Hugging Face's logo

Dongkkka
/
pick_multi_obj_fix_task_30k

Safetensors
gr00t_n1_5
Model card Files
xet
 Community
Dongkkka commited on
Commit
6c73751
·
verified ·
1 Parent(s): af8c34b

Upload folder using huggingface_hub

Browse files
Files changed (6) hide show
  1. config.json +64 -0
  2. experiment_cfg/metadata.json +307 -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
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 +1,307 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "new_embodiment": {
3
+ "statistics": {
4
+ "state": {
5
+ "left_arm": {
6
+ "max": [
7
+ 0.4827006161212921,
8
+ 0.8112002015113831,
9
+ 0.4532553553581238,
10
+ -0.8231125473976135,
11
+ 1.1653939485549927,
12
+ 0.7899761199951172,
13
+ 0.19781433045864105,
14
+ 1.0499905347824097
15
+ ],
16
+ "min": [
17
+ -1.0513520240783691,
18
+ -0.000107858024421148,
19
+ -1.1072344779968262,
20
+ -2.296549081802368,
21
+ -0.07028748095035553,
22
+ -0.5859926342964172,
23
+ -1.47036874294281,
24
+ 0.0
25
+ ],
26
+ "mean": [
27
+ 0.011497579514980316,
28
+ 0.19215743243694305,
29
+ -0.1451243907213211,
30
+ -1.4894176721572876,
31
+ 0.3010796308517456,
32
+ -0.09618806838989258,
33
+ -0.3399161398410797,
34
+ 0.2933500409126282
35
+ ],
36
+ "std": [
37
+ 0.3714962303638458,
38
+ 0.19363699853420258,
39
+ 0.2132025510072708,
40
+ 0.216729074716568,
41
+ 0.26094645261764526,
42
+ 0.3879685401916504,
43
+ 0.44094032049179077,
44
+ 0.32030728459358215
45
+ ],
46
+ "q01": [
47
+ -0.9140344530344009,
48
+ 0.0,
49
+ -0.7611510992050171,
50
+ -2.0928124308586122,
51
+ -0.008116316050291061,
52
+ -0.5690907031297684,
53
+ -1.2905553519725799,
54
+ 0.029702702537178993
55
+ ],
56
+ "q99": [
57
+ 0.3913862168788912,
58
+ 0.7450544595718385,
59
+ 0.3400661617517477,
60
+ -1.013698261976242,
61
+ 0.9407364845275887,
62
+ 0.6568368136882784,
63
+ 0.18469616919755938,
64
+ 1.042564868927002
65
+ ]
66
+ },
67
+ "right_arm": {
68
+ "max": [
69
+ 0.6417672038078308,
70
+ -0.0014980281703174114,
71
+ 0.792744517326355,
72
+ -1.0384330749511719,
73
+ 0.6077679991722107,
74
+ 0.9873323440551758,
75
+ 1.360622525215149,
76
+ 1.0603865385055542
77
+ ],
78
+ "min": [
79
+ -0.9047250747680664,
80
+ -0.8373377919197083,
81
+ -0.6043045520782471,
82
+ -2.3454806804656982,
83
+ -0.7402056455612183,
84
+ -0.7786630392074585,
85
+ -0.33405861258506775,
86
+ 0.0
87
+ ],
88
+ "mean": [
89
+ 0.09296321868896484,
90
+ -0.21980760991573334,
91
+ 0.15327434241771698,
92
+ -1.646316647529602,
93
+ -0.2041722536087036,
94
+ -0.03187763690948486,
95
+ 0.24834179878234863,
96
+ 0.325268417596817
97
+ ],
98
+ "std": [
99
+ 0.3301715552806854,
100
+ 0.172244131565094,
101
+ 0.2095300257205963,
102
+ 0.24973280727863312,
103
+ 0.1718880832195282,
104
+ 0.4755953550338745,
105
+ 0.4145625829696655,
106
+ 0.33061647415161133
107
+ ],
108
+ "q01": [
109
+ -0.7729669392108918,
110
+ -0.7347564280033112,
111
+ -0.4285223290324211,
112
+ -2.2316149473190308,
113
+ -0.6489248216152191,
114
+ -0.731025755405426,
115
+ -0.25769850611686707,
116
+ 0.06831621378660202
117
+ ],
118
+ "q99": [
119
+ 0.5628942608833314,
120
+ -0.052740177512168744,
121
+ 0.5677921921014794,
122
+ -1.2287228107452393,
123
+ 0.12717000618577062,
124
+ 0.8326435863971711,
125
+ 1.1078305482864381,
126
+ 1.0574162006378174
127
+ ]
128
+ }
129
+ },
130
+ "action": {
131
+ "left_arm": {
132
+ "max": [
133
+ 0.4832039475440979,
134
+ 0.8114758133888245,
135
+ 0.45405831933021545,
136
+ -0.8222137093544006,
137
+ 1.165825366973877,
138
+ 0.7930680513381958,
139
+ 0.1978835165500641,
140
+ 1.1789709329605103
141
+ ],
142
+ "min": [
143
+ -1.052310824394226,
144
+ -0.08130098134279251,
145
+ -1.1090681552886963,
146
+ -2.297903299331665,
147
+ -0.07056311517953873,
148
+ -0.5859806537628174,
149
+ -1.4710875749588013,
150
+ -0.11570879071950912
151
+ ],
152
+ "mean": [
153
+ 0.011788768693804741,
154
+ 0.1899206042289734,
155
+ -0.14491689205169678,
156
+ -1.489245057106018,
157
+ 0.3011428117752075,
158
+ -0.09616442769765854,
159
+ -0.3399713933467865,
160
+ 0.29545462131500244
161
+ ],
162
+ "std": [
163
+ 0.371960312128067,
164
+ 0.19602903723716736,
165
+ 0.21364881098270416,
166
+ 0.21710477769374847,
167
+ 0.26112067699432373,
168
+ 0.38814306259155273,
169
+ 0.4409879148006439,
170
+ 0.3344240188598633
171
+ ],
172
+ "q01": [
173
+ -0.9142525792121887,
174
+ -0.032213594764471054,
175
+ -0.7639224529266357,
176
+ -2.093883752822876,
177
+ -0.00920388475060463,
178
+ -0.5691068768501282,
179
+ -1.2916117906570435,
180
+ 0.01007763110101223
181
+ ],
182
+ "q99": [
183
+ 0.3957670331001282,
184
+ 0.7455146908760071,
185
+ 0.34207770228385925,
186
+ -1.0124273300170898,
187
+ 0.9418641924858093,
188
+ 0.6565437912940979,
189
+ 0.18561168015003204,
190
+ 1.124131202697765
191
+ ]
192
+ },
193
+ "right_arm": {
194
+ "max": [
195
+ 0.6442719101905823,
196
+ -2.0694557179012918e-13,
197
+ 0.7930680513381958,
198
+ -1.0369709730148315,
199
+ 0.6089903712272644,
200
+ 0.9894176125526428,
201
+ 1.3959225416183472,
202
+ 1.155961275100708
203
+ ],
204
+ "min": [
205
+ -0.905048668384552,
206
+ -0.8375535011291504,
207
+ -0.6059224009513855,
208
+ -2.3469905853271484,
209
+ -0.7409127354621887,
210
+ -0.7792622447013855,
211
+ -0.3344078063964844,
212
+ -0.0788932517170906
213
+ ],
214
+ "mean": [
215
+ 0.09317154437303543,
216
+ -0.21982984244823456,
217
+ 0.15331701934337616,
218
+ -1.6462136507034302,
219
+ -0.20444351434707642,
220
+ -0.03180892765522003,
221
+ 0.2483728677034378,
222
+ 0.326642245054245
223
+ ],
224
+ "std": [
225
+ 0.33065080642700195,
226
+ 0.17234687507152557,
227
+ 0.20994938910007477,
228
+ 0.2501552402973175,
229
+ 0.1719447821378708,
230
+ 0.4758480191230774,
231
+ 0.4146299362182617,
232
+ 0.34059736132621765
233
+ ],
234
+ "q01": [
235
+ -0.7731263041496277,
236
+ -0.7342399001121521,
237
+ -0.429514616727829,
238
+ -2.2314050436019897,
239
+ -0.6519418358802795,
240
+ -0.731708824634552,
241
+ -0.25770875811576843,
242
+ 0.05149511247873306
243
+ ],
244
+ "q99": [
245
+ 0.5675728917121887,
246
+ -0.05215534567832947,
247
+ 0.5660389065742493,
248
+ -1.228718638420105,
249
+ 0.1298514693975471,
250
+ 0.832951545715332,
251
+ 1.1085312604904196,
252
+ 1.1222137212753296
253
+ ]
254
+ }
255
+ }
256
+ },
257
+ "modalities": {
258
+ "video": {
259
+ "cam_head": {
260
+ "resolution": [
261
+ 672,
262
+ 376
263
+ ],
264
+ "channels": 3,
265
+ "fps": 10.0
266
+ }
267
+ },
268
+ "state": {
269
+ "left_arm": {
270
+ "absolute": true,
271
+ "rotation_type": null,
272
+ "shape": [
273
+ 8
274
+ ],
275
+ "continuous": true
276
+ },
277
+ "right_arm": {
278
+ "absolute": true,
279
+ "rotation_type": null,
280
+ "shape": [
281
+ 8
282
+ ],
283
+ "continuous": true
284
+ }
285
+ },
286
+ "action": {
287
+ "left_arm": {
288
+ "absolute": true,
289
+ "rotation_type": null,
290
+ "shape": [
291
+ 8
292
+ ],
293
+ "continuous": true
294
+ },
295
+ "right_arm": {
296
+ "absolute": true,
297
+ "rotation_type": null,
298
+ "shape": [
299
+ 8
300
+ ],
301
+ "continuous": true
302
+ }
303
+ }
304
+ },
305
+ "embodiment_tag": "new_embodiment"
306
+ }
307
+ }
model-00001-of-00002.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b02a884cbc6ba98b3c152331f9d8fb153ec24479be5c64ad57adedcb0f6fb516
3
+ size 4999367032
model-00002-of-00002.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5758618c81c9f2ccb83e6df1a3bd6906aa14e24844d9cf15affe53dc873c8f2e
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