TTaekwan commited on
Commit
39ab8cf
·
verified ·
1 Parent(s): 69535ba

s1v2_contact: final checkpoint (checkpoint-60000)

Browse files
s1v2_contact/config.json ADDED
@@ -0,0 +1,98 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "action_dim": 32,
3
+ "action_head_cfg": {
4
+ "action_dim": 32,
5
+ "action_horizon": 16,
6
+ "action_merge_reduction": "sum",
7
+ "add_pos_embed": true,
8
+ "backbone_embedding_dim": 2048,
9
+ "conf_loss_coef": 0.1,
10
+ "conf_threshold": 0.5,
11
+ "diffusion_model_cfg": {
12
+ "attention_head_dim": 48,
13
+ "cross_attention_dim": 2048,
14
+ "dropout": 0.2,
15
+ "final_dropout": true,
16
+ "interleave_self_attention": true,
17
+ "norm_type": "ada_norm",
18
+ "num_attention_heads": 32,
19
+ "num_layers": 16,
20
+ "output_dim": 1024,
21
+ "positional_embeddings": null
22
+ },
23
+ "discrete_action_dims": [
24
+ 6
25
+ ],
26
+ "hidden_size": 1024,
27
+ "input_embedding_dim": 1536,
28
+ "max_action_dim": 32,
29
+ "max_state_dim": 64,
30
+ "merged_4_weight": 1.0,
31
+ "merged_8_weight": 1.0,
32
+ "model_dtype": "float32",
33
+ "moe_2expert_use_native_8": false,
34
+ "moe_balance_weight": 0.05,
35
+ "moe_block_scheme": "r221",
36
+ "moe_body_mode": "per_expert_h",
37
+ "moe_expert_n_layers": 2,
38
+ "moe_inference_stochastic": false,
39
+ "moe_inference_temp": 0.7,
40
+ "moe_label_gated": true,
41
+ "moe_loss_ema_momentum": 0.99,
42
+ "moe_normalize_loss_for_kl": false,
43
+ "moe_num_experts": 4,
44
+ "moe_router_attn_pool": false,
45
+ "moe_router_attn_pool_heads": 4,
46
+ "moe_router_hidden": 256,
47
+ "moe_router_masked_pool": false,
48
+ "moe_router_temp": 0.5,
49
+ "moe_router_warmup_steps": 5000,
50
+ "moe_speed": 2.5,
51
+ "moe_supervise_weight": 0.1,
52
+ "moe_target_temp": 0.3,
53
+ "moe_uniform_router": false,
54
+ "native_8_weight": 1.0,
55
+ "noise_beta_alpha": 1.5,
56
+ "noise_beta_beta": 1.0,
57
+ "noise_s": 0.999,
58
+ "num_inference_timesteps": 4,
59
+ "num_target_vision_tokens": 32,
60
+ "num_timestep_buckets": 1000,
61
+ "tune_diffusion_model": true,
62
+ "tune_projector": true,
63
+ "use_merged_4_head": true,
64
+ "use_merged_8_head": true,
65
+ "use_moe_routing": true,
66
+ "use_native_8_head": true,
67
+ "use_vlln": true,
68
+ "vl_self_attention_cfg": {
69
+ "attention_head_dim": 64,
70
+ "dropout": 0.2,
71
+ "final_dropout": true,
72
+ "num_attention_heads": 32,
73
+ "num_layers": 4,
74
+ "positional_embeddings": null
75
+ }
76
+ },
77
+ "action_horizon": 16,
78
+ "architectures": [
79
+ "GR00T_N1_5"
80
+ ],
81
+ "attn_implementation": null,
82
+ "backbone_cfg": {
83
+ "eagle_path": "NVEagle/eagle_er-qwen3_1_7B-Siglip2_400M_stage1_5_128gpu_er_v7_1mlp_nops",
84
+ "load_bf16": false,
85
+ "project_to_dim": null,
86
+ "reproject_vision": false,
87
+ "select_layer": 12,
88
+ "tune_llm": false,
89
+ "tune_visual": true,
90
+ "use_flash_attention": true
91
+ },
92
+ "compute_dtype": "bfloat16",
93
+ "hidden_size": 2048,
94
+ "model_dtype": "float32",
95
+ "model_type": "gr00t_n1_5",
96
+ "torch_dtype": "bfloat16",
97
+ "transformers_version": "4.51.3"
98
+ }
s1v2_contact/experiment_cfg/metadata.json ADDED
@@ -0,0 +1,291 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "libero": {
3
+ "statistics": {
4
+ "state": {
5
+ "eef_pos_absolute": {
6
+ "max": [
7
+ 0.21031762659549713,
8
+ 0.39128610491752625,
9
+ 1.3660105466842651
10
+ ],
11
+ "min": [
12
+ -0.4828203022480011,
13
+ -0.3255046010017395,
14
+ 0.008128180168569088
15
+ ],
16
+ "mean": [
17
+ -0.046518828719854355,
18
+ 0.034408919513225555,
19
+ 0.7645694613456726
20
+ ],
21
+ "std": [
22
+ 0.10494082421064377,
23
+ 0.1517619788646698,
24
+ 0.3785194456577301
25
+ ],
26
+ "q01": [
27
+ -0.39912487387657164,
28
+ -0.26883512616157534,
29
+ 0.03826696485280991
30
+ ],
31
+ "q99": [
32
+ 0.13556506335735302,
33
+ 0.33566484451293943,
34
+ 1.2706659078598022
35
+ ]
36
+ },
37
+ "eef_rot_absolute": {
38
+ "max": [
39
+ 3.6714255809783936,
40
+ 3.560650587081909,
41
+ 1.386339545249939
42
+ ],
43
+ "min": [
44
+ 0.35277295112609863,
45
+ -3.641430377960205,
46
+ -1.842738389968872
47
+ ],
48
+ "mean": [
49
+ 2.9716713428497314,
50
+ -0.2204727977514267,
51
+ -0.12557993829250336
52
+ ],
53
+ "std": [
54
+ 0.3442671298980713,
55
+ 0.9068173766136169,
56
+ 0.32538288831710815
57
+ ],
58
+ "q01": [
59
+ 1.5089580535888671,
60
+ -2.719791135787964,
61
+ -1.08050856590271
62
+ ],
63
+ "q99": [
64
+ 3.2773460674285886,
65
+ 2.406109695434569,
66
+ 0.59776972055435
67
+ ]
68
+ },
69
+ "gripper_close": {
70
+ "max": [
71
+ 0.04233968257904053,
72
+ 0.0013633022317662835
73
+ ],
74
+ "min": [
75
+ -0.0013586411951109767,
76
+ -0.042040832340717316
77
+ ],
78
+ "mean": [
79
+ 0.026915358379483223,
80
+ -0.027191326022148132
81
+ ],
82
+ "std": [
83
+ 0.014175750315189362,
84
+ 0.014058776199817657
85
+ ],
86
+ "q01": [
87
+ 0.0017423711018636821,
88
+ -0.04002561450004578
89
+ ],
90
+ "q99": [
91
+ 0.04031316146254539,
92
+ -0.0017781084170565016
93
+ ]
94
+ }
95
+ },
96
+ "action": {
97
+ "eef_pos_delta": {
98
+ "max": [
99
+ 0.9375,
100
+ 0.9375,
101
+ 0.9375
102
+ ],
103
+ "min": [
104
+ -0.9375,
105
+ -0.9375,
106
+ -0.9375
107
+ ],
108
+ "mean": [
109
+ 0.06278152763843536,
110
+ 0.08684158325195312,
111
+ -0.0903734639286995
112
+ ],
113
+ "std": [
114
+ 0.33551836013793945,
115
+ 0.37847793102264404,
116
+ 0.4446770250797272
117
+ ],
118
+ "q01": [
119
+ -0.7044642567634583,
120
+ -0.8008928298950195,
121
+ -0.9375
122
+ ],
123
+ "q99": [
124
+ 0.9375,
125
+ 0.8678571581840515,
126
+ 0.9375
127
+ ]
128
+ },
129
+ "eef_rot_delta": {
130
+ "max": [
131
+ 0.3557142913341522,
132
+ 0.375,
133
+ 0.375
134
+ ],
135
+ "min": [
136
+ -0.2582142949104309,
137
+ -0.375,
138
+ -0.3675000071525574
139
+ ],
140
+ "mean": [
141
+ 0.0005407554563134909,
142
+ 0.005643464159220457,
143
+ -0.005229106638580561
144
+ ],
145
+ "std": [
146
+ 0.03924214467406273,
147
+ 0.06341660022735596,
148
+ 0.07792268693447113
149
+ ],
150
+ "q01": [
151
+ -0.11464285850524902,
152
+ -0.1639285683631897,
153
+ -0.2239285707473755
154
+ ],
155
+ "q99": [
156
+ 0.13178572058677673,
157
+ 0.19285714626312256,
158
+ 0.335357129573822
159
+ ]
160
+ },
161
+ "gripper_close": {
162
+ "max": [
163
+ 1.0
164
+ ],
165
+ "min": [
166
+ -1.0
167
+ ],
168
+ "mean": [
169
+ -0.0496407225728035
170
+ ],
171
+ "std": [
172
+ 1.000144362449646
173
+ ],
174
+ "q01": [
175
+ -1.0
176
+ ],
177
+ "q99": [
178
+ 1.0
179
+ ]
180
+ },
181
+ "ratio_label": {
182
+ "max": [
183
+ 1.0,
184
+ 1.0
185
+ ],
186
+ "min": [
187
+ 0.0,
188
+ 0.0
189
+ ],
190
+ "mean": [
191
+ 0.0,
192
+ 0.0
193
+ ],
194
+ "std": [
195
+ 1.0,
196
+ 1.0
197
+ ],
198
+ "q01": [
199
+ 0.0,
200
+ 0.0
201
+ ],
202
+ "q99": [
203
+ 1.0,
204
+ 1.0
205
+ ]
206
+ }
207
+ }
208
+ },
209
+ "modalities": {
210
+ "video": {
211
+ "front_view": {
212
+ "resolution": [
213
+ 256,
214
+ 256
215
+ ],
216
+ "channels": 3,
217
+ "fps": 10.0
218
+ },
219
+ "left_wrist_view": {
220
+ "resolution": [
221
+ 256,
222
+ 256
223
+ ],
224
+ "channels": 3,
225
+ "fps": 10.0
226
+ }
227
+ },
228
+ "state": {
229
+ "eef_pos_absolute": {
230
+ "absolute": true,
231
+ "rotation_type": null,
232
+ "shape": [
233
+ 3
234
+ ],
235
+ "continuous": true
236
+ },
237
+ "eef_rot_absolute": {
238
+ "absolute": true,
239
+ "rotation_type": "euler_angles_rpy",
240
+ "shape": [
241
+ 3
242
+ ],
243
+ "continuous": true
244
+ },
245
+ "gripper_close": {
246
+ "absolute": true,
247
+ "rotation_type": null,
248
+ "shape": [
249
+ 2
250
+ ],
251
+ "continuous": true
252
+ }
253
+ },
254
+ "action": {
255
+ "eef_pos_delta": {
256
+ "absolute": false,
257
+ "rotation_type": null,
258
+ "shape": [
259
+ 3
260
+ ],
261
+ "continuous": true
262
+ },
263
+ "eef_rot_delta": {
264
+ "absolute": false,
265
+ "rotation_type": "euler_angles_rpy",
266
+ "shape": [
267
+ 3
268
+ ],
269
+ "continuous": true
270
+ },
271
+ "gripper_close": {
272
+ "absolute": true,
273
+ "rotation_type": null,
274
+ "shape": [
275
+ 1
276
+ ],
277
+ "continuous": true
278
+ },
279
+ "ratio_label": {
280
+ "absolute": true,
281
+ "rotation_type": null,
282
+ "shape": [
283
+ 2
284
+ ],
285
+ "continuous": true
286
+ }
287
+ }
288
+ },
289
+ "embodiment_tag": "libero"
290
+ }
291
+ }
s1v2_contact/model-00001-of-00002.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3383a3ed5c2f244710237748f4096316f73474c4e3d5ec58281a4bca40edea65
3
+ size 4999367216
s1v2_contact/model-00002-of-00002.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:610d7e88ca6d7ad233c747f4bc5abb946c7ff9dcd075affae3848a84d4864ce2
3
+ size 3009499644
s1v2_contact/model.safetensors.index.json ADDED
The diff for this file is too large to render. See raw diff
 
s1v2_contact/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b8b9e8457a650a7649b023b95cbddbb7699b12f2c11c37bc5454bfb89655fce5
3
+ size 9396321091
s1v2_contact/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b244f1acb356e0fee47aa6b5d3bd5670f23d3252cb043909c69e9d161eb30dad
3
+ size 14917
s1v2_contact/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:047957c278accb70ff33923b0cc3fd89cfb5447451806659081d04ecf99c4be2
3
+ size 14917
s1v2_contact/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:96b418fbb509f96d95ad52dd99d48eeb782a306175bb55d2661eb9b3585613ba
3
+ size 1465
s1v2_contact/trainer_state.json ADDED
The diff for this file is too large to render. See raw diff