Hugging Face's logo

paragon7060
/
Robocasa_baseline

Safetensors
gr00t_n1_5
Model card Files
xet
 Community
paragon7060 commited on
Commit
2c334d7
·
verified ·
1 Parent(s): c4f854a

Upload folder using huggingface_hub

Browse files
Files changed (9) hide show
  1. config.json +64 -0
  2. experiment_cfg/metadata.json +431 -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. optimizer.pt +3 -0
  7. rng_state.pth +3 -0
  8. scheduler.pt +3 -0
  9. trainer_state.json +0 -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 +1,431 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "new_embodiment": {
3
+ "statistics": {
4
+ "state": {
5
+ "base_position": {
6
+ "max": [
7
+ 7.778976917266846,
8
+ 0.7572364211082458,
9
+ 0.7678815722465515
10
+ ],
11
+ "min": [
12
+ -5.451417922973633,
13
+ -8.747910499572754,
14
+ 0.698784351348877
15
+ ],
16
+ "mean": [
17
+ 2.654967701417028,
18
+ -2.312603201514601,
19
+ 0.7017931665650737
20
+ ],
21
+ "std": [
22
+ 1.6812542217813051,
23
+ 1.4374053366477129,
24
+ 0.002770274630317656
25
+ ],
26
+ "q01": [
27
+ -1.2119644922112263,
28
+ -5.737460157529482,
29
+ 0.6999888640441118
30
+ ],
31
+ "q99": [
32
+ 6.229334757497698,
33
+ -0.6044882808532064,
34
+ 0.7073385971611953
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
+ 0.24505679210493964,
54
+ 0.6304187740544468
55
+ ],
56
+ "std": [
57
+ 0.0,
58
+ 0.0,
59
+ 0.6484714011255448,
60
+ 0.34942432790807293
61
+ ],
62
+ "q01": [
63
+ 0.0,
64
+ 0.0,
65
+ -0.9999876197837357,
66
+ 0.0001443402896033188
67
+ ],
68
+ "q99": [
69
+ 0.0,
70
+ 0.0,
71
+ 0.9999981170335596,
72
+ 0.99999996133623
73
+ ]
74
+ },
75
+ "end_effector_position_relative": {
76
+ "max": [
77
+ 0.9316536784172058,
78
+ 0.8727958798408508,
79
+ 1.1222138404846191
80
+ ],
81
+ "min": [
82
+ -0.65090411901474,
83
+ -0.903346061706543,
84
+ -0.37153130769729614
85
+ ],
86
+ "mean": [
87
+ 0.24668259593208566,
88
+ -0.012032116142909142,
89
+ 0.5279085030194163
90
+ ],
91
+ "std": [
92
+ 0.19344092845015276,
93
+ 0.1930747899058619,
94
+ 0.19997707447442778
95
+ ],
96
+ "q01": [
97
+ -0.09279797820659887,
98
+ -0.4254360876401248,
99
+ 0.20971629323845067
100
+ ],
101
+ "q99": [
102
+ 0.6337033079770765,
103
+ 0.41157325993908306,
104
+ 0.7842848743908305
105
+ ]
106
+ },
107
+ "end_effector_rotation_relative": {
108
+ "max": [
109
+ 0.9999998807907104,
110
+ 0.9999892115592957,
111
+ 0.9948683977127075,
112
+ 0.9997240900993347
113
+ ],
114
+ "min": [
115
+ -0.9999992847442627,
116
+ -0.9999884366989136,
117
+ -0.9975370168685913,
118
+ 7.53001938580411e-10
119
+ ],
120
+ "mean": [
121
+ -0.18716363477102008,
122
+ 0.042197768367289466,
123
+ -0.057659459752153995,
124
+ 0.12923932727163936
125
+ ],
126
+ "std": [
127
+ 0.8359975444297222,
128
+ 0.3154394462749756,
129
+ 0.34675285652401194,
130
+ 0.15659141437591545
131
+ ],
132
+ "q01": [
133
+ -0.9936889269479753,
134
+ -0.6207200408872208,
135
+ -0.6139638960815086,
136
+ 0.0013660874448937702
137
+ ],
138
+ "q99": [
139
+ 0.9852533529246098,
140
+ 0.7212727283914762,
141
+ 0.5828153666912674,
142
+ 0.5189229249802788
143
+ ]
144
+ },
145
+ "gripper_qpos": {
146
+ "max": [
147
+ 0.07638481259346008,
148
+ 0.030545983463525772
149
+ ],
150
+ "min": [
151
+ -0.04314137250185013,
152
+ -0.07785148173570633
153
+ ],
154
+ "mean": [
155
+ 0.03010611853568522,
156
+ -0.03007088309534884
157
+ ],
158
+ "std": [
159
+ 0.012650814924005526,
160
+ 0.01265015748860496
161
+ ],
162
+ "q01": [
163
+ 0.005405729914847853,
164
+ -0.04027899166624506
165
+ ],
166
+ "q99": [
167
+ 0.04022964370674609,
168
+ -0.005456649609281796
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
+ 0.06118170421366438,
188
+ 0.005168978720450757,
189
+ 0.00015468911936722662,
190
+ 0.0
191
+ ],
192
+ "std": [
193
+ 0.28311138077516473,
194
+ 0.23894521906956986,
195
+ 0.17381784712136492,
196
+ 0.0
197
+ ],
198
+ "q01": [
199
+ -0.562252432369087,
200
+ -0.6371663064504606,
201
+ -0.4676562510790457,
202
+ 0.0
203
+ ],
204
+ "q99": [
205
+ 0.6772610732113116,
206
+ 0.6611900172189822,
207
+ 0.47041472061780404,
208
+ 0.0
209
+ ]
210
+ },
211
+ "control_mode": {
212
+ "max": [
213
+ 1.0
214
+ ],
215
+ "min": [
216
+ -1.0
217
+ ],
218
+ "mean": [
219
+ -0.565887839502144
220
+ ],
221
+ "std": [
222
+ 0.8245004495696986
223
+ ],
224
+ "q01": [
225
+ -1.0
226
+ ],
227
+ "q99": [
228
+ 0.6670156612373467
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
+ 0.003785921291709651,
244
+ -0.0025378753488897797,
245
+ -0.014835714235104965
246
+ ],
247
+ "std": [
248
+ 0.3511517723668662,
249
+ 0.2665986378244751,
250
+ 0.28413315125899574
251
+ ],
252
+ "q01": [
253
+ -0.9302930273370403,
254
+ -0.7655163753477748,
255
+ -0.6654673233949292
256
+ ],
257
+ "q99": [
258
+ 0.7677186796978582,
259
+ 0.7610623906968231,
260
+ 0.8621295315171748
261
+ ]
262
+ },
263
+ "end_effector_rotation": {
264
+ "max": [
265
+ 1.0,
266
+ 1.0,
267
+ 1.0
268
+ ],
269
+ "min": [
270
+ -1.0,
271
+ -1.0,
272
+ -1.0
273
+ ],
274
+ "mean": [
275
+ 0.0005460055915413661,
276
+ -0.00519104931885604,
277
+ 0.0005011079496210638
278
+ ],
279
+ "std": [
280
+ 0.07288433955483174,
281
+ 0.0912551221625068,
282
+ 0.08750437622572464
283
+ ],
284
+ "q01": [
285
+ -0.22605902557440774,
286
+ -0.30891056138292344,
287
+ -0.28242365421680377
288
+ ],
289
+ "q99": [
290
+ 0.22656327959372116,
291
+ 0.26343943799376573,
292
+ 0.2894539337544894
293
+ ]
294
+ },
295
+ "gripper_close": {
296
+ "max": [
297
+ 1.0
298
+ ],
299
+ "min": [
300
+ -1.0
301
+ ],
302
+ "mean": [
303
+ -0.11074762091339947
304
+ ],
305
+ "std": [
306
+ 0.9938708219067919
307
+ ],
308
+ "q01": [
309
+ -1.0
310
+ ],
311
+ "q99": [
312
+ 0.9821338671287022
313
+ ]
314
+ }
315
+ }
316
+ },
317
+ "modalities": {
318
+ "video": {
319
+ "robot0_eye_in_hand": {
320
+ "resolution": [
321
+ 256,
322
+ 256
323
+ ],
324
+ "channels": 3,
325
+ "fps": 20.0
326
+ },
327
+ "robot0_agentview_left": {
328
+ "resolution": [
329
+ 256,
330
+ 256
331
+ ],
332
+ "channels": 3,
333
+ "fps": 20.0
334
+ },
335
+ "robot0_agentview_right": {
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": "axis_angle",
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
+ }
model-00001-of-00002.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:08f1891947973e2e5ec2422201cd90261806f77f2634f9ec0477c27aa5a4fe42
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:deb9c9cf40cd8983a7779af85341f6344db15f65bf3307073a0e3c3085450435
3
+ size 2586705312
model.safetensors.index.json ADDED
The diff for this file is too large to render. See raw diff
 
optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4d59aa8fae6642def95034580444222027900a537b2fd99c2a49d4e53ca574fc
3
+ size 8550719959
rng_state.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e097f2ff2c1e1790217877d411f313d095262121e5999071b7d2c9c2aa94b442
3
+ size 14645
scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:abc2607874971bcbd0e7fc3a3cb9d7f94cfea97632a7acd7c7510e0b729784fc
3
+ size 1465
trainer_state.json ADDED
The diff for this file is too large to render. See raw diff