Hugging Face's logo

zaleni
/
WSA-Base-LIBERO

Robotics
LeRobot
Safetensors
English
embodied-ai
robot-learning
vision-language-action
world-model
Model card Files
xet
 Community
1
zaleni commited on
Commit
ffcd417
·
verified ·
1 Parent(s): f0d5952

Upload folder using huggingface_hub

Browse files
Files changed (4) hide show
  1. config.json +131 -0
  2. model.safetensors +3 -0
  3. stats.json +321 -0
  4. train_config.json +333 -0
config.json ADDED
@@ -0,0 +1,131 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "type": "TBot_SA1",
3
+ "n_obs_steps": 1,
4
+ "input_features": {
5
+ "observation.state": {
6
+ "type": "STATE",
7
+ "shape": [
8
+ 32
9
+ ]
10
+ }
11
+ },
12
+ "output_features": {
13
+ "action": {
14
+ "type": "ACTION",
15
+ "shape": [
16
+ 32
17
+ ]
18
+ }
19
+ },
20
+ "device": "cuda:0",
21
+ "use_amp": false,
22
+ "push_to_hub": false,
23
+ "repo_id": "zaleni/TBot-SA1-LIBERO",
24
+ "private": null,
25
+ "tags": null,
26
+ "license": null,
27
+ "pretrained_path": "zaleni/TBot-SA1-Base",
28
+ "qwen3_vl_variant": "qwen3_vl_28l",
29
+ "action_expert_variant": "qwen3_28l",
30
+ "qwen3_vl_pretrained_path": "Qwen/Qwen3-VL-2B-Instruct",
31
+ "dtype": "bfloat16",
32
+ "chunk_size": 10,
33
+ "n_action_steps": 10,
34
+ "max_state_dim": 32,
35
+ "max_action_dim": 32,
36
+ "mask_action_dim_padding_loss": true,
37
+ "action_loss_valid_dim": 7,
38
+ "num_inference_steps": 10,
39
+ "time_sampling_beta_alpha": 1.5,
40
+ "time_sampling_beta_beta": 1.0,
41
+ "time_sampling_scale": 0.999,
42
+ "time_sampling_offset": 0.001,
43
+ "min_period": 0.004,
44
+ "max_period": 4.0,
45
+ "attention_mask_mode": "causal",
46
+ "image_resolution": [
47
+ 224,
48
+ 224
49
+ ],
50
+ "image_delta_indices": [
51
+ -15,
52
+ 0,
53
+ 15
54
+ ],
55
+ "empty_cameras": 0,
56
+ "normalization_mapping": {
57
+ "VISUAL": "IDENTITY",
58
+ "STATE": "IDENTITY",
59
+ "ACTION": "IDENTITY"
60
+ },
61
+ "gradient_checkpointing": false,
62
+ "compile_model": false,
63
+ "compile_mode": "max-autotune",
64
+ "optimizer_lr": 7.5e-05,
65
+ "optimizer_betas": [
66
+ 0.9,
67
+ 0.95
68
+ ],
69
+ "optimizer_eps": 1e-08,
70
+ "optimizer_weight_decay": 0.01,
71
+ "optimizer_grad_clip_norm": 1.0,
72
+ "scheduler_warmup_steps": 2000,
73
+ "scheduler_decay_steps": 128000,
74
+ "scheduler_decay_lr": 5e-06,
75
+ "tokenizer_max_length": 48,
76
+ "freeze_vision_encoder": false,
77
+ "train_expert_only": false,
78
+ "train_vlm_only": false,
79
+ "lora_modules": [],
80
+ "lora_unselected_mode": "full",
81
+ "lora_targets": [
82
+ "attn",
83
+ "ffn"
84
+ ],
85
+ "lora_rank": 16,
86
+ "lora_alpha": 32.0,
87
+ "lora_rank_und": null,
88
+ "lora_alpha_und": null,
89
+ "lora_rank_gen": null,
90
+ "lora_alpha_gen": null,
91
+ "lora_rank_act": null,
92
+ "lora_alpha_act": null,
93
+ "lora_dropout": 0.0,
94
+ "scale_factor": 8,
95
+ "lambda_gen": 0.005,
96
+ "cosmos_tokenizer_path_or_name": "nvidia/Cosmos-Tokenizer-CI8x8",
97
+ "enable_3d_queries": true,
98
+ "num_3d_query_tokens": 432,
99
+ "da3_alignment_mode": "query_decoder",
100
+ "da3_query_resampler_layers": 1,
101
+ "da3_query_resampler_ff_mult": 1,
102
+ "query_layer_indices": [
103
+ 13,
104
+ 19,
105
+ 23,
106
+ 27
107
+ ],
108
+ "da3_variant": "auto",
109
+ "da3_teacher_layers": [
110
+ 11,
111
+ 15,
112
+ 19,
113
+ 23
114
+ ],
115
+ "da3_query_dim": 2048,
116
+ "da3_tokens_per_view": 1296,
117
+ "da3_num_views": 3,
118
+ "lambda_3d": 0.01,
119
+ "da3_model_path_or_name": "depth-anything/DA3-LARGE-1.1",
120
+ "da3_model_name": null,
121
+ "da3_code_root": null,
122
+ "da3_teacher_process_res": 504,
123
+ "da3_layer_weights": [
124
+ 1.0,
125
+ 1.2,
126
+ 1.4,
127
+ 1.6
128
+ ],
129
+ "future_query_init_std": 0.02,
130
+ "log_da3_teacher_timing": true
131
+ }
model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:603eacbb1a7da80dd2d41cbbd814a833bd57dec30601da5fda35651d6dd46e67
3
+ size 6608313224
stats.json ADDED
@@ -0,0 +1,321 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "libero_franka": {
3
+ "observation.state": {
4
+ "min": [
5
+ -0.4828203022480011,
6
+ -0.3255046010017395,
7
+ 0.008128180168569088,
8
+ 0.35277295112609863,
9
+ -3.641430377960205,
10
+ -1.842738389968872,
11
+ -0.0013586411951109767,
12
+ -0.042040832340717316
13
+ ],
14
+ "max": [
15
+ 0.21031762659549713,
16
+ 0.39128610491752625,
17
+ 1.3660105466842651,
18
+ 3.6714255809783936,
19
+ 3.560650587081909,
20
+ 1.386339545249939,
21
+ 0.04233968257904053,
22
+ 0.0013633022317662835
23
+ ],
24
+ "mean": [
25
+ -0.046518679708242416,
26
+ 0.0344090536236763,
27
+ 0.7645527720451355,
28
+ 2.972210168838501,
29
+ -0.22046972811222076,
30
+ -0.12557938694953918,
31
+ 0.026914240792393684,
32
+ -0.02719077467918396
33
+ ],
34
+ "std": [
35
+ 0.10494392365217209,
36
+ 0.15176630020141602,
37
+ 0.37851521372795105,
38
+ 0.3442504107952118,
39
+ 0.9069470763206482,
40
+ 0.3253919780254364,
41
+ 0.014175928197801113,
42
+ 0.014058895409107208
43
+ ],
44
+ "count": [
45
+ 273465
46
+ ]
47
+ },
48
+ "observation.images.wrist_image": {
49
+ "min": [
50
+ [
51
+ [
52
+ 0.0
53
+ ]
54
+ ],
55
+ [
56
+ [
57
+ 0.0
58
+ ]
59
+ ],
60
+ [
61
+ [
62
+ 0.0
63
+ ]
64
+ ]
65
+ ],
66
+ "max": [
67
+ [
68
+ [
69
+ 1.0
70
+ ]
71
+ ],
72
+ [
73
+ [
74
+ 1.0
75
+ ]
76
+ ],
77
+ [
78
+ [
79
+ 1.0
80
+ ]
81
+ ]
82
+ ],
83
+ "mean": [
84
+ [
85
+ [
86
+ 0.5846596441867226
87
+ ]
88
+ ],
89
+ [
90
+ [
91
+ 0.5453226857018393
92
+ ]
93
+ ],
94
+ [
95
+ [
96
+ 0.49830541633479514
97
+ ]
98
+ ]
99
+ ],
100
+ "std": [
101
+ [
102
+ [
103
+ 0.24875102462076606
104
+ ]
105
+ ],
106
+ [
107
+ [
108
+ 0.2252864323219699
109
+ ]
110
+ ],
111
+ [
112
+ [
113
+ 0.20095594997710015
114
+ ]
115
+ ]
116
+ ],
117
+ "count": [
118
+ 42695
119
+ ]
120
+ },
121
+ "index": {
122
+ "min": [
123
+ 0.0
124
+ ],
125
+ "max": [
126
+ 101468.0
127
+ ],
128
+ "mean": [
129
+ 37110.2578125
130
+ ],
131
+ "std": [
132
+ 24810.0703125
133
+ ],
134
+ "count": [
135
+ 273465
136
+ ]
137
+ },
138
+ "action": {
139
+ "min": [
140
+ -0.9375,
141
+ -0.9375,
142
+ -0.9375,
143
+ -0.2582142949104309,
144
+ -0.375,
145
+ -0.3675000071525574,
146
+ 0.0
147
+ ],
148
+ "max": [
149
+ 0.9375,
150
+ 0.9375,
151
+ 0.9375,
152
+ 0.3557142913341522,
153
+ 0.375,
154
+ 0.375,
155
+ 1.0
156
+ ],
157
+ "mean": [
158
+ 0.06278150528669357,
159
+ 0.0868406891822815,
160
+ -0.09037303924560547,
161
+ 0.0005407433491200209,
162
+ 0.005643374752253294,
163
+ -0.005229101050645113,
164
+ 0.5248202085494995
165
+ ],
166
+ "std": [
167
+ 0.33552396297454834,
168
+ 0.3784470856189728,
169
+ 0.4447287619113922,
170
+ 0.03924354165792465,
171
+ 0.06339297443628311,
172
+ 0.07797030359506607,
173
+ 0.4993835687637329
174
+ ],
175
+ "count": [
176
+ 273465
177
+ ]
178
+ },
179
+ "observation.images.image": {
180
+ "min": [
181
+ [
182
+ [
183
+ 0.0
184
+ ]
185
+ ],
186
+ [
187
+ [
188
+ 0.0
189
+ ]
190
+ ],
191
+ [
192
+ [
193
+ 0.0
194
+ ]
195
+ ]
196
+ ],
197
+ "max": [
198
+ [
199
+ [
200
+ 1.0
201
+ ]
202
+ ],
203
+ [
204
+ [
205
+ 1.0
206
+ ]
207
+ ],
208
+ [
209
+ [
210
+ 1.0
211
+ ]
212
+ ]
213
+ ],
214
+ "mean": [
215
+ [
216
+ [
217
+ 0.49465999363758456
218
+ ]
219
+ ],
220
+ [
221
+ [
222
+ 0.46759409514760075
223
+ ]
224
+ ],
225
+ [
226
+ [
227
+ 0.4307482421582366
228
+ ]
229
+ ]
230
+ ],
231
+ "std": [
232
+ [
233
+ [
234
+ 0.22094010046605658
235
+ ]
236
+ ],
237
+ [
238
+ [
239
+ 0.20259253787959647
240
+ ]
241
+ ],
242
+ [
243
+ [
244
+ 0.18565729345932921
245
+ ]
246
+ ]
247
+ ],
248
+ "count": [
249
+ 42695
250
+ ]
251
+ },
252
+ "task_index": {
253
+ "min": [
254
+ 0.0
255
+ ],
256
+ "max": [
257
+ 9.0
258
+ ],
259
+ "mean": [
260
+ 4.49923038482666
261
+ ],
262
+ "std": [
263
+ 2.8671939373016357
264
+ ],
265
+ "count": [
266
+ 273465
267
+ ]
268
+ },
269
+ "episode_index": {
270
+ "min": [
271
+ 0.0
272
+ ],
273
+ "max": [
274
+ 453.0
275
+ ],
276
+ "mean": [
277
+ 207.07574462890625
278
+ ],
279
+ "std": [
280
+ 121.92366027832031
281
+ ],
282
+ "count": [
283
+ 273465
284
+ ]
285
+ },
286
+ "frame_index": {
287
+ "min": [
288
+ 0.0
289
+ ],
290
+ "max": [
291
+ 504.0
292
+ ],
293
+ "mean": [
294
+ 94.64840698242188
295
+ ],
296
+ "std": [
297
+ 72.59561920166016
298
+ ],
299
+ "count": [
300
+ 273465
301
+ ]
302
+ },
303
+ "timestamp": {
304
+ "min": [
305
+ 0.0
306
+ ],
307
+ "max": [
308
+ 25.200000762939453
309
+ ],
310
+ "mean": [
311
+ 4.732416152954102
312
+ ],
313
+ "std": [
314
+ 3.6297850608825684
315
+ ],
316
+ "count": [
317
+ 273465
318
+ ]
319
+ }
320
+ }
321
+ }
train_config.json ADDED
@@ -0,0 +1,333 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "dataset": {
3
+ "type": "TBot_SA1",
4
+ "repo_id": "multidata_from_file",
5
+ "repo_id_file": "outputs/TBot_SA1/_repo_id_files/TBot_SA1-libero4-abs-chunk10-finetune-2026_05_24_11_47_11.txt",
6
+ "root": null,
7
+ "episodes": null,
8
+ "image_transforms": {
9
+ "enable": false,
10
+ "preset": null,
11
+ "max_num_transforms": 3,
12
+ "random_order": false,
13
+ "tfs": {
14
+ "brightness": {
15
+ "weight": 1.0,
16
+ "type": "ColorJitter",
17
+ "kwargs": {
18
+ "brightness": [
19
+ 0.8,
20
+ 1.2
21
+ ]
22
+ }
23
+ },
24
+ "contrast": {
25
+ "weight": 1.0,
26
+ "type": "ColorJitter",
27
+ "kwargs": {
28
+ "contrast": [
29
+ 0.8,
30
+ 1.2
31
+ ]
32
+ }
33
+ },
34
+ "saturation": {
35
+ "weight": 1.0,
36
+ "type": "ColorJitter",
37
+ "kwargs": {
38
+ "saturation": [
39
+ 0.5,
40
+ 1.5
41
+ ]
42
+ }
43
+ },
44
+ "hue": {
45
+ "weight": 1.0,
46
+ "type": "ColorJitter",
47
+ "kwargs": {
48
+ "hue": [
49
+ -0.05,
50
+ 0.05
51
+ ]
52
+ }
53
+ },
54
+ "sharpness": {
55
+ "weight": 1.0,
56
+ "type": "SharpnessJitter",
57
+ "kwargs": {
58
+ "sharpness": [
59
+ 0.5,
60
+ 1.5
61
+ ]
62
+ }
63
+ },
64
+ "affine": {
65
+ "weight": 1.0,
66
+ "type": "RandomAffine",
67
+ "kwargs": {
68
+ "degrees": [
69
+ -5.0,
70
+ 5.0
71
+ ],
72
+ "translate": [
73
+ 0.05,
74
+ 0.05
75
+ ]
76
+ }
77
+ }
78
+ }
79
+ },
80
+ "revision": null,
81
+ "use_imagenet_stats": true,
82
+ "use_external_stats": true,
83
+ "external_stats_path": "norm_stats/libero_all_chunk10/franka/abs/stats.json",
84
+ "external_stats_root": null,
85
+ "weight_rules_path": null,
86
+ "video_backend": "torchcodec",
87
+ "streaming": false,
88
+ "dist_loading": false,
89
+ "buffer_size": 1024,
90
+ "action_mode": "abs",
91
+ "repack_transforms": {
92
+ "inputs": [],
93
+ "outputs": []
94
+ },
95
+ "data_transforms": {
96
+ "inputs": [
97
+ {
98
+ "type": "resize_with_pad",
99
+ "height": 224,
100
+ "width": 224,
101
+ "mode": "bilinear"
102
+ },
103
+ {
104
+ "type": "remap_image_key",
105
+ "mapping": {}
106
+ },
107
+ {
108
+ "type": "normalize",
109
+ "selected_keys": null,
110
+ "mode": "mean_std",
111
+ "norm_stats": {}
112
+ },
113
+ {
114
+ "type": "compose_fields",
115
+ "mapping": {}
116
+ },
117
+ {
118
+ "type": "pad_state_and_action",
119
+ "max_state_dim": 32,
120
+ "max_action_dim": 32
121
+ },
122
+ {
123
+ "type": "tbot_sa1_processor",
124
+ "pretrained_model_name_or_path": "Qwen/Qwen3-VL-2B-Instruct",
125
+ "max_length": 48,
126
+ "task_key": "task",
127
+ "padding_side": "right",
128
+ "padding": "max_length",
129
+ "truncation": true,
130
+ "spatial_merge_size": 2,
131
+ "vision_start_token_id": 151652,
132
+ "vision_end_token_id": 151653,
133
+ "image_token_id": 151655,
134
+ "processor": null,
135
+ "_processor_source": null
136
+ },
137
+ {
138
+ "type": "unify_tbot_sa1_inputs"
139
+ }
140
+ ],
141
+ "outputs": []
142
+ },
143
+ "model_transforms": {
144
+ "inputs": [],
145
+ "outputs": []
146
+ },
147
+ "height": 224,
148
+ "width": 224,
149
+ "max_state_dim": 32,
150
+ "max_action_dim": 32,
151
+ "qwen3_vl_processor_path": "Qwen/Qwen3-VL-2B-Instruct"
152
+ },
153
+ "env": null,
154
+ "policy": {
155
+ "type": "TBot_SA1",
156
+ "n_obs_steps": 1,
157
+ "input_features": {
158
+ "observation.state": {
159
+ "type": "STATE",
160
+ "shape": [
161
+ 32
162
+ ]
163
+ }
164
+ },
165
+ "output_features": {
166
+ "action": {
167
+ "type": "ACTION",
168
+ "shape": [
169
+ 32
170
+ ]
171
+ }
172
+ },
173
+ "device": "cuda:0",
174
+ "use_amp": false,
175
+ "push_to_hub": false,
176
+ "repo_id": "zaleni/TBot-SA1-LIBERO",
177
+ "private": null,
178
+ "tags": null,
179
+ "license": null,
180
+ "pretrained_path": "zaleni/TBot-SA1-Base",
181
+ "qwen3_vl_variant": "qwen3_vl_28l",
182
+ "action_expert_variant": "qwen3_28l",
183
+ "qwen3_vl_pretrained_path": "Qwen/Qwen3-VL-2B-Instruct",
184
+ "dtype": "bfloat16",
185
+ "chunk_size": 10,
186
+ "n_action_steps": 10,
187
+ "max_state_dim": 32,
188
+ "max_action_dim": 32,
189
+ "mask_action_dim_padding_loss": true,
190
+ "action_loss_valid_dim": 7,
191
+ "num_inference_steps": 10,
192
+ "time_sampling_beta_alpha": 1.5,
193
+ "time_sampling_beta_beta": 1.0,
194
+ "time_sampling_scale": 0.999,
195
+ "time_sampling_offset": 0.001,
196
+ "min_period": 0.004,
197
+ "max_period": 4.0,
198
+ "attention_mask_mode": "causal",
199
+ "image_resolution": [
200
+ 224,
201
+ 224
202
+ ],
203
+ "image_delta_indices": [
204
+ -15,
205
+ 0,
206
+ 15
207
+ ],
208
+ "empty_cameras": 0,
209
+ "normalization_mapping": {
210
+ "VISUAL": "IDENTITY",
211
+ "STATE": "IDENTITY",
212
+ "ACTION": "IDENTITY"
213
+ },
214
+ "gradient_checkpointing": false,
215
+ "compile_model": false,
216
+ "compile_mode": "max-autotune",
217
+ "optimizer_lr": 7.5e-05,
218
+ "optimizer_betas": [
219
+ 0.9,
220
+ 0.95
221
+ ],
222
+ "optimizer_eps": 1e-08,
223
+ "optimizer_weight_decay": 0.01,
224
+ "optimizer_grad_clip_norm": 1.0,
225
+ "scheduler_warmup_steps": 2000,
226
+ "scheduler_decay_steps": 128000,
227
+ "scheduler_decay_lr": 5e-06,
228
+ "tokenizer_max_length": 48,
229
+ "freeze_vision_encoder": false,
230
+ "train_expert_only": false,
231
+ "train_vlm_only": false,
232
+ "lora_modules": [],
233
+ "lora_unselected_mode": "full",
234
+ "lora_targets": [
235
+ "attn",
236
+ "ffn"
237
+ ],
238
+ "lora_rank": 16,
239
+ "lora_alpha": 32.0,
240
+ "lora_rank_und": null,
241
+ "lora_alpha_und": null,
242
+ "lora_rank_gen": null,
243
+ "lora_alpha_gen": null,
244
+ "lora_rank_act": null,
245
+ "lora_alpha_act": null,
246
+ "lora_dropout": 0.0,
247
+ "scale_factor": 8,
248
+ "lambda_gen": 0.005,
249
+ "cosmos_tokenizer_path_or_name": "nvidia/Cosmos-Tokenizer-CI8x8",
250
+ "enable_3d_queries": true,
251
+ "num_3d_query_tokens": 432,
252
+ "da3_alignment_mode": "query_decoder",
253
+ "da3_query_resampler_layers": 1,
254
+ "da3_query_resampler_ff_mult": 1,
255
+ "query_layer_indices": [
256
+ 13,
257
+ 19,
258
+ 23,
259
+ 27
260
+ ],
261
+ "da3_variant": "auto",
262
+ "da3_teacher_layers": [
263
+ 11,
264
+ 15,
265
+ 19,
266
+ 23
267
+ ],
268
+ "da3_query_dim": 2048,
269
+ "da3_tokens_per_view": 1296,
270
+ "da3_num_views": 3,
271
+ "lambda_3d": 0.01,
272
+ "da3_model_path_or_name": "depth-anything/DA3-LARGE-1.1",
273
+ "da3_model_name": null,
274
+ "da3_code_root": null,
275
+ "da3_teacher_process_res": 504,
276
+ "da3_layer_weights": [
277
+ 1.0,
278
+ 1.2,
279
+ 1.4,
280
+ 1.6
281
+ ],
282
+ "future_query_init_std": 0.02,
283
+ "log_da3_teacher_timing": true
284
+ },
285
+ "output_dir": "outputs/TBot_SA1/TBot_SA1-libero4-abs-chunk10-finetune-2026_05_24_11_47_11",
286
+ "job_name": "TBot_SA1-libero4-abs-chunk10-finetune-2026_05_24_11_47_11",
287
+ "resume": false,
288
+ "seed": 42,
289
+ "num_workers": 12,
290
+ "batch_size": 16,
291
+ "gradient_accumulation_steps": 1,
292
+ "steps": 128000,
293
+ "eval_freq": 20000,
294
+ "eval_max_batches": 2,
295
+ "log_freq": 100,
296
+ "save_checkpoint": true,
297
+ "save_freq": 20000,
298
+ "use_policy_training_preset": true,
299
+ "optimizer": {
300
+ "type": "adamw",
301
+ "lr": 7.5e-05,
302
+ "weight_decay": 0.01,
303
+ "grad_clip_norm": 1.0,
304
+ "betas": [
305
+ 0.9,
306
+ 0.95
307
+ ],
308
+ "eps": 1e-08
309
+ },
310
+ "scheduler": {
311
+ "type": "cosine_decay_with_warmup",
312
+ "num_warmup_steps": 2000,
313
+ "num_decay_steps": 128000,
314
+ "peak_lr": 7.5e-05,
315
+ "decay_lr": 5e-06
316
+ },
317
+ "eval": {
318
+ "n_episodes": 50,
319
+ "batch_size": 50,
320
+ "use_async_envs": false
321
+ },
322
+ "wandb": {
323
+ "enable": true,
324
+ "disable_artifact": true,
325
+ "project": "TBot_SA1",
326
+ "entity": null,
327
+ "notes": null,
328
+ "run_id": "t6vqahnt",
329
+ "mode": "offline"
330
+ },
331
+ "checkpoint_path": null,
332
+ "rename_map": {}
333
+ }