Hugging Face's logo

lucasngoo
/
GR00T-N1.5-humanoid-box-folding

Safetensors
gr00t_n1_5
Model card Files Community
lucasngoo commited on
Commit
d9a910c
·
verified ·
1 Parent(s): d03daf2

Upload folder using huggingface_hub

Browse files
Files changed (9) hide show
  1. config.json +64 -0
  2. experiment_cfg/metadata.json +347 -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": 50,
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": 50,
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 +1,347 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "new_embodiment": {
3
+ "statistics": {
4
+ "state": {
5
+ "left_arm": {
6
+ "max": [
7
+ 25.91289520263672,
8
+ 6.778072834014893,
9
+ 17.10537338256836,
10
+ 72.08064270019531,
11
+ 50.418907165527344
12
+ ],
13
+ "min": [
14
+ -25.265907287597656,
15
+ -28.317729949951172,
16
+ -39.24317169189453,
17
+ -2.8656461238861084,
18
+ -8.798368453979492
19
+ ],
20
+ "mean": [
21
+ -0.28367045521736145,
22
+ -5.662345886230469,
23
+ -10.992593765258789,
24
+ 44.8178596496582,
25
+ 25.731718063354492
26
+ ],
27
+ "std": [
28
+ 7.767245292663574,
29
+ 7.235373020172119,
30
+ 10.450653076171875,
31
+ 19.069732666015625,
32
+ 13.327289581298828
33
+ ],
34
+ "q01": [
35
+ -15.805314064025879,
36
+ -22.9526424407959,
37
+ -31.300299339294433,
38
+ 0.3387473225593567,
39
+ -4.354917049407959
40
+ ],
41
+ "q99": [
42
+ 19.583444595336914,
43
+ 6.558338165283203,
44
+ 13.192961692810059,
45
+ 71.65340423583984,
46
+ 50.27852249145508
47
+ ]
48
+ },
49
+ "left_gripper": {
50
+ "max": [
51
+ 72.15686798095703
52
+ ],
53
+ "min": [
54
+ 0.0
55
+ ],
56
+ "mean": [
57
+ 11.430745124816895
58
+ ],
59
+ "std": [
60
+ 15.247529983520508
61
+ ],
62
+ "q01": [
63
+ 0.0
64
+ ],
65
+ "q99": [
66
+ 51.37255096435547
67
+ ]
68
+ },
69
+ "right_arm": {
70
+ "max": [
71
+ 32.14467239379883,
72
+ 14.773802757263184,
73
+ 25.473440170288086,
74
+ 71.21392059326172,
75
+ 41.50152587890625
76
+ ],
77
+ "min": [
78
+ -4.99580192565918,
79
+ -23.654558181762695,
80
+ -10.31204605102539,
81
+ -0.9308037757873535,
82
+ -26.40117835998535
83
+ ],
84
+ "mean": [
85
+ 13.737095832824707,
86
+ -4.50025749206543,
87
+ 8.567008018493652,
88
+ 45.279117584228516,
89
+ -0.6063000559806824
90
+ ],
91
+ "std": [
92
+ 8.034978866577148,
93
+ 6.2780256271362305,
94
+ 7.485732078552246,
95
+ 23.365392684936523,
96
+ 10.581278800964355
97
+ ],
98
+ "q01": [
99
+ 0.16992037922143954,
100
+ -20.199907302856445,
101
+ -8.277472190856933,
102
+ 0.9552103281021118,
103
+ -21.11545753479004
104
+ ],
105
+ "q99": [
106
+ 30.2525577545166,
107
+ 11.146169357299781,
108
+ 21.43284797668457,
109
+ 70.66458892822266,
110
+ 34.79975128173828
111
+ ]
112
+ },
113
+ "right_gripper": {
114
+ "max": [
115
+ 44.313724517822266
116
+ ],
117
+ "min": [
118
+ 0.0
119
+ ],
120
+ "mean": [
121
+ 9.014527320861816
122
+ ],
123
+ "std": [
124
+ 12.955764770507812
125
+ ],
126
+ "q01": [
127
+ 0.0
128
+ ],
129
+ "q99": [
130
+ 40.0
131
+ ]
132
+ }
133
+ },
134
+ "action": {
135
+ "left_arm": {
136
+ "max": [
137
+ 26.153480529785156,
138
+ 21.16117286682129,
139
+ 17.25690460205078,
140
+ 76.69793701171875,
141
+ 50.54612731933594
142
+ ],
143
+ "min": [
144
+ -27.18766975402832,
145
+ -29.027196884155273,
146
+ -38.694984436035156,
147
+ -2.9206554889678955,
148
+ -8.928709030151367
149
+ ],
150
+ "mean": [
151
+ -0.9697821736335754,
152
+ -5.286567211151123,
153
+ -10.967265129089355,
154
+ 45.5112419128418,
155
+ 25.749237060546875
156
+ ],
157
+ "std": [
158
+ 8.039172172546387,
159
+ 8.12271499633789,
160
+ 10.513473510742188,
161
+ 19.38804054260254,
162
+ 13.361382484436035
163
+ ],
164
+ "q01": [
165
+ -17.0665283203125,
166
+ -23.195911407470703,
167
+ -31.021818161010742,
168
+ 0.35705602169036865,
169
+ -4.24599552154541
170
+ ],
171
+ "q99": [
172
+ 19.51510578155497,
173
+ 14.871246337890625,
174
+ 13.35810661315918,
175
+ 73.53941345214844,
176
+ 50.41483688354492
177
+ ]
178
+ },
179
+ "left_gripper": {
180
+ "max": [
181
+ 72.32975006103516
182
+ ],
183
+ "min": [
184
+ 0.0
185
+ ],
186
+ "mean": [
187
+ 11.728493690490723
188
+ ],
189
+ "std": [
190
+ 15.288747787475586
191
+ ],
192
+ "q01": [
193
+ 0.0
194
+ ],
195
+ "q99": [
196
+ 51.87526977539035
197
+ ]
198
+ },
199
+ "right_arm": {
200
+ "max": [
201
+ 32.64799880981445,
202
+ 16.040616989135742,
203
+ 25.249710083007812,
204
+ 80.02906799316406,
205
+ 41.51949691772461
206
+ ],
207
+ "min": [
208
+ -6.512839317321777,
209
+ -24.186840057373047,
210
+ -10.549227714538574,
211
+ -1.0488430261611938,
212
+ -26.454090118408203
213
+ ],
214
+ "mean": [
215
+ 13.61085319519043,
216
+ -4.462998390197754,
217
+ 8.643723487854004,
218
+ 45.99699401855469,
219
+ -0.5744842290878296
220
+ ],
221
+ "std": [
222
+ 8.170876502990723,
223
+ 6.442399501800537,
224
+ 7.559754371643066,
225
+ 23.84986114501953,
226
+ 10.585643768310547
227
+ ],
228
+ "q01": [
229
+ -0.6387133002281189,
230
+ -20.612510681152344,
231
+ -8.342443466186523,
232
+ 1.0601201057434082,
233
+ -21.21265411376953
234
+ ],
235
+ "q99": [
236
+ 30.47239875793457,
237
+ 11.751421928405762,
238
+ 21.31264400482162,
239
+ 74.40516662597656,
240
+ 34.79229736328125
241
+ ]
242
+ },
243
+ "right_gripper": {
244
+ "max": [
245
+ 45.036643981933594
246
+ ],
247
+ "min": [
248
+ 0.0
249
+ ],
250
+ "mean": [
251
+ 9.347930908203125
252
+ ],
253
+ "std": [
254
+ 12.987605094909668
255
+ ],
256
+ "q01": [
257
+ 0.06662224978208542
258
+ ],
259
+ "q99": [
260
+ 40.4397087097168
261
+ ]
262
+ }
263
+ }
264
+ },
265
+ "modalities": {
266
+ "video": {
267
+ "front": {
268
+ "resolution": [
269
+ 640,
270
+ 480
271
+ ],
272
+ "channels": 3,
273
+ "fps": 30.0
274
+ }
275
+ },
276
+ "state": {
277
+ "left_arm": {
278
+ "absolute": true,
279
+ "rotation_type": null,
280
+ "shape": [
281
+ 5
282
+ ],
283
+ "continuous": true
284
+ },
285
+ "left_gripper": {
286
+ "absolute": true,
287
+ "rotation_type": null,
288
+ "shape": [
289
+ 1
290
+ ],
291
+ "continuous": true
292
+ },
293
+ "right_arm": {
294
+ "absolute": true,
295
+ "rotation_type": null,
296
+ "shape": [
297
+ 5
298
+ ],
299
+ "continuous": true
300
+ },
301
+ "right_gripper": {
302
+ "absolute": true,
303
+ "rotation_type": null,
304
+ "shape": [
305
+ 1
306
+ ],
307
+ "continuous": true
308
+ }
309
+ },
310
+ "action": {
311
+ "left_arm": {
312
+ "absolute": true,
313
+ "rotation_type": null,
314
+ "shape": [
315
+ 5
316
+ ],
317
+ "continuous": true
318
+ },
319
+ "left_gripper": {
320
+ "absolute": true,
321
+ "rotation_type": null,
322
+ "shape": [
323
+ 1
324
+ ],
325
+ "continuous": true
326
+ },
327
+ "right_arm": {
328
+ "absolute": true,
329
+ "rotation_type": null,
330
+ "shape": [
331
+ 5
332
+ ],
333
+ "continuous": true
334
+ },
335
+ "right_gripper": {
336
+ "absolute": true,
337
+ "rotation_type": null,
338
+ "shape": [
339
+ 1
340
+ ],
341
+ "continuous": true
342
+ }
343
+ }
344
+ },
345
+ "embodiment_tag": "new_embodiment"
346
+ }
347
+ }
model-00001-of-00002.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1f42ea87ec07e60e9c7f35b917e0240a97171d7d05ebf6e4323d00aebe565eeb
3
+ size 4999367032
model-00002-of-00002.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:95fdc3f348876a1f36b4e08a23cb7e91d9e003386342c500528b1bdc219cf59d
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 sha256:2c431a870272435331af9af0a564757e48ba8effbb7ed92edf440f436b4e571a
3
+ size 8550720062
rng_state.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f01f3638036e80af5515003c0053445eee7bde2fb1bea7807679491c0c6c038c
3
+ size 14244
scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:07356cb668b8efc440ddb2f7f81f7da3152060e81d0229922fa3b30df3313f82
3
+ size 1064
trainer_state.json ADDED
The diff for this file is too large to render. See raw diff