Hugging Face's logo

robocasa
/
robocasa365_checkpoints

Safetensors
Model card Files
xet
 Community
1
Abhi03 commited on
Commit
c484448
·
verified ·
1 Parent(s): 1489599

Add files using upload-large-folder tool

Browse files
Files changed (47) hide show
  1. gr00t_n1-5/target_fraction/atomic_seen_10p/README.md +35 -0
  2. gr00t_n1-5/target_fraction/atomic_seen_10p/config.json +64 -0
  3. gr00t_n1-5/target_fraction/atomic_seen_10p/experiment_cfg/metadata.json +431 -0
  4. gr00t_n1-5/target_fraction/atomic_seen_10p/model-00001-of-00002.safetensors +3 -0
  5. gr00t_n1-5/target_fraction/atomic_seen_10p/model-00002-of-00002.safetensors +3 -0
  6. gr00t_n1-5/target_fraction/atomic_seen_10p/model.safetensors.index.json +0 -0
  7. gr00t_n1-5/target_fraction/atomic_seen_10p/trainer_state.json +0 -0
  8. gr00t_n1-5/target_fraction/atomic_seen_30p/README.md +33 -0
  9. gr00t_n1-5/target_fraction/atomic_seen_30p/config.json +64 -0
  10. gr00t_n1-5/target_fraction/atomic_seen_30p/eval_results.json +74 -0
  11. gr00t_n1-5/target_fraction/atomic_seen_30p/experiment_cfg/metadata.json +431 -0
  12. gr00t_n1-5/target_fraction/atomic_seen_30p/model-00001-of-00002.safetensors +3 -0
  13. gr00t_n1-5/target_fraction/atomic_seen_30p/model-00002-of-00002.safetensors +3 -0
  14. gr00t_n1-5/target_fraction/atomic_seen_30p/model.safetensors.index.json +0 -0
  15. gr00t_n1-5/target_fraction/atomic_seen_30p/trainer_state.json +0 -0
  16. gr00t_n1-5/target_fraction/composite_seen_10p/README.md +31 -0
  17. gr00t_n1-5/target_fraction/composite_seen_10p/config.json +64 -0
  18. gr00t_n1-5/target_fraction/composite_seen_10p/eval_results.json +66 -0
  19. gr00t_n1-5/target_fraction/composite_seen_10p/experiment_cfg/metadata.json +431 -0
  20. gr00t_n1-5/target_fraction/composite_seen_10p/model-00001-of-00002.safetensors +3 -0
  21. gr00t_n1-5/target_fraction/composite_seen_10p/model-00002-of-00002.safetensors +3 -0
  22. gr00t_n1-5/target_fraction/composite_seen_10p/model.safetensors.index.json +0 -0
  23. gr00t_n1-5/target_fraction/composite_seen_10p/trainer_state.json +0 -0
  24. gr00t_n1-5/target_fraction/composite_seen_30p/README.md +31 -0
  25. gr00t_n1-5/target_fraction/composite_seen_30p/config.json +64 -0
  26. gr00t_n1-5/target_fraction/composite_seen_30p/eval_results.json +66 -0
  27. gr00t_n1-5/target_fraction/composite_seen_30p/experiment_cfg/metadata.json +431 -0
  28. gr00t_n1-5/target_fraction/composite_seen_30p/model-00001-of-00002.safetensors +3 -0
  29. gr00t_n1-5/target_fraction/composite_seen_30p/model-00002-of-00002.safetensors +3 -0
  30. gr00t_n1-5/target_fraction/composite_seen_30p/model.safetensors.index.json +0 -0
  31. gr00t_n1-5/target_fraction/composite_seen_30p/trainer_state.json +0 -0
  32. gr00t_n1-5/target_fraction/composite_unseen_10p/README.md +31 -0
  33. gr00t_n1-5/target_fraction/composite_unseen_10p/config.json +64 -0
  34. gr00t_n1-5/target_fraction/composite_unseen_10p/eval_results.json +94 -0
  35. gr00t_n1-5/target_fraction/composite_unseen_10p/experiment_cfg/metadata.json +431 -0
  36. gr00t_n1-5/target_fraction/composite_unseen_10p/model-00001-of-00002.safetensors +3 -0
  37. gr00t_n1-5/target_fraction/composite_unseen_10p/model-00002-of-00002.safetensors +3 -0
  38. gr00t_n1-5/target_fraction/composite_unseen_10p/model.safetensors.index.json +0 -0
  39. gr00t_n1-5/target_fraction/composite_unseen_10p/trainer_state.json +0 -0
  40. gr00t_n1-5/target_fraction/composite_unseen_30p/README.md +31 -0
  41. gr00t_n1-5/target_fraction/composite_unseen_30p/config.json +64 -0
  42. gr00t_n1-5/target_fraction/composite_unseen_30p/eval_results.json +90 -0
  43. gr00t_n1-5/target_fraction/composite_unseen_30p/experiment_cfg/metadata.json +431 -0
  44. gr00t_n1-5/target_fraction/composite_unseen_30p/model-00001-of-00002.safetensors +3 -0
  45. gr00t_n1-5/target_fraction/composite_unseen_30p/model-00002-of-00002.safetensors +3 -0
  46. gr00t_n1-5/target_fraction/composite_unseen_30p/model.safetensors.index.json +0 -0
  47. gr00t_n1-5/target_fraction/composite_unseen_30p/trainer_state.json +0 -0
gr00t_n1-5/target_fraction/atomic_seen_10p/README.md ADDED
@@ -0,0 +1,35 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # GR00T N1.5 — RoboCasa `atomic_seen`, pretraining + post-training on 10% target data
2
+
3
+ Setting: **Pretraining + Target Post-Training (10%)**.
4
+
5
+ - Stage 1 (pretraining): GR00T N1.5 (`nvidia/GR00T-N1.5-3B`) trained on the RoboCasa
6
+ pretraining corpus (300 human demos/task + MimicGen), 80k steps.
7
+ - Stage 2 (post-training): this checkpoint — continued training on the `atomic_seen` target
8
+ tasks using 50 of the 500 human demos per task (10%), 60k steps, batch size 128,
9
+ peak LR 3e-5 (cosine, 5% warmup), vision tower and LLM frozen.
10
+
11
+ Evaluated with 30 rollouts per task on the held-out `test` split.
12
+
13
+ > **Note:** evaluation rollouts for this checkpoint were not retained, so per-task numbers are unavailable here.
14
+
15
+ | Task | Success rate (%) | Rollouts |
16
+ |---|---|---|
17
+ | CloseBlenderLid | – | – |
18
+ | CloseFridge | – | – |
19
+ | CloseToasterOvenDoor | – | – |
20
+ | CoffeeSetupMug | – | – |
21
+ | NavigateKitchen | – | – |
22
+ | OpenCabinet | – | – |
23
+ | OpenDrawer | – | – |
24
+ | OpenStandMixerHead | – | – |
25
+ | PickPlaceCounterToCabinet | – | – |
26
+ | PickPlaceCounterToStove | – | – |
27
+ | PickPlaceDrawerToCounter | – | – |
28
+ | PickPlaceSinkToCounter | – | – |
29
+ | PickPlaceToasterToCounter | – | – |
30
+ | SlideDishwasherRack | – | – |
31
+ | TurnOffStove | – | – |
32
+ | TurnOnElectricKettle | – | – |
33
+ | TurnOnMicrowave | – | – |
34
+ | TurnOnSinkFaucet | – | – |
35
+ | **Average** | **–** | |
gr00t_n1-5/target_fraction/atomic_seen_10p/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
+ }
gr00t_n1-5/target_fraction/atomic_seen_10p/experiment_cfg/metadata.json ADDED
@@ -0,0 +1,431 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "new_embodiment": {
3
+ "statistics": {
4
+ "state": {
5
+ "base_position": {
6
+ "max": [
7
+ 5.6508049964904785,
8
+ -0.3116215467453003,
9
+ 0.7259154915809631
10
+ ],
11
+ "min": [
12
+ -0.18281720578670502,
13
+ -5.046002388000488,
14
+ 0.6994456052780151
15
+ ],
16
+ "mean": [
17
+ 2.5493806007557045,
18
+ -1.5390440980247186,
19
+ 0.7009784298143833
20
+ ],
21
+ "std": [
22
+ 1.4253797879186358,
23
+ 0.8980984326280239,
24
+ 0.0015021808227890538
25
+ ],
26
+ "q01": [
27
+ 0.6657724927447427,
28
+ -3.631848556511482,
29
+ 0.6999751984012619
30
+ ],
31
+ "q99": [
32
+ 5.12559897514716,
33
+ -0.7038390281231623,
34
+ 0.7043861784587546
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
+ 0.281700787603147,
54
+ 0.630464182135629
55
+ ],
56
+ "std": [
57
+ 0.0,
58
+ 0.0,
59
+ 0.6367096666286696,
60
+ 0.3438833742309608
61
+ ],
62
+ "q01": [
63
+ 0.0,
64
+ 0.0,
65
+ -0.932615719817279,
66
+ 0.04762755174397524
67
+ ],
68
+ "q99": [
69
+ 0.0,
70
+ 0.0,
71
+ 0.9803092118492496,
72
+ 0.9736528770428411
73
+ ]
74
+ },
75
+ "end_effector_position_relative": {
76
+ "max": [
77
+ 0.9091058969497681,
78
+ 0.8446807861328125,
79
+ 1.0843316316604614
80
+ ],
81
+ "min": [
82
+ -0.6169384717941284,
83
+ -0.8689756989479065,
84
+ -0.2231825590133667
85
+ ],
86
+ "mean": [
87
+ 0.2628751127215861,
88
+ -0.03473324004586387,
89
+ 0.4644125295758262
90
+ ],
91
+ "std": [
92
+ 0.16684970460713566,
93
+ 0.25566680631641875,
94
+ 0.22053197440759104
95
+ ],
96
+ "q01": [
97
+ -0.020079493302181794,
98
+ -0.4894094962776976,
99
+ 0.22971941192880552
100
+ ],
101
+ "q99": [
102
+ 0.5461262942953219,
103
+ 0.42565034948724817,
104
+ 0.7052325618142467
105
+ ]
106
+ },
107
+ "end_effector_rotation_relative": {
108
+ "max": [
109
+ 0.999998927116394,
110
+ 0.9999622106552124,
111
+ 0.9589148759841919,
112
+ 0.9470692873001099
113
+ ],
114
+ "min": [
115
+ -0.9999971985816956,
116
+ -0.9999222159385681,
117
+ -0.961084246635437,
118
+ 1.524969519550723e-07
119
+ ],
120
+ "mean": [
121
+ -0.26002845234917366,
122
+ 0.026552968081773636,
123
+ -0.07165353651450326,
124
+ 0.16271486350899184
125
+ ],
126
+ "std": [
127
+ 0.7981302540159743,
128
+ 0.33059453529696203,
129
+ 0.3510504321701125,
130
+ 0.17470599946915794
131
+ ],
132
+ "q01": [
133
+ -0.9947738625671444,
134
+ -0.5913023445767219,
135
+ -0.6086838450334485,
136
+ 0.0021780892904438953
137
+ ],
138
+ "q99": [
139
+ 0.9282418689956956,
140
+ 0.6958814819500573,
141
+ 0.5733852310980178,
142
+ 0.5142617974766991
143
+ ]
144
+ },
145
+ "gripper_qpos": {
146
+ "max": [
147
+ 0.06876722723245621,
148
+ 0.026710085570812225
149
+ ],
150
+ "min": [
151
+ -0.029329104349017143,
152
+ -0.07029139250516891
153
+ ],
154
+ "mean": [
155
+ 0.031609066021187834,
156
+ -0.03202025248711912
157
+ ],
158
+ "std": [
159
+ 0.012650339858458329,
160
+ 0.012210710510126082
161
+ ],
162
+ "q01": [
163
+ 0.003146712640177711,
164
+ -0.04063613303123828
165
+ ],
166
+ "q99": [
167
+ 0.040539700037192664,
168
+ -0.005423341258821252
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
+ 0.007875813897841435,
188
+ -0.0006477674302115731,
189
+ -0.0004135653917642353,
190
+ 0.0
191
+ ],
192
+ "std": [
193
+ 0.10277249601569914,
194
+ 0.09112102918226186,
195
+ 0.08247047002145606,
196
+ 0.0
197
+ ],
198
+ "q01": [
199
+ -0.0607993814937069,
200
+ -0.10008969570792621,
201
+ -0.05568222663543766,
202
+ 0.0
203
+ ],
204
+ "q99": [
205
+ 0.06492485525365585,
206
+ 0.06612125042907188,
207
+ 0.04739571014707229,
208
+ 0.0
209
+ ]
210
+ },
211
+ "control_mode": {
212
+ "max": [
213
+ 1.0
214
+ ],
215
+ "min": [
216
+ -1.0
217
+ ],
218
+ "mean": [
219
+ -0.9359017917961401
220
+ ],
221
+ "std": [
222
+ 0.352250522340036
223
+ ],
224
+ "q01": [
225
+ -1.0
226
+ ],
227
+ "q99": [
228
+ -0.4322972825423608
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
+ -0.006181831797632424,
244
+ -0.01634249962910833,
245
+ -0.04985652844458544
246
+ ],
247
+ "std": [
248
+ 0.4112843674434335,
249
+ 0.40499048162861406,
250
+ 0.369152790657577
251
+ ],
252
+ "q01": [
253
+ -0.9708802076625289,
254
+ -0.944519675986929,
255
+ -0.8411703234712035
256
+ ],
257
+ "q99": [
258
+ 0.764806134291762,
259
+ 0.9025713738076095,
260
+ 0.8172220525729417
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
+ 0.00693045185983346,
276
+ -0.023410988123315653,
277
+ -0.0024104932381308593
278
+ ],
279
+ "std": [
280
+ 0.1189423798155561,
281
+ 0.12845264915281232,
282
+ 0.12717348556426816
283
+ ],
284
+ "q01": [
285
+ -0.3214619629767296,
286
+ -0.4588578347191383,
287
+ -0.3744009816145433
288
+ ],
289
+ "q99": [
290
+ 0.35093587568958734,
291
+ 0.30115206231544006,
292
+ 0.34910549446377226
293
+ ]
294
+ },
295
+ "gripper_close": {
296
+ "max": [
297
+ 1.0
298
+ ],
299
+ "min": [
300
+ -1.0
301
+ ],
302
+ "mean": [
303
+ -0.36268394967465506
304
+ ],
305
+ "std": [
306
+ 0.9319752289134154
307
+ ],
308
+ "q01": [
309
+ -1.0
310
+ ],
311
+ "q99": [
312
+ 0.7553010808181337
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
+ }
gr00t_n1-5/target_fraction/atomic_seen_10p/model-00001-of-00002.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:04fc01a2417986584c64af36600b228908e8d8430b75f0e36e6d94f6f766044e
3
+ size 4999367032
gr00t_n1-5/target_fraction/atomic_seen_10p/model-00002-of-00002.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:95a59705d4095db5fe785d2b12107356959a4428c3151e9998aa2065df4d4219
3
+ size 2586705312
gr00t_n1-5/target_fraction/atomic_seen_10p/model.safetensors.index.json ADDED
The diff for this file is too large to render. See raw diff
 
gr00t_n1-5/target_fraction/atomic_seen_10p/trainer_state.json ADDED
The diff for this file is too large to render. See raw diff
 
gr00t_n1-5/target_fraction/atomic_seen_30p/README.md ADDED
@@ -0,0 +1,33 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # GR00T N1.5 — RoboCasa `atomic_seen`, pretraining + post-training on 30% target data
2
+
3
+ Setting: **Pretraining + Target Post-Training (30%)**.
4
+
5
+ - Stage 1 (pretraining): GR00T N1.5 (`nvidia/GR00T-N1.5-3B`) trained on the RoboCasa
6
+ pretraining corpus (300 human demos/task + MimicGen), 80k steps.
7
+ - Stage 2 (post-training): this checkpoint — continued training on the `atomic_seen` target
8
+ tasks using 150 of the 500 human demos per task (30%), 60k steps, batch size 128,
9
+ peak LR 3e-5 (cosine, 5% warmup), vision tower and LLM frozen.
10
+
11
+ Evaluated with 30 rollouts per task on the held-out `test` split.
12
+
13
+ | Task | Success rate (%) | Rollouts |
14
+ |---|---|---|
15
+ | CloseBlenderLid | 16.7 | 30 |
16
+ | CloseFridge | 60.0 | 30 |
17
+ | CloseToasterOvenDoor | 63.3 | 30 |
18
+ | CoffeeSetupMug | 40.0 | 30 |
19
+ | NavigateKitchen | 30.0 | 30 |
20
+ | OpenCabinet | 73.3 | 30 |
21
+ | OpenDrawer | 83.3 | 30 |
22
+ | OpenStandMixerHead | 90.0 | 30 |
23
+ | PickPlaceCounterToCabinet | 60.0 | 30 |
24
+ | PickPlaceCounterToStove | 63.3 | 30 |
25
+ | PickPlaceDrawerToCounter | 53.3 | 30 |
26
+ | PickPlaceSinkToCounter | 56.7 | 30 |
27
+ | PickPlaceToasterToCounter | 76.7 | 30 |
28
+ | SlideDishwasherRack | 76.7 | 30 |
29
+ | TurnOffStove | 33.3 | 30 |
30
+ | TurnOnElectricKettle | 70.0 | 30 |
31
+ | TurnOnMicrowave | 56.7 | 30 |
32
+ | TurnOnSinkFaucet | 60.0 | 30 |
33
+ | **Average** | **59.1** | |
gr00t_n1-5/target_fraction/atomic_seen_30p/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
+ }
gr00t_n1-5/target_fraction/atomic_seen_30p/eval_results.json ADDED
@@ -0,0 +1,74 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "CloseBlenderLid": {
3
+ "num_episodes": 30,
4
+ "success_rate": 0.16666666666666666
5
+ },
6
+ "CloseFridge": {
7
+ "num_episodes": 30,
8
+ "success_rate": 0.6
9
+ },
10
+ "CloseToasterOvenDoor": {
11
+ "num_episodes": 30,
12
+ "success_rate": 0.6333333333333333
13
+ },
14
+ "CoffeeSetupMug": {
15
+ "num_episodes": 30,
16
+ "success_rate": 0.4
17
+ },
18
+ "NavigateKitchen": {
19
+ "num_episodes": 30,
20
+ "success_rate": 0.3
21
+ },
22
+ "OpenCabinet": {
23
+ "num_episodes": 30,
24
+ "success_rate": 0.7333333333333333
25
+ },
26
+ "OpenDrawer": {
27
+ "num_episodes": 30,
28
+ "success_rate": 0.8333333333333334
29
+ },
30
+ "OpenStandMixerHead": {
31
+ "num_episodes": 30,
32
+ "success_rate": 0.9
33
+ },
34
+ "PnPCounterToCabinet": {
35
+ "num_episodes": 30,
36
+ "success_rate": 0.6
37
+ },
38
+ "PnPCounterToStove": {
39
+ "num_episodes": 30,
40
+ "success_rate": 0.6333333333333333
41
+ },
42
+ "PnPDrawerToCounter": {
43
+ "num_episodes": 30,
44
+ "success_rate": 0.5333333333333333
45
+ },
46
+ "PnPSinkToCounter": {
47
+ "num_episodes": 30,
48
+ "success_rate": 0.5666666666666667
49
+ },
50
+ "PnPToasterToCounter": {
51
+ "num_episodes": 30,
52
+ "success_rate": 0.7666666666666667
53
+ },
54
+ "SlideDishwasherRack": {
55
+ "num_episodes": 30,
56
+ "success_rate": 0.7666666666666667
57
+ },
58
+ "TurnOffStove": {
59
+ "num_episodes": 30,
60
+ "success_rate": 0.3333333333333333
61
+ },
62
+ "TurnOnElectricKettle": {
63
+ "num_episodes": 30,
64
+ "success_rate": 0.7
65
+ },
66
+ "TurnOnMicrowave": {
67
+ "num_episodes": 30,
68
+ "success_rate": 0.5666666666666667
69
+ },
70
+ "TurnOnSinkFaucet": {
71
+ "num_episodes": 30,
72
+ "success_rate": 0.6
73
+ }
74
+ }
gr00t_n1-5/target_fraction/atomic_seen_30p/experiment_cfg/metadata.json ADDED
@@ -0,0 +1,431 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "new_embodiment": {
3
+ "statistics": {
4
+ "state": {
5
+ "base_position": {
6
+ "max": [
7
+ 5.6508049964904785,
8
+ -0.3116215467453003,
9
+ 0.7259154915809631
10
+ ],
11
+ "min": [
12
+ -0.18281720578670502,
13
+ -5.046002388000488,
14
+ 0.6994456052780151
15
+ ],
16
+ "mean": [
17
+ 2.5522164625560984,
18
+ -1.5351394637099773,
19
+ 0.7009723094583877
20
+ ],
21
+ "std": [
22
+ 1.421104107579562,
23
+ 0.8966581209856037,
24
+ 0.0015046092401204503
25
+ ],
26
+ "q01": [
27
+ 0.6716205581760701,
28
+ -3.62540590590063,
29
+ 0.6999747383148612
30
+ ],
31
+ "q99": [
32
+ 5.123225719117989,
33
+ -0.705310359913161,
34
+ 0.704394924157765
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
+ 0.28337931917672304,
54
+ 0.6314460741965231
55
+ ],
56
+ "std": [
57
+ 0.0,
58
+ 0.0,
59
+ 0.6354280013096522,
60
+ 0.34307309698340693
61
+ ],
62
+ "q01": [
63
+ 0.0,
64
+ 0.0,
65
+ -0.929790770409868,
66
+ 0.04962419972392856
67
+ ],
68
+ "q99": [
69
+ 0.0,
70
+ 0.0,
71
+ 0.9794837148625262,
72
+ 0.9736637892896679
73
+ ]
74
+ },
75
+ "end_effector_position_relative": {
76
+ "max": [
77
+ 0.9091058969497681,
78
+ 0.8446807861328125,
79
+ 1.0843316316604614
80
+ ],
81
+ "min": [
82
+ -0.6169384717941284,
83
+ -0.8689756989479065,
84
+ -0.2231825590133667
85
+ ],
86
+ "mean": [
87
+ 0.26313885023210104,
88
+ -0.034286421424737375,
89
+ 0.46391291306527993
90
+ ],
91
+ "std": [
92
+ 0.16671149318997278,
93
+ 0.25482370734650756,
94
+ 0.22040869918085113
95
+ ],
96
+ "q01": [
97
+ -0.018992565019639168,
98
+ -0.48762155926249273,
99
+ 0.2302469172053066
100
+ ],
101
+ "q99": [
102
+ 0.5466692453479169,
103
+ 0.4235498594314706,
104
+ 0.7039285837884217
105
+ ]
106
+ },
107
+ "end_effector_rotation_relative": {
108
+ "max": [
109
+ 0.999998927116394,
110
+ 0.9999622106552124,
111
+ 0.9589148759841919,
112
+ 0.9470692873001099
113
+ ],
114
+ "min": [
115
+ -0.9999971985816956,
116
+ -0.9999222159385681,
117
+ -0.961084246635437,
118
+ 1.524969519550723e-07
119
+ ],
120
+ "mean": [
121
+ -0.2590153409959691,
122
+ 0.027026150278575616,
123
+ -0.07058457653893126,
124
+ 0.16206217748187443
125
+ ],
126
+ "std": [
127
+ 0.7989553614284364,
128
+ 0.3303452362779523,
129
+ 0.3507338935809783,
130
+ 0.17451624399783655
131
+ ],
132
+ "q01": [
133
+ -0.9947814845421474,
134
+ -0.5920348288719453,
135
+ -0.6081925797899597,
136
+ 0.0021657908788476985
137
+ ],
138
+ "q99": [
139
+ 0.9270059520053112,
140
+ 0.6942225018748309,
141
+ 0.572749546037488,
142
+ 0.5135132014947437
143
+ ]
144
+ },
145
+ "gripper_qpos": {
146
+ "max": [
147
+ 0.06876722723245621,
148
+ 0.026710085570812225
149
+ ],
150
+ "min": [
151
+ -0.029329104349017143,
152
+ -0.07029139250516891
153
+ ],
154
+ "mean": [
155
+ 0.03155140007399895,
156
+ -0.031968310456181856
157
+ ],
158
+ "std": [
159
+ 0.012693781093778977,
160
+ 0.012252604074497478
161
+ ],
162
+ "q01": [
163
+ 0.0031699178006506365,
164
+ -0.04063450812742773
165
+ ],
166
+ "q99": [
167
+ 0.04053575418062887,
168
+ -0.00536568416525792
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
+ 0.00825659364631405,
188
+ -0.0006346648623404031,
189
+ -0.0004467892483576082,
190
+ 0.0
191
+ ],
192
+ "std": [
193
+ 0.10470026325791158,
194
+ 0.09274961921079236,
195
+ 0.08414924951358817,
196
+ 0.0
197
+ ],
198
+ "q01": [
199
+ -0.06181609331494543,
200
+ -0.10196249404128609,
201
+ -0.057284617965416146,
202
+ 0.0
203
+ ],
204
+ "q99": [
205
+ 0.06550187875670213,
206
+ 0.06788172030985429,
207
+ 0.04881912790338798,
208
+ 0.0
209
+ ]
210
+ },
211
+ "control_mode": {
212
+ "max": [
213
+ 1.0
214
+ ],
215
+ "min": [
216
+ -1.0
217
+ ],
218
+ "mean": [
219
+ -0.9338199338198251
220
+ ],
221
+ "std": [
222
+ 0.3577342419432837
223
+ ],
224
+ "q01": [
225
+ -1.0
226
+ ],
227
+ "q99": [
228
+ -0.433171283067158
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
+ -0.006553009138441189,
244
+ -0.016608022290262266,
245
+ -0.05016173417530997
246
+ ],
247
+ "std": [
248
+ 0.40987177722476126,
249
+ 0.40383962826157893,
250
+ 0.36895317559738666
251
+ ],
252
+ "q01": [
253
+ -0.9702605718564837,
254
+ -0.9429333006242672,
255
+ -0.842116058588727
256
+ ],
257
+ "q99": [
258
+ 0.7623192102305567,
259
+ 0.900901254260253,
260
+ 0.8155007925076305
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
+ 0.007067280474845929,
276
+ -0.02330899663479962,
277
+ -0.0026119012407210896
278
+ ],
279
+ "std": [
280
+ 0.11940363156877098,
281
+ 0.12838116089027235,
282
+ 0.127664081388098
283
+ ],
284
+ "q01": [
285
+ -0.3214241458414286,
286
+ -0.4585669505603246,
287
+ -0.37592454277308074
288
+ ],
289
+ "q99": [
290
+ 0.35188972878520564,
291
+ 0.30085425170856034,
292
+ 0.3484529695997845
293
+ ]
294
+ },
295
+ "gripper_close": {
296
+ "max": [
297
+ 1.0
298
+ ],
299
+ "min": [
300
+ -1.0
301
+ ],
302
+ "mean": [
303
+ -0.3598056821304454
304
+ ],
305
+ "std": [
306
+ 0.9330870600356279
307
+ ],
308
+ "q01": [
309
+ -1.0
310
+ ],
311
+ "q99": [
312
+ 0.7636621223538708
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
+ }
gr00t_n1-5/target_fraction/atomic_seen_30p/model-00001-of-00002.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8c4623d66b379039461f8561aab72296c172e93963190ca68f834e5d2a5914d8
3
+ size 4999367032
gr00t_n1-5/target_fraction/atomic_seen_30p/model-00002-of-00002.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1a28e35ba116404d24d35f6060cc6ba55867e83c68775cda42e9b92aa56fc1c8
3
+ size 2586705312
gr00t_n1-5/target_fraction/atomic_seen_30p/model.safetensors.index.json ADDED
The diff for this file is too large to render. See raw diff
 
gr00t_n1-5/target_fraction/atomic_seen_30p/trainer_state.json ADDED
The diff for this file is too large to render. See raw diff
 
gr00t_n1-5/target_fraction/composite_seen_10p/README.md ADDED
@@ -0,0 +1,31 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # GR00T N1.5 — RoboCasa `composite_seen`, pretraining + post-training on 10% target data
2
+
3
+ Setting: **Pretraining + Target Post-Training (10%)**.
4
+
5
+ - Stage 1 (pretraining): GR00T N1.5 (`nvidia/GR00T-N1.5-3B`) trained on the RoboCasa
6
+ pretraining corpus (300 human demos/task + MimicGen), 80k steps.
7
+ - Stage 2 (post-training): this checkpoint — continued training on the `composite_seen` target
8
+ tasks using 50 of the 500 human demos per task (10%), 60k steps, batch size 128,
9
+ peak LR 3e-5 (cosine, 5% warmup), vision tower and LLM frozen.
10
+
11
+ Evaluated with 30 rollouts per task on the held-out `test` split.
12
+
13
+ | Task | Success rate (%) | Rollouts |
14
+ |---|---|---|
15
+ | DeliverStraw | 0.0 | 30 |
16
+ | GetToastedBread | 0.0 | 30 |
17
+ | KettleBoiling | 13.3 | 30 |
18
+ | LoadDishwasher | 50.0 | 30 |
19
+ | PackIdenticalLunches | 0.0 | 30 |
20
+ | PreSoakPan | 26.7 | 30 |
21
+ | PrepareCoffee | 3.3 | 30 |
22
+ | RinseSinkBasin | 40.0 | 30 |
23
+ | ScrubCuttingBoard | 36.7 | 30 |
24
+ | SearingMeat | 3.3 | 30 |
25
+ | SetUpCuttingStation | 40.0 | 30 |
26
+ | StackBowlsCabinet | 63.3 | 30 |
27
+ | SteamInMicrowave | 26.7 | 30 |
28
+ | StirVegetables | 10.0 | 30 |
29
+ | StoreLeftoversInBowl | 56.7 | 30 |
30
+ | WashLettuce | 36.7 | 30 |
31
+ | **Average** | **25.4** | |
gr00t_n1-5/target_fraction/composite_seen_10p/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
+ }
gr00t_n1-5/target_fraction/composite_seen_10p/eval_results.json ADDED
@@ -0,0 +1,66 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "DeliverStraw": {
3
+ "num_episodes": 30,
4
+ "success_rate": 0.0
5
+ },
6
+ "GetToastedBread": {
7
+ "num_episodes": 30,
8
+ "success_rate": 0.0
9
+ },
10
+ "KettleBoiling": {
11
+ "num_episodes": 30,
12
+ "success_rate": 0.13333333333333333
13
+ },
14
+ "LoadDishwasher": {
15
+ "num_episodes": 30,
16
+ "success_rate": 0.5
17
+ },
18
+ "PackIdenticalLunches": {
19
+ "num_episodes": 30,
20
+ "success_rate": 0.0
21
+ },
22
+ "PreSoakPan": {
23
+ "num_episodes": 30,
24
+ "success_rate": 0.26666666666666666
25
+ },
26
+ "PrepareCoffee": {
27
+ "num_episodes": 30,
28
+ "success_rate": 0.03333333333333333
29
+ },
30
+ "RinseSinkBasin": {
31
+ "num_episodes": 30,
32
+ "success_rate": 0.4
33
+ },
34
+ "ScrubCuttingBoard": {
35
+ "num_episodes": 30,
36
+ "success_rate": 0.36666666666666664
37
+ },
38
+ "SearingMeat": {
39
+ "num_episodes": 30,
40
+ "success_rate": 0.03333333333333333
41
+ },
42
+ "SetUpCuttingStation": {
43
+ "num_episodes": 30,
44
+ "success_rate": 0.4
45
+ },
46
+ "StackBowlsCabinet": {
47
+ "num_episodes": 30,
48
+ "success_rate": 0.6333333333333333
49
+ },
50
+ "SteamInMicrowave": {
51
+ "num_episodes": 30,
52
+ "success_rate": 0.26666666666666666
53
+ },
54
+ "StirVegetables": {
55
+ "num_episodes": 30,
56
+ "success_rate": 0.1
57
+ },
58
+ "StoreLeftoversInBowl": {
59
+ "num_episodes": 30,
60
+ "success_rate": 0.5666666666666667
61
+ },
62
+ "WashLettuce": {
63
+ "num_episodes": 30,
64
+ "success_rate": 0.36666666666666664
65
+ }
66
+ }
gr00t_n1-5/target_fraction/composite_seen_10p/experiment_cfg/metadata.json ADDED
@@ -0,0 +1,431 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "new_embodiment": {
3
+ "statistics": {
4
+ "state": {
5
+ "base_position": {
6
+ "max": [
7
+ 5.65041971206665,
8
+ -0.2986648678779602,
9
+ 0.7586075663566589
10
+ ],
11
+ "min": [
12
+ -0.43130701780319214,
13
+ -5.801346778869629,
14
+ 0.6994776725769043
15
+ ],
16
+ "mean": [
17
+ 2.6920709930094637,
18
+ -1.8379249105552624,
19
+ 0.7010909795626654
20
+ ],
21
+ "std": [
22
+ 1.5068903285983084,
23
+ 1.053907905563134,
24
+ 0.003944710721403756
25
+ ],
26
+ "q01": [
27
+ 0.6182051942097797,
28
+ -3.6699359014139623,
29
+ 0.6999907291082175
30
+ ],
31
+ "q99": [
32
+ 5.282695460759009,
33
+ -0.7468700918857174,
34
+ 0.7084273627451048
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
+ 0.2513844690583616,
54
+ 0.6202965634590774
55
+ ],
56
+ "std": [
57
+ 0.0,
58
+ 0.0,
59
+ 0.6564040721145076,
60
+ 0.3483093177139186
61
+ ],
62
+ "q01": [
63
+ 0.0,
64
+ 0.0,
65
+ -0.9984730461294307,
66
+ 0.003902442919735478
67
+ ],
68
+ "q99": [
69
+ 0.0,
70
+ 0.0,
71
+ 0.9987377098753247,
72
+ 0.9998819394606573
73
+ ]
74
+ },
75
+ "end_effector_position_relative": {
76
+ "max": [
77
+ 0.8965585827827454,
78
+ 0.8544884920120239,
79
+ 1.0672767162322998
80
+ ],
81
+ "min": [
82
+ -0.5257036089897156,
83
+ -0.8542829751968384,
84
+ -0.30580952763557434
85
+ ],
86
+ "mean": [
87
+ 0.30679412708191117,
88
+ -0.01579092495114904,
89
+ 0.43521944528582074
90
+ ],
91
+ "std": [
92
+ 0.18326600225369144,
93
+ 0.23446922958145905,
94
+ 0.19147930242603217
95
+ ],
96
+ "q01": [
97
+ -0.026957846365062776,
98
+ -0.503495536971308,
99
+ 0.14170778049593882
100
+ ],
101
+ "q99": [
102
+ 0.6783935806803654,
103
+ 0.49414056059516126,
104
+ 0.725649085514759
105
+ ]
106
+ },
107
+ "end_effector_rotation_relative": {
108
+ "max": [
109
+ 0.9999983906745911,
110
+ 0.9999905824661255,
111
+ 0.9724889397621155,
112
+ 0.9802296757698059
113
+ ],
114
+ "min": [
115
+ -0.9999982118606567,
116
+ -0.9999850392341614,
117
+ -0.9627900123596191,
118
+ 2.564086543088706e-08
119
+ ],
120
+ "mean": [
121
+ -0.17919662801308273,
122
+ 0.07048151017494528,
123
+ -0.052269537146037784,
124
+ 0.133214650784382
125
+ ],
126
+ "std": [
127
+ 0.8425070365925962,
128
+ 0.344056034454254,
129
+ 0.29737370308508315,
130
+ 0.16065261474785456
131
+ ],
132
+ "q01": [
133
+ -0.9964180075714307,
134
+ -0.6607670303605057,
135
+ -0.609025917389068,
136
+ 0.001401202009437433
137
+ ],
138
+ "q99": [
139
+ 0.9927372962673179,
140
+ 0.8035775148923475,
141
+ 0.560790397570088,
142
+ 0.5464405954978925
143
+ ]
144
+ },
145
+ "gripper_qpos": {
146
+ "max": [
147
+ 0.05964253097772598,
148
+ 0.0066233607940375805
149
+ ],
150
+ "min": [
151
+ -0.019988659769296646,
152
+ -0.06823232024908066
153
+ ],
154
+ "mean": [
155
+ 0.02850532132699791,
156
+ -0.02850481227476295
157
+ ],
158
+ "std": [
159
+ 0.013392260016620725,
160
+ 0.013456426937901687
161
+ ],
162
+ "q01": [
163
+ 0.003970029605016947,
164
+ -0.04043246125943667
165
+ ],
166
+ "q99": [
167
+ 0.04036231710733215,
168
+ -0.004198575431747249
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
+ 0.03639835765225665,
188
+ 0.0018700378673464006,
189
+ -0.001275134650769611,
190
+ 0.0
191
+ ],
192
+ "std": [
193
+ 0.24092157627779898,
194
+ 0.23575745204418216,
195
+ 0.14144658921652953,
196
+ 0.0
197
+ ],
198
+ "q01": [
199
+ -0.4599637351048044,
200
+ -0.6449453792277065,
201
+ -0.4224815788330467,
202
+ 0.0
203
+ ],
204
+ "q99": [
205
+ 0.601479977877128,
206
+ 0.5828127113144466,
207
+ 0.3986812551063104,
208
+ 0.0
209
+ ]
210
+ },
211
+ "control_mode": {
212
+ "max": [
213
+ 1.0
214
+ ],
215
+ "min": [
216
+ -1.0
217
+ ],
218
+ "mean": [
219
+ -0.696800929106668
220
+ ],
221
+ "std": [
222
+ 0.717331824002298
223
+ ],
224
+ "q01": [
225
+ -1.0
226
+ ],
227
+ "q99": [
228
+ 0.6374072403866691
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
+ 0.011980253026554786,
244
+ -0.009904974687042014,
245
+ -0.02786220043625079
246
+ ],
247
+ "std": [
248
+ 0.41772955790483135,
249
+ 0.36016348170155765,
250
+ 0.37686543120121996
251
+ ],
252
+ "q01": [
253
+ -0.992418904589238,
254
+ -0.9264507487381988,
255
+ -0.887921306157055
256
+ ],
257
+ "q99": [
258
+ 0.9023091362882582,
259
+ 0.9272019537919207,
260
+ 0.9869026883377432
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
+ 0.0006158212892280167,
276
+ -0.0051020386117644435,
277
+ 0.0021842735483321033
278
+ ],
279
+ "std": [
280
+ 0.10053352790521747,
281
+ 0.11334779165555971,
282
+ 0.11122198562442343
283
+ ],
284
+ "q01": [
285
+ -0.32635084054792424,
286
+ -0.3907612560919037,
287
+ -0.34682255251091093
288
+ ],
289
+ "q99": [
290
+ 0.3093633216398961,
291
+ 0.3267943366279165,
292
+ 0.38177169080650525
293
+ ]
294
+ },
295
+ "gripper_close": {
296
+ "max": [
297
+ 1.0
298
+ ],
299
+ "min": [
300
+ -1.0
301
+ ],
302
+ "mean": [
303
+ -0.07410798548524393
304
+ ],
305
+ "std": [
306
+ 0.99722827635026
307
+ ],
308
+ "q01": [
309
+ -1.0
310
+ ],
311
+ "q99": [
312
+ 1.0
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
+ }
gr00t_n1-5/target_fraction/composite_seen_10p/model-00001-of-00002.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:72522c4c5cd1b7d286ba329a84637aa8d673a55d338ddfa19633a89f7a97c487
3
+ size 4999367032
gr00t_n1-5/target_fraction/composite_seen_10p/model-00002-of-00002.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:25849051c398bf7e444ff714553877261133fd1e50347207fe5b9bec3582895d
3
+ size 2586705312
gr00t_n1-5/target_fraction/composite_seen_10p/model.safetensors.index.json ADDED
The diff for this file is too large to render. See raw diff
 
gr00t_n1-5/target_fraction/composite_seen_10p/trainer_state.json ADDED
The diff for this file is too large to render. See raw diff
 
gr00t_n1-5/target_fraction/composite_seen_30p/README.md ADDED
@@ -0,0 +1,31 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # GR00T N1.5 — RoboCasa `composite_seen`, pretraining + post-training on 30% target data
2
+
3
+ Setting: **Pretraining + Target Post-Training (30%)**.
4
+
5
+ - Stage 1 (pretraining): GR00T N1.5 (`nvidia/GR00T-N1.5-3B`) trained on the RoboCasa
6
+ pretraining corpus (300 human demos/task + MimicGen), 80k steps.
7
+ - Stage 2 (post-training): this checkpoint — continued training on the `composite_seen` target
8
+ tasks using 150 of the 500 human demos per task (30%), 60k steps, batch size 128,
9
+ peak LR 3e-5 (cosine, 5% warmup), vision tower and LLM frozen.
10
+
11
+ Evaluated with 30 rollouts per task on the held-out `test` split.
12
+
13
+ | Task | Success rate (%) | Rollouts |
14
+ |---|---|---|
15
+ | DeliverStraw | 3.3 | 30 |
16
+ | GetToastedBread | 0.0 | 30 |
17
+ | KettleBoiling | 40.0 | 30 |
18
+ | LoadDishwasher | 53.3 | 30 |
19
+ | PackIdenticalLunches | 20.0 | 30 |
20
+ | PreSoakPan | 66.7 | 30 |
21
+ | PrepareCoffee | 20.0 | 30 |
22
+ | RinseSinkBasin | 36.7 | 30 |
23
+ | ScrubCuttingBoard | 46.7 | 30 |
24
+ | SearingMeat | 13.3 | 30 |
25
+ | SetUpCuttingStation | 30.0 | 30 |
26
+ | StackBowlsCabinet | 70.0 | 30 |
27
+ | SteamInMicrowave | 50.0 | 30 |
28
+ | StirVegetables | 13.3 | 30 |
29
+ | StoreLeftoversInBowl | 63.3 | 30 |
30
+ | WashLettuce | 26.7 | 30 |
31
+ | **Average** | **34.6** | |
gr00t_n1-5/target_fraction/composite_seen_30p/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
+ }
gr00t_n1-5/target_fraction/composite_seen_30p/eval_results.json ADDED
@@ -0,0 +1,66 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "DeliverStraw": {
3
+ "num_episodes": 30,
4
+ "success_rate": 0.03333333333333333
5
+ },
6
+ "GetToastedBread": {
7
+ "num_episodes": 30,
8
+ "success_rate": 0.0
9
+ },
10
+ "KettleBoiling": {
11
+ "num_episodes": 30,
12
+ "success_rate": 0.4
13
+ },
14
+ "LoadDishwasher": {
15
+ "num_episodes": 30,
16
+ "success_rate": 0.5333333333333333
17
+ },
18
+ "PackIdenticalLunches": {
19
+ "num_episodes": 30,
20
+ "success_rate": 0.2
21
+ },
22
+ "PreSoakPan": {
23
+ "num_episodes": 30,
24
+ "success_rate": 0.6666666666666666
25
+ },
26
+ "PrepareCoffee": {
27
+ "num_episodes": 30,
28
+ "success_rate": 0.2
29
+ },
30
+ "RinseSinkBasin": {
31
+ "num_episodes": 30,
32
+ "success_rate": 0.36666666666666664
33
+ },
34
+ "ScrubCuttingBoard": {
35
+ "num_episodes": 30,
36
+ "success_rate": 0.4666666666666667
37
+ },
38
+ "SearingMeat": {
39
+ "num_episodes": 30,
40
+ "success_rate": 0.13333333333333333
41
+ },
42
+ "SetUpCuttingStation": {
43
+ "num_episodes": 30,
44
+ "success_rate": 0.3
45
+ },
46
+ "StackBowlsCabinet": {
47
+ "num_episodes": 30,
48
+ "success_rate": 0.7
49
+ },
50
+ "SteamInMicrowave": {
51
+ "num_episodes": 30,
52
+ "success_rate": 0.5
53
+ },
54
+ "StirVegetables": {
55
+ "num_episodes": 30,
56
+ "success_rate": 0.13333333333333333
57
+ },
58
+ "StoreLeftoversInBowl": {
59
+ "num_episodes": 30,
60
+ "success_rate": 0.6333333333333333
61
+ },
62
+ "WashLettuce": {
63
+ "num_episodes": 30,
64
+ "success_rate": 0.26666666666666666
65
+ }
66
+ }
gr00t_n1-5/target_fraction/composite_seen_30p/experiment_cfg/metadata.json ADDED
@@ -0,0 +1,431 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "new_embodiment": {
3
+ "statistics": {
4
+ "state": {
5
+ "base_position": {
6
+ "max": [
7
+ 5.65041971206665,
8
+ -0.2986648678779602,
9
+ 0.7586075663566589
10
+ ],
11
+ "min": [
12
+ -0.43130701780319214,
13
+ -5.801346778869629,
14
+ 0.6994776725769043
15
+ ],
16
+ "mean": [
17
+ 2.692740579874402,
18
+ -1.8382568700240314,
19
+ 0.7010874652342763
20
+ ],
21
+ "std": [
22
+ 1.509974703450701,
23
+ 1.050997470898509,
24
+ 0.003967231599040678
25
+ ],
26
+ "q01": [
27
+ 0.6141094540013862,
28
+ -3.662778771057285,
29
+ 0.6999907374355346
30
+ ],
31
+ "q99": [
32
+ 5.286612962829395,
33
+ -0.7463194474541274,
34
+ 0.7084921806473795
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
+ 0.24889863961966005,
54
+ 0.6196654326307379
55
+ ],
56
+ "std": [
57
+ 0.0,
58
+ 0.0,
59
+ 0.6574404964156503,
60
+ 0.34926861676666165
61
+ ],
62
+ "q01": [
63
+ 0.0,
64
+ 0.0,
65
+ -0.9994904149324124,
66
+ 0.0017963339488804642
67
+ ],
68
+ "q99": [
69
+ 0.0,
70
+ 0.0,
71
+ 0.9997086026088378,
72
+ 0.9998873740738382
73
+ ]
74
+ },
75
+ "end_effector_position_relative": {
76
+ "max": [
77
+ 0.8965585827827454,
78
+ 0.8544884920120239,
79
+ 1.0672767162322998
80
+ ],
81
+ "min": [
82
+ -0.5257036089897156,
83
+ -0.8542829751968384,
84
+ -0.30580952763557434
85
+ ],
86
+ "mean": [
87
+ 0.30693297772466793,
88
+ -0.015591254528777701,
89
+ 0.4355198098682602
90
+ ],
91
+ "std": [
92
+ 0.1835422479701844,
93
+ 0.23451611970242953,
94
+ 0.19171566434572784
95
+ ],
96
+ "q01": [
97
+ -0.026768881775287107,
98
+ -0.5040839778252101,
99
+ 0.1410761818433882
100
+ ],
101
+ "q99": [
102
+ 0.6788473862636837,
103
+ 0.4954540983202727,
104
+ 0.7264870756054683
105
+ ]
106
+ },
107
+ "end_effector_rotation_relative": {
108
+ "max": [
109
+ 0.9999983906745911,
110
+ 0.9999905824661255,
111
+ 0.9724889397621155,
112
+ 0.9802296757698059
113
+ ],
114
+ "min": [
115
+ -0.9999982118606567,
116
+ -0.9999850392341614,
117
+ -0.9627900123596191,
118
+ 2.564086543088706e-08
119
+ ],
120
+ "mean": [
121
+ -0.1807256070855625,
122
+ 0.06878253824905319,
123
+ -0.053266011984745225,
124
+ 0.13332345927306935
125
+ ],
126
+ "std": [
127
+ 0.8415034506402461,
128
+ 0.3449928620024532,
129
+ 0.2982919853801513,
130
+ 0.16080691291557447
131
+ ],
132
+ "q01": [
133
+ -0.9964254306631649,
134
+ -0.6630621863344923,
135
+ -0.6105533846295836,
136
+ 0.0014010741177523966
137
+ ],
138
+ "q99": [
139
+ 0.9926395302942274,
140
+ 0.8049622183854135,
141
+ 0.5619022549322382,
142
+ 0.5482749558880121
143
+ ]
144
+ },
145
+ "gripper_qpos": {
146
+ "max": [
147
+ 0.05964253097772598,
148
+ 0.0066233607940375805
149
+ ],
150
+ "min": [
151
+ -0.019988659769296646,
152
+ -0.06823232024908066
153
+ ],
154
+ "mean": [
155
+ 0.028536247599882117,
156
+ -0.028538968763708122
157
+ ],
158
+ "std": [
159
+ 0.013369226362542433,
160
+ 0.013433203127554975
161
+ ],
162
+ "q01": [
163
+ 0.00396260754716939,
164
+ -0.04043649033701297
165
+ ],
166
+ "q99": [
167
+ 0.040364050884200135,
168
+ -0.00420067820363677
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
+ 0.03610723676028634,
188
+ 0.0019463876676260835,
189
+ -0.0012300652232849223,
190
+ 0.0
191
+ ],
192
+ "std": [
193
+ 0.24096715052127174,
194
+ 0.23606465039001745,
195
+ 0.14113148767178124,
196
+ 0.0
197
+ ],
198
+ "q01": [
199
+ -0.4636808668188184,
200
+ -0.6425051025441084,
201
+ -0.4201780727327269,
202
+ 0.0
203
+ ],
204
+ "q99": [
205
+ 0.5979685443969391,
206
+ 0.5788799557052615,
207
+ 0.39660424442217307,
208
+ 0.0
209
+ ]
210
+ },
211
+ "control_mode": {
212
+ "max": [
213
+ 1.0
214
+ ],
215
+ "min": [
216
+ -1.0
217
+ ],
218
+ "mean": [
219
+ -0.6965170410483006
220
+ ],
221
+ "std": [
222
+ 0.7176023621441738
223
+ ],
224
+ "q01": [
225
+ -1.0
226
+ ],
227
+ "q99": [
228
+ 0.6350037095647443
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
+ 0.012042026656403164,
244
+ -0.010033269630020658,
245
+ -0.02810790116779667
246
+ ],
247
+ "std": [
248
+ 0.41774134301967925,
249
+ 0.35992703282962496,
250
+ 0.37663580810255404
251
+ ],
252
+ "q01": [
253
+ -0.9919896001643727,
254
+ -0.9271079157646638,
255
+ -0.8890412252517207
256
+ ],
257
+ "q99": [
258
+ 0.9016773039641756,
259
+ 0.9265291757422638,
260
+ 0.9877088574541197
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
+ 0.0006487467792736983,
276
+ -0.005188974158750088,
277
+ 0.0022346536317766055
278
+ ],
279
+ "std": [
280
+ 0.10056504185488277,
281
+ 0.11340120102082349,
282
+ 0.11128870879600825
283
+ ],
284
+ "q01": [
285
+ -0.3262168861057055,
286
+ -0.3916413240513208,
287
+ -0.3471981253740904
288
+ ],
289
+ "q99": [
290
+ 0.3094441076447754,
291
+ 0.3267791042761232,
292
+ 0.38244333629933464
293
+ ]
294
+ },
295
+ "gripper_close": {
296
+ "max": [
297
+ 1.0
298
+ ],
299
+ "min": [
300
+ -1.0
301
+ ],
302
+ "mean": [
303
+ -0.0739909993221065
304
+ ],
305
+ "std": [
306
+ 0.9972351222914387
307
+ ],
308
+ "q01": [
309
+ -1.0
310
+ ],
311
+ "q99": [
312
+ 1.0
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
+ }
gr00t_n1-5/target_fraction/composite_seen_30p/model-00001-of-00002.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:73b58aedb2fa6181d94d1a6d8c5d2ca0e7f6f09e195df3cd2f0fcdb69fbaa43a
3
+ size 4999367032
gr00t_n1-5/target_fraction/composite_seen_30p/model-00002-of-00002.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:fe45c186c987d0eb050274b5947055d07e776f46c4bfa6de9b98179387e64d3a
3
+ size 2586705312
gr00t_n1-5/target_fraction/composite_seen_30p/model.safetensors.index.json ADDED
The diff for this file is too large to render. See raw diff
 
gr00t_n1-5/target_fraction/composite_seen_30p/trainer_state.json ADDED
The diff for this file is too large to render. See raw diff
 
gr00t_n1-5/target_fraction/composite_unseen_10p/README.md ADDED
@@ -0,0 +1,31 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # GR00T N1.5 — RoboCasa `composite_unseen`, pretraining + post-training on 10% target data
2
+
3
+ Setting: **Pretraining + Target Post-Training (10%)**.
4
+
5
+ - Stage 1 (pretraining): GR00T N1.5 (`nvidia/GR00T-N1.5-3B`) trained on the RoboCasa
6
+ pretraining corpus (300 human demos/task + MimicGen), 80k steps.
7
+ - Stage 2 (post-training): this checkpoint — continued training on the `composite_unseen` target
8
+ tasks using 50 of the 500 human demos per task (10%), 60k steps, batch size 128,
9
+ peak LR 3e-5 (cosine, 5% warmup), vision tower and LLM frozen.
10
+
11
+ Evaluated with 30 rollouts per task on the held-out `test` split.
12
+
13
+ | Task | Success rate (%) | Rollouts |
14
+ |---|---|---|
15
+ | ArrangeBreadBasket | 60.0 | 30 |
16
+ | ArrangeTea | 13.3 | 30 |
17
+ | BreadSelection | 36.7 | 30 |
18
+ | CategorizeCondiments | 3.3 | 30 |
19
+ | CuttingToolSelection | 23.3 | 30 |
20
+ | GarnishPancake | 60.0 | 30 |
21
+ | GatherTableware | 10.0 | 30 |
22
+ | HeatKebabSandwich | 3.3 | 30 |
23
+ | MakeIceLemonade | 26.7 | 30 |
24
+ | PanTransfer | 3.3 | 30 |
25
+ | PortionHotDogs | 3.3 | 30 |
26
+ | RecycleBottlesByType | 10.0 | 30 |
27
+ | SeparateFreezerRack | 10.0 | 30 |
28
+ | WaffleReheat | 53.3 | 30 |
29
+ | WashFruitColander | 13.3 | 30 |
30
+ | WeighIngredients | 33.3 | 30 |
31
+ | **Average** | **22.7** | |
gr00t_n1-5/target_fraction/composite_unseen_10p/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
+ }
gr00t_n1-5/target_fraction/composite_unseen_10p/eval_results.json ADDED
@@ -0,0 +1,94 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "ArrangeBreadBasket": {
3
+ "num_episodes": 30,
4
+ "success_rate": 0.6
5
+ },
6
+ "ArrangeTea": {
7
+ "num_episodes": 30,
8
+ "success_rate": 0.13333333333333333
9
+ },
10
+ "BreadSelection": {
11
+ "num_episodes": 30,
12
+ "success_rate": 0.36666666666666664
13
+ },
14
+ "CategorizeCondiments": {
15
+ "num_episodes": 30,
16
+ "success_rate": 0.03333333333333333
17
+ },
18
+ "CuttingToolSelection": {
19
+ "num_episodes": 30,
20
+ "success_rate": 0.23333333333333334
21
+ },
22
+ "DeliverStraw": {
23
+ "num_episodes": 30,
24
+ "success_rate": 0.0
25
+ },
26
+ "GarnishPancake": {
27
+ "num_episodes": 30,
28
+ "success_rate": 0.6
29
+ },
30
+ "GatherTableware": {
31
+ "num_episodes": 30,
32
+ "success_rate": 0.1
33
+ },
34
+ "GetToastedBread": {
35
+ "num_episodes": 30,
36
+ "success_rate": 0.0
37
+ },
38
+ "HeatKebabSandwich": {
39
+ "num_episodes": 30,
40
+ "success_rate": 0.03333333333333333
41
+ },
42
+ "KettleBoiling": {
43
+ "num_episodes": 30,
44
+ "success_rate": 0.0
45
+ },
46
+ "LoadDishwasher": {
47
+ "num_episodes": 30,
48
+ "success_rate": 0.0
49
+ },
50
+ "MakeIceLemonade": {
51
+ "num_episodes": 30,
52
+ "success_rate": 0.26666666666666666
53
+ },
54
+ "PackIdenticalLunches": {
55
+ "num_episodes": 30,
56
+ "success_rate": 0.0
57
+ },
58
+ "PanTransfer": {
59
+ "num_episodes": 30,
60
+ "success_rate": 0.03333333333333333
61
+ },
62
+ "PortionHotDogs": {
63
+ "num_episodes": 30,
64
+ "success_rate": 0.03333333333333333
65
+ },
66
+ "PreSoakPan": {
67
+ "num_episodes": 30,
68
+ "success_rate": 0.0
69
+ },
70
+ "PrepareCoffee": {
71
+ "num_episodes": 30,
72
+ "success_rate": 0.0
73
+ },
74
+ "RecycleBottlesByType": {
75
+ "num_episodes": 30,
76
+ "success_rate": 0.1
77
+ },
78
+ "SeparateFreezerRack": {
79
+ "num_episodes": 30,
80
+ "success_rate": 0.1
81
+ },
82
+ "WaffleReheat": {
83
+ "num_episodes": 30,
84
+ "success_rate": 0.5333333333333333
85
+ },
86
+ "WashFruitColander": {
87
+ "num_episodes": 30,
88
+ "success_rate": 0.13333333333333333
89
+ },
90
+ "WeighIngredients": {
91
+ "num_episodes": 30,
92
+ "success_rate": 0.3333333333333333
93
+ }
94
+ }
gr00t_n1-5/target_fraction/composite_unseen_10p/experiment_cfg/metadata.json ADDED
@@ -0,0 +1,431 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "new_embodiment": {
3
+ "statistics": {
4
+ "state": {
5
+ "base_position": {
6
+ "max": [
7
+ 5.650563716888428,
8
+ -0.2951663136482239,
9
+ 0.7298293113708496
10
+ ],
11
+ "min": [
12
+ -0.38555988669395447,
13
+ -5.501502513885498,
14
+ 0.6992281079292297
15
+ ],
16
+ "mean": [
17
+ 2.6336262536806494,
18
+ -1.9054981430884423,
19
+ 0.7007931980989522
20
+ ],
21
+ "std": [
22
+ 1.5284093706836843,
23
+ 1.1743208021371767,
24
+ 0.0030475272238204235
25
+ ],
26
+ "q01": [
27
+ 0.6220077041901974,
28
+ -4.049521694062227,
29
+ 0.6999838248370592
30
+ ],
31
+ "q99": [
32
+ 5.136307403294112,
33
+ -0.9477999695671174,
34
+ 0.706538721581072
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
+ 0.316881847044985,
54
+ 0.6310692992705585
55
+ ],
56
+ "std": [
57
+ 0.0,
58
+ 0.0,
59
+ 0.6251992332522089,
60
+ 0.3325076229026489
61
+ ],
62
+ "q01": [
63
+ 0.0,
64
+ 0.0,
65
+ -0.9833443356265679,
66
+ 0.003905101324083478
67
+ ],
68
+ "q99": [
69
+ 0.0,
70
+ 0.0,
71
+ 0.9997866644218206,
72
+ 0.9835663784429151
73
+ ]
74
+ },
75
+ "end_effector_position_relative": {
76
+ "max": [
77
+ 0.8826583027839661,
78
+ 0.8520669341087341,
79
+ 1.050371766090393
80
+ ],
81
+ "min": [
82
+ -0.5266212224960327,
83
+ -0.846970796585083,
84
+ 0.004259257111698389
85
+ ],
86
+ "mean": [
87
+ 0.2690251605600002,
88
+ -0.023420017035223662,
89
+ 0.5242273728357386
90
+ ],
91
+ "std": [
92
+ 0.18356831958627004,
93
+ 0.23203504501587371,
94
+ 0.18447536255463537
95
+ ],
96
+ "q01": [
97
+ -0.06294697921152705,
98
+ -0.4925957533685401,
99
+ 0.26524910982703237
100
+ ],
101
+ "q99": [
102
+ 0.6225278348569468,
103
+ 0.4684270740415465,
104
+ 0.7624899680348455
105
+ ]
106
+ },
107
+ "end_effector_rotation_relative": {
108
+ "max": [
109
+ 0.9999992847442627,
110
+ 0.9999575614929199,
111
+ 0.999175488948822,
112
+ 0.9986861944198608
113
+ ],
114
+ "min": [
115
+ -0.9999995231628418,
116
+ -0.9999401569366455,
117
+ -0.9994947910308838,
118
+ 1.0679671369473454e-08
119
+ ],
120
+ "mean": [
121
+ -0.2167419651637259,
122
+ 0.060487081190059654,
123
+ -0.062320573017817744,
124
+ 0.13341569272510032
125
+ ],
126
+ "std": [
127
+ 0.8194752517175329,
128
+ 0.3327863495321197,
129
+ 0.3464280286005768,
130
+ 0.15903985866929315
131
+ ],
132
+ "q01": [
133
+ -0.9952245292680625,
134
+ -0.6251701970824366,
135
+ -0.6320119147930746,
136
+ 0.0012923678921187137
137
+ ],
138
+ "q99": [
139
+ 0.9892267303829377,
140
+ 0.7308262819163254,
141
+ 0.586981749797333,
142
+ 0.576234793324005
143
+ ]
144
+ },
145
+ "gripper_qpos": {
146
+ "max": [
147
+ 0.07373982667922974,
148
+ 0.009727069176733494
149
+ ],
150
+ "min": [
151
+ -0.0086278785020113,
152
+ -0.09789291024208069
153
+ ],
154
+ "mean": [
155
+ 0.029400387847344143,
156
+ -0.029440330405669023
157
+ ],
158
+ "std": [
159
+ 0.013287871883843788,
160
+ 0.013322406142863369
161
+ ],
162
+ "q01": [
163
+ 0.004163312093880173,
164
+ -0.04039848146102206
165
+ ],
166
+ "q99": [
167
+ 0.04025154108259885,
168
+ -0.0043862213744777
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
+ 0.024136075867151503,
188
+ 0.0022696116416857036,
189
+ 0.00032668872635757825,
190
+ 0.0
191
+ ],
192
+ "std": [
193
+ 0.193508441081677,
194
+ 0.1710160316710299,
195
+ 0.13250778077371247,
196
+ 0.0
197
+ ],
198
+ "q01": [
199
+ -0.3957412160193452,
200
+ -0.4484857010107363,
201
+ -0.28722809293816515,
202
+ 0.0
203
+ ],
204
+ "q99": [
205
+ 0.4547273422615313,
206
+ 0.4527358800196218,
207
+ 0.3134420507916789,
208
+ 0.0
209
+ ]
210
+ },
211
+ "control_mode": {
212
+ "max": [
213
+ 1.0
214
+ ],
215
+ "min": [
216
+ -1.0
217
+ ],
218
+ "mean": [
219
+ -0.7679086773594997
220
+ ],
221
+ "std": [
222
+ 0.6404993324091306
223
+ ],
224
+ "q01": [
225
+ -1.0
226
+ ],
227
+ "q99": [
228
+ 0.35797764034706575
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
+ 0.011600842301825853,
244
+ -0.005896718477673545,
245
+ -0.010023766594070147
246
+ ],
247
+ "std": [
248
+ 0.41528609378435527,
249
+ 0.3757614449872329,
250
+ 0.32533404876452887
251
+ ],
252
+ "q01": [
253
+ -0.9897175609012556,
254
+ -0.9476974506479521,
255
+ -0.8021519238433477
256
+ ],
257
+ "q99": [
258
+ 0.9075523886180787,
259
+ 0.9401430230351486,
260
+ 0.9627511816695983
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
+ 0.00269485670172446,
276
+ -0.008619524461325209,
277
+ -0.0016058665973128076
278
+ ],
279
+ "std": [
280
+ 0.0996712919862095,
281
+ 0.11186835545020272,
282
+ 0.12064238150340853
283
+ ],
284
+ "q01": [
285
+ -0.31622562755187317,
286
+ -0.37984201508359744,
287
+ -0.39186986092810866
288
+ ],
289
+ "q99": [
290
+ 0.3366098317483241,
291
+ 0.31964602858511193,
292
+ 0.40786028720238415
293
+ ]
294
+ },
295
+ "gripper_close": {
296
+ "max": [
297
+ 1.0
298
+ ],
299
+ "min": [
300
+ -1.0
301
+ ],
302
+ "mean": [
303
+ -0.16924234569166097
304
+ ],
305
+ "std": [
306
+ 0.9864073033291665
307
+ ],
308
+ "q01": [
309
+ -1.0
310
+ ],
311
+ "q99": [
312
+ 1.0
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
+ }
gr00t_n1-5/target_fraction/composite_unseen_10p/model-00001-of-00002.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5d40eae26160cdbaa54a7cbc84b06227273a354d0524e334e4cb268a675e92b3
3
+ size 4999367032
gr00t_n1-5/target_fraction/composite_unseen_10p/model-00002-of-00002.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:bcf3c9fa5d01e4dab682f56f82806e27916f5890573b88fc1098c1a34438724c
3
+ size 2586705312
gr00t_n1-5/target_fraction/composite_unseen_10p/model.safetensors.index.json ADDED
The diff for this file is too large to render. See raw diff
 
gr00t_n1-5/target_fraction/composite_unseen_10p/trainer_state.json ADDED
The diff for this file is too large to render. See raw diff
 
gr00t_n1-5/target_fraction/composite_unseen_30p/README.md ADDED
@@ -0,0 +1,31 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # GR00T N1.5 — RoboCasa `composite_unseen`, pretraining + post-training on 30% target data
2
+
3
+ Setting: **Pretraining + Target Post-Training (30%)**.
4
+
5
+ - Stage 1 (pretraining): GR00T N1.5 (`nvidia/GR00T-N1.5-3B`) trained on the RoboCasa
6
+ pretraining corpus (300 human demos/task + MimicGen), 80k steps.
7
+ - Stage 2 (post-training): this checkpoint — continued training on the `composite_unseen` target
8
+ tasks using 150 of the 500 human demos per task (30%), 60k steps, batch size 128,
9
+ peak LR 3e-5 (cosine, 5% warmup), vision tower and LLM frozen.
10
+
11
+ Evaluated with 30 rollouts per task on the held-out `test` split.
12
+
13
+ | Task | Success rate (%) | Rollouts |
14
+ |---|---|---|
15
+ | ArrangeBreadBasket | 63.3 | 30 |
16
+ | ArrangeTea | 23.3 | 30 |
17
+ | BreadSelection | 46.7 | 30 |
18
+ | CategorizeCondiments | 6.7 | 30 |
19
+ | CuttingToolSelection | 36.7 | 30 |
20
+ | GarnishPancake | 43.3 | 30 |
21
+ | GatherTableware | 20.0 | 30 |
22
+ | HeatKebabSandwich | 10.0 | 30 |
23
+ | MakeIceLemonade | 26.7 | 30 |
24
+ | PanTransfer | 10.0 | 30 |
25
+ | PortionHotDogs | 0.0 | 30 |
26
+ | RecycleBottlesByType | 63.3 | 30 |
27
+ | SeparateFreezerRack | 20.0 | 30 |
28
+ | WaffleReheat | 70.0 | 30 |
29
+ | WashFruitColander | 20.0 | 30 |
30
+ | WeighIngredients | 33.3 | 30 |
31
+ | **Average** | **30.8** | |
gr00t_n1-5/target_fraction/composite_unseen_30p/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
+ }
gr00t_n1-5/target_fraction/composite_unseen_30p/eval_results.json ADDED
@@ -0,0 +1,90 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "ArrangeBreadBasket": {
3
+ "num_episodes": 30,
4
+ "success_rate": 0.6333333333333333
5
+ },
6
+ "ArrangeTea": {
7
+ "num_episodes": 30,
8
+ "success_rate": 0.23333333333333334
9
+ },
10
+ "BreadSelection": {
11
+ "num_episodes": 30,
12
+ "success_rate": 0.4666666666666667
13
+ },
14
+ "CategorizeCondiments": {
15
+ "num_episodes": 30,
16
+ "success_rate": 0.06666666666666667
17
+ },
18
+ "CuttingToolSelection": {
19
+ "num_episodes": 30,
20
+ "success_rate": 0.36666666666666664
21
+ },
22
+ "DeliverStraw": {
23
+ "num_episodes": 30,
24
+ "success_rate": 0.0
25
+ },
26
+ "GarnishPancake": {
27
+ "num_episodes": 30,
28
+ "success_rate": 0.43333333333333335
29
+ },
30
+ "GatherTableware": {
31
+ "num_episodes": 30,
32
+ "success_rate": 0.2
33
+ },
34
+ "GetToastedBread": {
35
+ "num_episodes": 30,
36
+ "success_rate": 0.0
37
+ },
38
+ "HeatKebabSandwich": {
39
+ "num_episodes": 30,
40
+ "success_rate": 0.1
41
+ },
42
+ "KettleBoiling": {
43
+ "num_episodes": 30,
44
+ "success_rate": 0.0
45
+ },
46
+ "LoadDishwasher": {
47
+ "num_episodes": 30,
48
+ "success_rate": 0.0
49
+ },
50
+ "MakeIceLemonade": {
51
+ "num_episodes": 30,
52
+ "success_rate": 0.26666666666666666
53
+ },
54
+ "PackIdenticalLunches": {
55
+ "num_episodes": 30,
56
+ "success_rate": 0.0
57
+ },
58
+ "PanTransfer": {
59
+ "num_episodes": 30,
60
+ "success_rate": 0.1
61
+ },
62
+ "PortionHotDogs": {
63
+ "num_episodes": 30,
64
+ "success_rate": 0.0
65
+ },
66
+ "PreSoakPan": {
67
+ "num_episodes": 30,
68
+ "success_rate": 0.1
69
+ },
70
+ "RecycleBottlesByType": {
71
+ "num_episodes": 30,
72
+ "success_rate": 0.6333333333333333
73
+ },
74
+ "SeparateFreezerRack": {
75
+ "num_episodes": 30,
76
+ "success_rate": 0.2
77
+ },
78
+ "WaffleReheat": {
79
+ "num_episodes": 30,
80
+ "success_rate": 0.7
81
+ },
82
+ "WashFruitColander": {
83
+ "num_episodes": 30,
84
+ "success_rate": 0.2
85
+ },
86
+ "WeighIngredients": {
87
+ "num_episodes": 30,
88
+ "success_rate": 0.3333333333333333
89
+ }
90
+ }
gr00t_n1-5/target_fraction/composite_unseen_30p/experiment_cfg/metadata.json ADDED
@@ -0,0 +1,431 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "new_embodiment": {
3
+ "statistics": {
4
+ "state": {
5
+ "base_position": {
6
+ "max": [
7
+ 5.650563716888428,
8
+ -0.2951663136482239,
9
+ 0.7298293113708496
10
+ ],
11
+ "min": [
12
+ -0.38555988669395447,
13
+ -5.501502513885498,
14
+ 0.6992281079292297
15
+ ],
16
+ "mean": [
17
+ 2.6334457924708787,
18
+ -1.907420854658446,
19
+ 0.7007862693259018
20
+ ],
21
+ "std": [
22
+ 1.5136653737945276,
23
+ 1.178582783725861,
24
+ 0.003036676712013605
25
+ ],
26
+ "q01": [
27
+ 0.6364016741080525,
28
+ -4.065610123736803,
29
+ 0.6999835925583867
30
+ ],
31
+ "q99": [
32
+ 5.131142461622728,
33
+ -0.9491482602535835,
34
+ 0.70646103398464
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
+ 0.31603188373540975,
54
+ 0.6300760444020418
55
+ ],
56
+ "std": [
57
+ 0.0,
58
+ 0.0,
59
+ 0.6261545822990109,
60
+ 0.33338669562419837
61
+ ],
62
+ "q01": [
63
+ 0.0,
64
+ 0.0,
65
+ -0.9859275692820585,
66
+ 0.0033289505710110343
67
+ ],
68
+ "q99": [
69
+ 0.0,
70
+ 0.0,
71
+ 0.999819560717181,
72
+ 0.982821686790793
73
+ ]
74
+ },
75
+ "end_effector_position_relative": {
76
+ "max": [
77
+ 0.8826583027839661,
78
+ 0.8520669341087341,
79
+ 1.050371766090393
80
+ ],
81
+ "min": [
82
+ -0.5266212224960327,
83
+ -0.846970796585083,
84
+ 0.004259257111698389
85
+ ],
86
+ "mean": [
87
+ 0.27017802679541536,
88
+ -0.023170603663626402,
89
+ 0.5261694888843391
90
+ ],
91
+ "std": [
92
+ 0.18327102374303522,
93
+ 0.23291469248937874,
94
+ 0.1855804155922538
95
+ ],
96
+ "q01": [
97
+ -0.06287757595596973,
98
+ -0.4940651139437684,
99
+ 0.2671180877516492
100
+ ],
101
+ "q99": [
102
+ 0.6218416170457015,
103
+ 0.4697029890124258,
104
+ 0.7643181485925548
105
+ ]
106
+ },
107
+ "end_effector_rotation_relative": {
108
+ "max": [
109
+ 0.9999992847442627,
110
+ 0.9999575614929199,
111
+ 0.999175488948822,
112
+ 0.9986861944198608
113
+ ],
114
+ "min": [
115
+ -0.9999995231628418,
116
+ -0.9999401569366455,
117
+ -0.9994947910308838,
118
+ 1.0679671369473454e-08
119
+ ],
120
+ "mean": [
121
+ -0.21398363077274452,
122
+ 0.05321420775847886,
123
+ -0.06266848966137746,
124
+ 0.13403865831638706
125
+ ],
126
+ "std": [
127
+ 0.8200769039900443,
128
+ 0.32950821985976114,
129
+ 0.35031410289719067,
130
+ 0.1599271177409951
131
+ ],
132
+ "q01": [
133
+ -0.9951612991473958,
134
+ -0.6238077475248331,
135
+ -0.6375534703750526,
136
+ 0.001286658371664053
137
+ ],
138
+ "q99": [
139
+ 0.9890265403590498,
140
+ 0.7256812211515437,
141
+ 0.5937754394576653,
142
+ 0.5783405407317761
143
+ ]
144
+ },
145
+ "gripper_qpos": {
146
+ "max": [
147
+ 0.07373982667922974,
148
+ 0.009727069176733494
149
+ ],
150
+ "min": [
151
+ -0.0086278785020113,
152
+ -0.09789291024208069
153
+ ],
154
+ "mean": [
155
+ 0.02950392852300861,
156
+ -0.029553632597886837
157
+ ],
158
+ "std": [
159
+ 0.013224631401662275,
160
+ 0.01324540339640474
161
+ ],
162
+ "q01": [
163
+ 0.004241727591071626,
164
+ -0.04039806619769442
165
+ ],
166
+ "q99": [
167
+ 0.04025287763029444,
168
+ -0.004489349520789004
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
+ 0.022870880016881655,
188
+ 0.0020983593339277094,
189
+ 0.0001707236413505001,
190
+ 0.0
191
+ ],
192
+ "std": [
193
+ 0.19064277194002152,
194
+ 0.16855241355529363,
195
+ 0.1290883846671232,
196
+ 0.0
197
+ ],
198
+ "q01": [
199
+ -0.38533596237442375,
200
+ -0.4363339796755096,
201
+ -0.2777101663088471,
202
+ 0.0
203
+ ],
204
+ "q99": [
205
+ 0.44333399362408077,
206
+ 0.44046784192563015,
207
+ 0.30166437224904386,
208
+ 0.0
209
+ ]
210
+ },
211
+ "control_mode": {
212
+ "max": [
213
+ 1.0
214
+ ],
215
+ "min": [
216
+ -1.0
217
+ ],
218
+ "mean": [
219
+ -0.7759901828249384
220
+ ],
221
+ "std": [
222
+ 0.6306945978462276
223
+ ],
224
+ "q01": [
225
+ -1.0
226
+ ],
227
+ "q99": [
228
+ 0.3375137629764175
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
+ 0.0115328440084565,
244
+ -0.005925575822350964,
245
+ -0.010129008011274209
246
+ ],
247
+ "std": [
248
+ 0.4154402288235015,
249
+ 0.37686469662007266,
250
+ 0.32541940055791746
251
+ ],
252
+ "q01": [
253
+ -0.989521859029367,
254
+ -0.9474992902095497,
255
+ -0.8022171724595398
256
+ ],
257
+ "q99": [
258
+ 0.9053201610858321,
259
+ 0.9406872002734736,
260
+ 0.9612026824801992
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
+ 0.002659859602018507,
276
+ -0.008822131613286726,
277
+ -0.0014573283663866898
278
+ ],
279
+ "std": [
280
+ 0.09993256683863617,
281
+ 0.1122848425446163,
282
+ 0.12025232422797603
283
+ ],
284
+ "q01": [
285
+ -0.3164788400832972,
286
+ -0.38128650849232165,
287
+ -0.38896859078461987
288
+ ],
289
+ "q99": [
290
+ 0.33674912765546977,
291
+ 0.32014255820318754,
292
+ 0.4057152384373908
293
+ ]
294
+ },
295
+ "gripper_close": {
296
+ "max": [
297
+ 1.0
298
+ ],
299
+ "min": [
300
+ -1.0
301
+ ],
302
+ "mean": [
303
+ -0.17317210312521217
304
+ ],
305
+ "std": [
306
+ 0.9857335812322223
307
+ ],
308
+ "q01": [
309
+ -1.0
310
+ ],
311
+ "q99": [
312
+ 1.0
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
+ }
gr00t_n1-5/target_fraction/composite_unseen_30p/model-00001-of-00002.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:53ca73c82c452d9fbb9aa62a2d4a7db47bb7b8dc220f4944aef7af300bd55ebb
3
+ size 4999367032
gr00t_n1-5/target_fraction/composite_unseen_30p/model-00002-of-00002.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e2dfd6c8d41b9fc78fcfad235f436ce03efbecb4abd6a332957e37552702f26f
3
+ size 2586705312
gr00t_n1-5/target_fraction/composite_unseen_30p/model.safetensors.index.json ADDED
The diff for this file is too large to render. See raw diff
 
gr00t_n1-5/target_fraction/composite_unseen_30p/trainer_state.json ADDED
The diff for this file is too large to render. See raw diff