Hugging Face's logo

JisooSong
/
parcel-sorting-model

Robotics
Safetensors
gr00t_n1_5
Model card Files
xet
Community
JisooSong commited on
Commit
3125e1a
·
verified ·
1 Parent(s): 8acc6a4

Upload folder using huggingface_hub

Browse files
Files changed (10) hide show
  1. config.json +64 -0
  2. experiment_cfg/metadata.json +371 -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_0.pth +3 -0
  8. rng_state_1.pth +3 -0
  9. scheduler.pt +3 -0
  10. 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 +1,371 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "new_embodiment": {
3
+ "statistics": {
4
+ "state": {
5
+ "left_arm": {
6
+ "max": [
7
+ 95.09536743164062,
8
+ 83.27091217041016,
9
+ 100.0,
10
+ 100.0,
11
+ 99.32415008544922
12
+ ],
13
+ "min": [
14
+ -55.839820861816406,
15
+ -100.0,
16
+ -98.37606811523438,
17
+ 1.2027491331100464,
18
+ -44.68416976928711
19
+ ],
20
+ "mean": [
21
+ 22.159082505954693,
22
+ -36.1522865844335,
23
+ 38.110837989207816,
24
+ 67.59448525514152,
25
+ 3.4253542602952947
26
+ ],
27
+ "std": [
28
+ 25.716990628812276,
29
+ 47.698537858795426,
30
+ 46.19350431771322,
31
+ 19.1841209477407,
32
+ 6.8097502388992615
33
+ ],
34
+ "q01": [
35
+ -12.002842900445998,
36
+ -99.43286145634731,
37
+ -52.656071342898684,
38
+ 25.23272637526786,
39
+ -5.732454395524938
40
+ ],
41
+ "q99": [
42
+ 69.51329636367103,
43
+ 48.929294089234084,
44
+ 98.55713962476287,
45
+ 95.4776345731188,
46
+ 15.56874376041509
47
+ ]
48
+ },
49
+ "left_gripper": {
50
+ "max": [
51
+ 64.16666412353516
52
+ ],
53
+ "min": [
54
+ 0.8333333134651184
55
+ ],
56
+ "mean": [
57
+ 13.689288553762635
58
+ ],
59
+ "std": [
60
+ 12.984300951913312
61
+ ],
62
+ "q01": [
63
+ 1.5826803881258833
64
+ ],
65
+ "q99": [
66
+ 42.40724866789503
67
+ ]
68
+ },
69
+ "right_arm": {
70
+ "max": [
71
+ 43.771827697753906,
72
+ 100.0,
73
+ 100.0,
74
+ 99.57319641113281,
75
+ 48.945037841796875
76
+ ],
77
+ "min": [
78
+ -88.61972045898438,
79
+ -100.0,
80
+ -99.28121948242188,
81
+ -22.321809768676758,
82
+ -36.29939651489258
83
+ ],
84
+ "mean": [
85
+ -11.67856744467494,
86
+ -27.4429899886685,
87
+ 28.373878773228622,
88
+ 76.41580615328233,
89
+ 4.1577284065544955
90
+ ],
91
+ "std": [
92
+ 27.48116988702998,
93
+ 50.246549138232034,
94
+ 50.428660409743784,
95
+ 20.372480781249013,
96
+ 7.417763186003502
97
+ ],
98
+ "q01": [
99
+ -63.34283040685163,
100
+ -99.65360342652822,
101
+ -53.90581389345294,
102
+ 38.0710301283186,
103
+ -7.54200355355562
104
+ ],
105
+ "q99": [
106
+ 22.858352383463803,
107
+ 47.312231534870655,
108
+ 99.20298660571919,
109
+ 97.11130862243914,
110
+ 16.009986043500536
111
+ ]
112
+ },
113
+ "right_gripper": {
114
+ "max": [
115
+ 60.18067932128906
116
+ ],
117
+ "min": [
118
+ 0.47297295928001404
119
+ ],
120
+ "mean": [
121
+ 15.0308968130323
122
+ ],
123
+ "std": [
124
+ 12.944993422715605
125
+ ],
126
+ "q01": [
127
+ 1.1132611453090515
128
+ ],
129
+ "q99": [
130
+ 40.36817944594688
131
+ ]
132
+ }
133
+ },
134
+ "action": {
135
+ "left_arm": {
136
+ "max": [
137
+ 95.11292266845703,
138
+ 83.48008728027344,
139
+ 100.0,
140
+ 100.0,
141
+ 99.7901382446289
142
+ ],
143
+ "min": [
144
+ -59.876922607421875,
145
+ -100.0,
146
+ -99.20844268798828,
147
+ 0.9055627584457397,
148
+ -45.068206787109375
149
+ ],
150
+ "mean": [
151
+ 22.28279387487105,
152
+ -37.873241275764556,
153
+ 36.35136246987811,
154
+ 67.46190909381968,
155
+ 3.4606613264174837
156
+ ],
157
+ "std": [
158
+ 25.73276263621616,
159
+ 46.3602547836817,
160
+ 46.991273477061156,
161
+ 19.487802396685474,
162
+ 6.816464369724153
163
+ ],
164
+ "q01": [
165
+ -12.023950153302826,
166
+ -99.52489027062533,
167
+ -55.16293811446299,
168
+ 24.360247713540204,
169
+ -5.789599367663552
170
+ ],
171
+ "q99": [
172
+ 69.71304509923148,
173
+ 46.8898585382468,
174
+ 98.96163892620517,
175
+ 95.75905762129268,
176
+ 15.64590279202312
177
+ ]
178
+ },
179
+ "left_gripper": {
180
+ "max": [
181
+ 64.88095092773438
182
+ ],
183
+ "min": [
184
+ 0.0
185
+ ],
186
+ "mean": [
187
+ 12.387821083805033
188
+ ],
189
+ "std": [
190
+ 14.105989284781801
191
+ ],
192
+ "q01": [
193
+ 0.6867846786860053
194
+ ],
195
+ "q99": [
196
+ 42.865011836182404
197
+ ]
198
+ },
199
+ "right_arm": {
200
+ "max": [
201
+ 43.94752502441406,
202
+ 100.0,
203
+ 100.0,
204
+ 100.0,
205
+ 49.066036224365234
206
+ ],
207
+ "min": [
208
+ -88.1812973022461,
209
+ -100.0,
210
+ -99.82110595703125,
211
+ -22.641510009765625,
212
+ -36.32310485839844
213
+ ],
214
+ "mean": [
215
+ -11.571535549452912,
216
+ -29.573800742848473,
217
+ 26.32569949873592,
218
+ 76.4456978336207,
219
+ 4.198700937075703
220
+ ],
221
+ "std": [
222
+ 27.546868517767308,
223
+ 48.91467407896806,
224
+ 51.189359786111034,
225
+ 20.74346295925323,
226
+ 7.4243332666729085
227
+ ],
228
+ "q01": [
229
+ -63.39823076319857,
230
+ -99.89433600217218,
231
+ -57.04760030889267,
232
+ 37.37837890082765,
233
+ -7.541886147434907
234
+ ],
235
+ "q99": [
236
+ 23.038255638783205,
237
+ 45.9272852524054,
238
+ 99.5496477035711,
239
+ 97.82369947320387,
240
+ 16.083525047556254
241
+ ]
242
+ },
243
+ "right_gripper": {
244
+ "max": [
245
+ 61.71564865112305
246
+ ],
247
+ "min": [
248
+ 0.0
249
+ ],
250
+ "mean": [
251
+ 12.529417928774118
252
+ ],
253
+ "std": [
254
+ 14.318071919612695
255
+ ],
256
+ "q01": [
257
+ 0.6170909220211473
258
+ ],
259
+ "q99": [
260
+ 40.59343580611316
261
+ ]
262
+ }
263
+ }
264
+ },
265
+ "modalities": {
266
+ "video": {
267
+ "left": {
268
+ "resolution": [
269
+ 640,
270
+ 480
271
+ ],
272
+ "channels": 3,
273
+ "fps": 30.0
274
+ },
275
+ "right": {
276
+ "resolution": [
277
+ 640,
278
+ 480
279
+ ],
280
+ "channels": 3,
281
+ "fps": 30.0
282
+ },
283
+ "top": {
284
+ "resolution": [
285
+ 640,
286
+ 480
287
+ ],
288
+ "channels": 3,
289
+ "fps": 30.0
290
+ },
291
+ "front": {
292
+ "resolution": [
293
+ 640,
294
+ 480
295
+ ],
296
+ "channels": 3,
297
+ "fps": 30.0
298
+ }
299
+ },
300
+ "state": {
301
+ "left_arm": {
302
+ "absolute": true,
303
+ "rotation_type": null,
304
+ "shape": [
305
+ 5
306
+ ],
307
+ "continuous": true
308
+ },
309
+ "left_gripper": {
310
+ "absolute": true,
311
+ "rotation_type": null,
312
+ "shape": [
313
+ 1
314
+ ],
315
+ "continuous": true
316
+ },
317
+ "right_arm": {
318
+ "absolute": true,
319
+ "rotation_type": null,
320
+ "shape": [
321
+ 5
322
+ ],
323
+ "continuous": true
324
+ },
325
+ "right_gripper": {
326
+ "absolute": true,
327
+ "rotation_type": null,
328
+ "shape": [
329
+ 1
330
+ ],
331
+ "continuous": true
332
+ }
333
+ },
334
+ "action": {
335
+ "left_arm": {
336
+ "absolute": true,
337
+ "rotation_type": null,
338
+ "shape": [
339
+ 5
340
+ ],
341
+ "continuous": true
342
+ },
343
+ "left_gripper": {
344
+ "absolute": true,
345
+ "rotation_type": null,
346
+ "shape": [
347
+ 1
348
+ ],
349
+ "continuous": true
350
+ },
351
+ "right_arm": {
352
+ "absolute": true,
353
+ "rotation_type": null,
354
+ "shape": [
355
+ 5
356
+ ],
357
+ "continuous": true
358
+ },
359
+ "right_gripper": {
360
+ "absolute": true,
361
+ "rotation_type": null,
362
+ "shape": [
363
+ 1
364
+ ],
365
+ "continuous": true
366
+ }
367
+ }
368
+ },
369
+ "embodiment_tag": "new_embodiment"
370
+ }
371
+ }
model-00001-of-00002.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8507f66df7c18532374828171b920e05f81d166e745a930a13737b4fa16f4faf
3
+ size 4999367032
model-00002-of-00002.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3e2b95d61b0e04fcd58a954db062a90c73001bfa5f6a10a7ca94330b96e82660
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 sha256:1b95a32d8b5766a8dab382c732ec2a5faca4c9b4c08fe62af2b1849db24966b3
3
+ size 8550724414
rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:be996e2c5fdc628d9ff72f6447a1ecc203cd2ac5e57212f93ac446e1524332ef
3
+ size 14512
rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:41907395b0e7fc3eadce1ebdf4721c0d1f54effc755f116c1677eea302d9829c
3
+ size 14512
scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a795a9825ff1276eafcf67772ce481f34c0c2ba70e0e2990913ac41e49a5bdae
3
+ size 1064
trainer_state.json ADDED
The diff for this file is too large to render. See raw diff