735875
107
+ ],
108
+ "q01": [
109
+ 0.06352259762327005,
110
+ 0.24115008948440325,
111
+ -2.291646721173831,
112
+ -0.3013912399316343,
113
+ 0.01190044061459547,
114
+ -0.8511293082780937
115
+ ],
116
+ "q99": [
117
+ 0.9199439867286312,
118
+ 2.43179472409651,
119
+ -0.16801454626693743,
120
+ 0.29150672081758855,
121
+ 1.2137429386497405,
122
+ 0.49233419072282203
123
+ ]
124
+ },
125
+ "right_gripper": {
126
+ "max": [
127
+ 0.08480769395828247
128
+ ],
129
+ "min": [
130
+ -0.0005769230774603784
131
+ ],
132
+ "mean": [
133
+ 0.024830522569081026
134
+ ],
135
+ "std": [
136
+ 0.024716617675411744
137
+ ],
138
+ "q01": [
139
+ -0.0003405798719650984
140
+ ],
141
+ "q99": [
142
+ 0.0748798751459126
143
+ ]
144
+ }
145
+ },
146
+ "action": {
147
+ "left_arm": {
148
+ "max": [
149
+ 0.0017627825727686286,
150
+ -0.009878563694655895,
151
+ 0.030438542366027832,
152
+ 0.08234463632106781,
153
+ 0.2016378939151764,
154
+ 0.13041099905967712
155
+ ],
156
+ "min": [
157
+ -0.12000884115695953,
158
+ -0.01612684316933155,
159
+ 0.024260075762867928,
160
+ -0.01539380382746458,
161
+ 0.031782444566488266,
162
+ -0.08897688239812851
163
+ ],
164
+ "mean": [
165
+ -0.01910265552887121,
166
+ -0.010410111039515967,
167
+ 0.029360902905809384,
168
+ 0.03840656131764875,
169
+ 0.13294742407695984,
170
+ -0.018013431673147548
171
+ ],
172
+ "std": [
173
+ 0.0311541949923402,
174
+ 0.001746231916101937,
175
+ 0.0017539935406906222,
176
+ 0.02797651276273133,
177
+ 0.059424378672045995,
178
+ 0.07420755979773257
179
+ ],
180
+ "q01": [
181
+ -0.0669675100838659,
182
+ -0.012851783063538121,
183
+ 0.027498543179217995,
184
+ 0.01679428606029558,
185
+ 0.03301159795065989,
186
+ -0.0700492343361025
187
+ ],
188
+ "q99": [
189
+ -0.009022958832457574,
190
+ -0.009878563694655895,
191
+ 0.03043854236602783,
192
+ 0.061614787198399024,
193
+ 0.1952524415383192,
194
+ 0.0838942406851405
195
+ ]
196
+ },
197
+ "left_gripper": {
198
+ "max": [
199
+ 0.00048461538972333074
200
+ ],
201
+ "min": [
202
+ 0.0
203
+ ],
204
+ "mean": [
205
+ 4.138201219854646e-05
206
+ ],
207
+ "std": [
208
+ 0.0001354398809075508
209
+ ],
210
+ "q01": [
211
+ 0.0
212
+ ],
213
+ "q99": [
214
+ 0.00023060233925573722
215
+ ]
216
+ },
217
+ "right_arm": {
218
+ "max": [
219
+ 1.2223937511444092,
220
+ 2.5278475284576416,
221
+ 0.01567305624485016,
222
+ 0.699929416179657,
223
+ 1.2358676195144653,
224
+ 0.8392939567565918
225
+ ],
226
+ "min": [
227
+ -0.3023608326911926,
228
+ 0.002687806962057948,
229
+ -2.456254243850708,
230
+ -0.8461880087852478,
231
+ -0.441201776266098,
232
+ -1.7722247838974
233
+ ],
234
+ "mean": [
235
+ 0.4698464159285657,
236
+ 1.4209469780626074,
237
+ -1.1517979039144939,
238
+ 0.0014538363791808207,
239
+ 0.622843688467666,
240
+ -0.07656144985117133
241
+ ],
242
+ "std": [
243
+ 0.2865071698128086,
244
+ 0.8084023491044288,
245
+ 0.7157899431591536,
246
+ 0.1210333278809214,
247
+ 0.4109320155155253,
248
+ 0.24402982330471445
249
+ ],
250
+ "q01": [
251
+ 0.06352259762327005,
252
+ 0.2435203664263759,
253
+ -2.291646721173831,
254
+ -0.30353229287542904,
255
+ 0.010397212330326991,
256
+ -0.8511293082780937
257
+ ],
258
+ "q99": [
259
+ 0.9201898113784049,
260
+ 2.43179472409651,
261
+ -0.16816070059683633,
262
+ 0.29150672081758855,
263
+ 1.2138143671154493,
264
+ 0.49234914350519243
265
+ ]
266
+ },
267
+ "right_gripper": {
268
+ "max": [
269
+ 0.08480769395828247
270
+ ],
271
+ "min": [
272
+ -0.0005769230774603784
273
+ ],
274
+ "mean": [
275
+ 0.02488749917243111
276
+ ],
277
+ "std": [
278
+ 0.024706559580291047
279
+ ],
280
+ "q01": [
281
+ -0.0003405798719650984
282
+ ],
283
+ "q99": [
284
+ 0.07486889455089227
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:ce973a6708e0806ad723a14251f7b914a3976031699b2fe3b436fbc0576bbdf3
3
+ size 4999367032
model-00002-of-00002.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2889768cdef115d40cf4b89f7d2b345d288878515159fc029c076385a3562d8d
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:bf8944b70b96476305d8e3f6f975f1a3fa552c7a065e90ada7325273d3c78a79
3
+ size 5713