Hugging Face's logo

Ronaldo-GOAT
/
actaug-gr00t-train-eval

Safetensors
Model card Files
xet
 Community
Ronaldo-GOAT commited on
Commit
ea27c19
·
verified ·
1 Parent(s): 38dc077

Add base60k checkpoint (== mlnha/gr00t-n15-robocasa-base60k)

Browse files
Files changed (6) hide show
  1. base60k/config.json +64 -0
  2. base60k/experiment_cfg/metadata.json +431 -0
  3. base60k/model-00001-of-00002.safetensors +3 -0
  4. base60k/model-00002-of-00002.safetensors +3 -0
  5. base60k/model.safetensors.index.json +0 -0
  6. base60k/trainer_state.json +0 -0
base60k/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
+ }
base60k/experiment_cfg/metadata.json ADDED
@@ -0,0 +1,431 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "new_embodiment": {
3
+ "statistics": {
4
+ "state": {
5
+ "base_position": {
6
+ "max": [
7
+ 5.634693145751953,
8
+ -0.4005681872367859,
9
+ 0.7043487429618835
10
+ ],
11
+ "min": [
12
+ 0.40834179520606995,
13
+ -4.441381454467773,
14
+ 0.6998239755630493
15
+ ],
16
+ "mean": [
17
+ 2.743187665939331,
18
+ -1.430670976638794,
19
+ 0.7007994055747986
20
+ ],
21
+ "std": [
22
+ 1.60910964012146,
23
+ 0.9574378132820129,
24
+ 0.0007126645650714636
25
+ ],
26
+ "q01": [
27
+ 0.5017430782318115,
28
+ -4.307175159454346,
29
+ 0.6999944448471069
30
+ ],
31
+ "q99": [
32
+ 5.6084208488464355,
33
+ -0.6116418838500977,
34
+ 0.7026175260543823
35
+ ]
36
+ },
37
+ "base_rotation": {
38
+ "max": [
39
+ 0.0,
40
+ 0.0,
41
+ 1.0,
42
+ 1.0
43
+ ],
44
+ "min": [
45
+ 0.0,
46
+ 0.0,
47
+ -1.0,
48
+ 0.0
49
+ ],
50
+ "mean": [
51
+ 0.0,
52
+ 0.0,
53
+ 0.2988399863243103,
54
+ 0.641080915927887
55
+ ],
56
+ "std": [
57
+ 0.0,
58
+ 0.0,
59
+ 0.6171786785125732,
60
+ 0.3455561399459839
61
+ ],
62
+ "q01": [
63
+ 0.0,
64
+ 0.0,
65
+ -1.0,
66
+ 4.676552691762481e-07
67
+ ],
68
+ "q99": [
69
+ 0.0,
70
+ 0.0,
71
+ 1.0,
72
+ 1.0
73
+ ]
74
+ },
75
+ "end_effector_position_relative": {
76
+ "max": [
77
+ 0.6971381902694702,
78
+ 0.7329497933387756,
79
+ 1.0843316316604614
80
+ ],
81
+ "min": [
82
+ -0.08105481415987015,
83
+ -0.6334956288337708,
84
+ 0.2257986068725586
85
+ ],
86
+ "mean": [
87
+ 0.3166314363479614,
88
+ -0.003687890712171793,
89
+ 0.5773549675941467
90
+ ],
91
+ "std": [
92
+ 0.12479864060878754,
93
+ 0.12240128964185715,
94
+ 0.19792696833610535
95
+ ],
96
+ "q01": [
97
+ 0.02998272143304348,
98
+ -0.31315547227859497,
99
+ 0.23194725811481476
100
+ ],
101
+ "q99": [
102
+ 0.5974694490432739,
103
+ 0.30662161111831665,
104
+ 0.8845611810684204
105
+ ]
106
+ },
107
+ "end_effector_rotation_relative": {
108
+ "max": [
109
+ 0.9999685883522034,
110
+ 0.9405615925788879,
111
+ 0.8613225817680359,
112
+ 0.6224425435066223
113
+ ],
114
+ "min": [
115
+ -0.9998859167098999,
116
+ -0.7644739747047424,
117
+ -0.9218037128448486,
118
+ 1.8247247624003649e-07
119
+ ],
120
+ "mean": [
121
+ -0.2097865641117096,
122
+ 0.035907942801713943,
123
+ -0.055897388607263565,
124
+ 0.08059503138065338
125
+ ],
126
+ "std": [
127
+ 0.8574697375297546,
128
+ 0.22963927686214447,
129
+ 0.3847891092300415,
130
+ 0.09500124305486679
131
+ ],
132
+ "q01": [
133
+ -0.9960652589797974,
134
+ -0.5019329190254211,
135
+ -0.6929202675819397,
136
+ 0.0007045630482025445
137
+ ],
138
+ "q99": [
139
+ 0.9955692291259766,
140
+ 0.6864141225814819,
141
+ 0.6946918964385986,
142
+ 0.4153072237968445
143
+ ]
144
+ },
145
+ "gripper_qpos": {
146
+ "max": [
147
+ 0.041749171912670135,
148
+ 0.0028377638664096594
149
+ ],
150
+ "min": [
151
+ -0.001856057089753449,
152
+ -0.04221527278423309
153
+ ],
154
+ "mean": [
155
+ 0.031458113342523575,
156
+ -0.03141428530216217
157
+ ],
158
+ "std": [
159
+ 0.010286576114594936,
160
+ 0.010318808257579803
161
+ ],
162
+ "q01": [
163
+ 0.004391033668071032,
164
+ -0.04016737639904022
165
+ ],
166
+ "q99": [
167
+ 0.040107566863298416,
168
+ -0.004319331608712673
169
+ ]
170
+ }
171
+ },
172
+ "action": {
173
+ "base_motion": {
174
+ "max": [
175
+ 1.0,
176
+ 1.0,
177
+ 1.0,
178
+ 0.0
179
+ ],
180
+ "min": [
181
+ -1.0,
182
+ -1.0,
183
+ -1.0,
184
+ 0.0
185
+ ],
186
+ "mean": [
187
+ 0.004284830763936043,
188
+ -0.0006925500929355621,
189
+ 0.0039020790718495846,
190
+ 0.0
191
+ ],
192
+ "std": [
193
+ 0.10394105315208435,
194
+ 0.08029993623495102,
195
+ 0.056489117443561554,
196
+ 0.0
197
+ ],
198
+ "q01": [
199
+ 0.0,
200
+ 0.0,
201
+ 0.0,
202
+ 0.0
203
+ ],
204
+ "q99": [
205
+ 0.4357142746448517,
206
+ 0.0,
207
+ 0.019999999552965164,
208
+ 0.0
209
+ ]
210
+ },
211
+ "control_mode": {
212
+ "max": [
213
+ 1.0
214
+ ],
215
+ "min": [
216
+ -1.0
217
+ ],
218
+ "mean": [
219
+ -0.9516769647598267
220
+ ],
221
+ "std": [
222
+ 0.3068857192993164
223
+ ],
224
+ "q01": [
225
+ -1.0
226
+ ],
227
+ "q99": [
228
+ 1.0
229
+ ]
230
+ },
231
+ "end_effector_position": {
232
+ "max": [
233
+ 1.0,
234
+ 1.0,
235
+ 1.0
236
+ ],
237
+ "min": [
238
+ -1.0,
239
+ -1.0,
240
+ -1.0
241
+ ],
242
+ "mean": [
243
+ -0.02753020077943802,
244
+ 0.0057201809249818325,
245
+ 0.0012639283668249846
246
+ ],
247
+ "std": [
248
+ 0.5038783550262451,
249
+ 0.22889363765716553,
250
+ 0.4663223922252655
251
+ ],
252
+ "q01": [
253
+ -1.0,
254
+ -0.8142856955528259,
255
+ -1.0
256
+ ],
257
+ "q99": [
258
+ 1.0,
259
+ 0.7642857432365417,
260
+ 1.0
261
+ ]
262
+ },
263
+ "end_effector_rotation": {
264
+ "max": [
265
+ 0.8514285683631897,
266
+ 1.0,
267
+ 0.868571400642395
268
+ ],
269
+ "min": [
270
+ -0.8828571438789368,
271
+ -0.9085714221000671,
272
+ -0.7457143068313599
273
+ ],
274
+ "mean": [
275
+ -0.0030895336531102657,
276
+ -0.029187792912125587,
277
+ -0.005299148615449667
278
+ ],
279
+ "std": [
280
+ 0.06469492614269257,
281
+ 0.11590102314949036,
282
+ 0.07264100760221481
283
+ ],
284
+ "q01": [
285
+ -0.2314285784959793,
286
+ -0.4514285624027252,
287
+ -0.23999999463558197
288
+ ],
289
+ "q99": [
290
+ 0.2085714340209961,
291
+ 0.2628571391105652,
292
+ 0.23999999463558197
293
+ ]
294
+ },
295
+ "gripper_close": {
296
+ "max": [
297
+ 1.0
298
+ ],
299
+ "min": [
300
+ -1.0
301
+ ],
302
+ "mean": [
303
+ -0.08604742586612701
304
+ ],
305
+ "std": [
306
+ 0.9964773058891296
307
+ ],
308
+ "q01": [
309
+ -1.0
310
+ ],
311
+ "q99": [
312
+ 1.0
313
+ ]
314
+ }
315
+ }
316
+ },
317
+ "modalities": {
318
+ "video": {
319
+ "left_view": {
320
+ "resolution": [
321
+ 256,
322
+ 256
323
+ ],
324
+ "channels": 3,
325
+ "fps": 20.0
326
+ },
327
+ "right_view": {
328
+ "resolution": [
329
+ 256,
330
+ 256
331
+ ],
332
+ "channels": 3,
333
+ "fps": 20.0
334
+ },
335
+ "wrist_view": {
336
+ "resolution": [
337
+ 256,
338
+ 256
339
+ ],
340
+ "channels": 3,
341
+ "fps": 20.0
342
+ }
343
+ },
344
+ "state": {
345
+ "base_position": {
346
+ "absolute": true,
347
+ "rotation_type": null,
348
+ "shape": [
349
+ 3
350
+ ],
351
+ "continuous": true
352
+ },
353
+ "base_rotation": {
354
+ "absolute": true,
355
+ "rotation_type": "quaternion",
356
+ "shape": [
357
+ 4
358
+ ],
359
+ "continuous": true
360
+ },
361
+ "end_effector_position_relative": {
362
+ "absolute": true,
363
+ "rotation_type": null,
364
+ "shape": [
365
+ 3
366
+ ],
367
+ "continuous": true
368
+ },
369
+ "end_effector_rotation_relative": {
370
+ "absolute": true,
371
+ "rotation_type": "quaternion",
372
+ "shape": [
373
+ 4
374
+ ],
375
+ "continuous": true
376
+ },
377
+ "gripper_qpos": {
378
+ "absolute": true,
379
+ "rotation_type": null,
380
+ "shape": [
381
+ 2
382
+ ],
383
+ "continuous": true
384
+ }
385
+ },
386
+ "action": {
387
+ "base_motion": {
388
+ "absolute": true,
389
+ "rotation_type": null,
390
+ "shape": [
391
+ 4
392
+ ],
393
+ "continuous": true
394
+ },
395
+ "control_mode": {
396
+ "absolute": true,
397
+ "rotation_type": null,
398
+ "shape": [
399
+ 1
400
+ ],
401
+ "continuous": true
402
+ },
403
+ "end_effector_position": {
404
+ "absolute": true,
405
+ "rotation_type": null,
406
+ "shape": [
407
+ 3
408
+ ],
409
+ "continuous": true
410
+ },
411
+ "end_effector_rotation": {
412
+ "absolute": true,
413
+ "rotation_type": null,
414
+ "shape": [
415
+ 3
416
+ ],
417
+ "continuous": true
418
+ },
419
+ "gripper_close": {
420
+ "absolute": true,
421
+ "rotation_type": null,
422
+ "shape": [
423
+ 1
424
+ ],
425
+ "continuous": true
426
+ }
427
+ }
428
+ },
429
+ "embodiment_tag": "new_embodiment"
430
+ }
431
+ }
base60k/model-00001-of-00002.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:45f60160e02c63cecd01b8979288a6292bb47e2461821be6db32cb1ce238ab1c
3
+ size 4999367032
base60k/model-00002-of-00002.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c80aa6362bb7158601a1d675034dc1fc112e4c2676f7cb85fe1db6f26f4060ff
3
+ size 2586508600
base60k/model.safetensors.index.json ADDED
The diff for this file is too large to render. See raw diff
 
base60k/trainer_state.json ADDED
The diff for this file is too large to render. See raw diff