Safetensors
English
jie530 commited on
Commit
cbec686
·
verified ·
1 Parent(s): 5ed3ad3

Upload folder using huggingface_hub

Browse files
psi0/simple-checkpoints/g1wholebodyopentrashcanteleop-v0.simple.flow1000.cosine.lr1.0e-04.b128.gpus8.2605291607/argv.txt ADDED
@@ -0,0 +1,53 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ scripts/train.py
2
+ finetune_simple_psi0_config
3
+ --seed=292285
4
+ --exp=g1wholebodyopentrashcanteleop-v0
5
+ --train.name=finetune
6
+ --train.data_parallel=ddp
7
+ --train.mixed_precision=bf16
8
+ --train.train_batch_size=16
9
+ --train.max_checkpoints_to_keep=5
10
+ --train.gradient_accumulation_steps=1
11
+ --train.learning_rate=1e-4
12
+ --train.max_training_steps=40000
13
+ --train.warmup_ratio=None
14
+ --train.warmup_steps=1000
15
+ --train.checkpointing_steps=10000
16
+ --train.validation_steps=500
17
+ --train.val_num_batches=20
18
+ --train.max_grad_norm=1.0
19
+ --train.lr_scheduler_type=cosine
20
+ --train.lr_scheduler_kwargs.weight_decay=1e-6
21
+ --train.lr_scheduler_kwargs.betas 0.95 0.999
22
+ --log.report_to=wandb
23
+ --data.root_dir=/data/jliu/data
24
+ --data.train-repo-ids=G1WholebodyOpenTrashCanTeleop-v0
25
+ --data.transform.repack.pad-action-dim=36
26
+ --data.transform.repack.pad-state-dim=36
27
+ --data.transform.field.stat-path=meta/stats_psi0.json
28
+ --data.transform.field.stat-action-key=action
29
+ --data.transform.field.stat-state-key=states
30
+ --data.transform.field.action_norm_type=bounds
31
+ --data.transform.field.no-use-norm-mask
32
+ --data.transform.field.normalize-state
33
+ --data.transform.field.pad-action-dim=36
34
+ --data.transform.field.pad-state-dim=36
35
+ --data.transform.model.img-aug
36
+ --data.transform.model.resize.size 180 320
37
+ --data.transform.model.center_crop.size 180 320
38
+ --model.model_name_or_path=/hfm/cache/checkpoints/psi0/pre.fast.1by1.2601091803.ckpt.ego200k.he30k
39
+ --model.pretrained-action-header-path=/hfm/cache/checkpoints/psi0/postpre.1by1.pad36.2601131206.ckpt.he30k
40
+ --model.noise-scheduler=flow
41
+ --model.train-diffusion-steps=1000
42
+ --model.n_conditions=0
43
+ --model.action-chunk-size=30
44
+ --model.action-dim=36
45
+ --model.action-exec-horizon=30
46
+ --model.observation-horizon=1
47
+ --model.odim=36
48
+ --model.view_feature_dim=2048
49
+ --model.no-tune-vlm
50
+ --model.no-use_film
51
+ --model.no-combined_temb
52
+ --model.rtc
53
+ --model.max-delay=8
psi0/simple-checkpoints/g1wholebodyopentrashcanteleop-v0.simple.flow1000.cosine.lr1.0e-04.b128.gpus8.2605291607/checkpoints/ckpt_40000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5e2f5911ccceb9b88e266a1eeae933fed1b41a91c553cd71e331526e29d61402
3
+ size 6253648840
psi0/simple-checkpoints/g1wholebodyopentrashcanteleop-v0.simple.flow1000.cosine.lr1.0e-04.b128.gpus8.2605291607/envs.txt ADDED
@@ -0,0 +1,17 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ OMP_NUM_THREADS=32
2
+ HF_HOME=/data/cache
3
+ TORCH_HOME=/data/cache
4
+ HF_TOKEN=hf_...TiKa
5
+ HF_LEROBOT_HOME=/data/data/lerobot
6
+ WE_HOME=Not Set
7
+ DATA_HOME=/data/data
8
+ UV_CACHE_DIR=/data/cache
9
+ WANDB_API_KEY=90e...5c06
10
+ PROTOCOL_BUFFERS_PYTHON_IMPLEMENTATION=python
11
+ CUDA_VISIBLE_DEVICES=0,1,2,3,4,5,6,7
12
+ WORLD_SIZE=8
13
+ LOCAL_WORLD_SIZE=8
14
+ RANK=0
15
+ LOCAL_RANK=0
16
+ MASTER_ADDR=127.0.0.1
17
+ MASTER_PORT=29509
psi0/simple-checkpoints/g1wholebodyopentrashcanteleop-v0.simple.flow1000.cosine.lr1.0e-04.b128.gpus8.2605291607/run_config.json ADDED
@@ -0,0 +1,348 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "exp": "g1wholebodyopentrashcanteleop-v0",
3
+ "seed": 292285,
4
+ "auto_tag_run": false,
5
+ "eval": false,
6
+ "debug": false,
7
+ "timestamp": "2605291607",
8
+ "log": {
9
+ "logging_dir": "logs",
10
+ "report_to": "wandb",
11
+ "log_freq": 100
12
+ },
13
+ "wandb": {
14
+ "project": "psi",
15
+ "entity": "jliu530-soochow-university",
16
+ "group": "finetune",
17
+ "id": "tda15vvs",
18
+ "name": "g1wholebodyopentrashcanteleop-v0.simple.flow1000.cosine.lr1.0e-04.b128.gpus8.2605291607",
19
+ "resume": "allow"
20
+ },
21
+ "train": {
22
+ "num_workers": 8,
23
+ "overfit_single_batch": false,
24
+ "name": "finetune",
25
+ "resume_from_checkpoint": null,
26
+ "skip_resumed_steps": false,
27
+ "hf_token": ".hf_token",
28
+ "lora": false,
29
+ "output_dir": ".runs",
30
+ "gradient_accumulation_steps": 1,
31
+ "mixed_precision": "bf16",
32
+ "max_grad_norm": 1.0,
33
+ "optimizer_foreach": null,
34
+ "train_batch_size": 16,
35
+ "val_batch_size": 16,
36
+ "val_num_batches": 20,
37
+ "checkpointing_steps": 10000,
38
+ "max_checkpoints_to_keep": 5,
39
+ "validation_steps": 500,
40
+ "learning_rate": 0.0001,
41
+ "lr_scheduler_type": "cosine",
42
+ "lr_scheduler_kwargs": {
43
+ "betas": [
44
+ 0.95,
45
+ 0.999
46
+ ],
47
+ "weight_decay": 1e-6,
48
+ "eps": 1e-8
49
+ },
50
+ "scheduler_specific_kwargs": {},
51
+ "data_parallel": "ddp",
52
+ "sharding_strategy": "full-shard",
53
+ "deepspeed_config": "/data/jliu/psi/src/InternVLA/config/deepseeds/zero3.json",
54
+ "enable_gradient_checkpointing": true,
55
+ "enable_mixed_precision_training": true,
56
+ "reduce_in_full_precision": true,
57
+ "max_training_steps": 40000,
58
+ "num_train_epochs": null,
59
+ "warmup_steps": 1000,
60
+ "warmup_ratio": null
61
+ },
62
+ "data": {
63
+ "transform": {
64
+ "repack": {
65
+ "dataset_name": "simple",
66
+ "num_past_frames": 0,
67
+ "action_chunk_size": 30,
68
+ "pad_action_dim": 36,
69
+ "pad_state_dim": 36
70
+ },
71
+ "model": {
72
+ "resize": {
73
+ "size": [
74
+ 180,
75
+ 320
76
+ ]
77
+ },
78
+ "center_crop": {
79
+ "size": [
80
+ 180,
81
+ 320
82
+ ]
83
+ },
84
+ "color_jitter": {
85
+ "brightness": 0.2,
86
+ "contrast": [
87
+ 0.8,
88
+ 1.2
89
+ ],
90
+ "saturation": [
91
+ 0.8,
92
+ 1.2
93
+ ],
94
+ "hue": 0.05
95
+ },
96
+ "gaussian_noise": {
97
+ "mean": 0.0,
98
+ "std": 3.0,
99
+ "prob_skip": 0.1
100
+ },
101
+ "img_aug": true,
102
+ "adaptive_resize": false,
103
+ "img_sizes": {
104
+ "egodex": [
105
+ 270,
106
+ 480
107
+ ],
108
+ "he": [
109
+ 240,
110
+ 320
111
+ ]
112
+ }
113
+ },
114
+ "field": {
115
+ "stat_path": "meta/stats_psi0.json",
116
+ "action_norm_type": "bounds",
117
+ "stat_action_key": "action",
118
+ "stat_state_key": "states",
119
+ "use_norm_mask": false,
120
+ "action_norm_masks": [
121
+ true,
122
+ true,
123
+ true,
124
+ true,
125
+ true,
126
+ true,
127
+ false
128
+ ],
129
+ "action_min": [
130
+ -0.5,
131
+ 0.0,
132
+ 0.0,
133
+ -0.6000000238418579,
134
+ -1.5,
135
+ -1.5,
136
+ -1.5,
137
+ -0.5,
138
+ -0.699999988079071,
139
+ -0.699999988079071,
140
+ -1.909680248814551e-16,
141
+ -1.909680248814551e-16,
142
+ -2.865715791895621e-19,
143
+ -1.909680248814551e-16,
144
+ -0.7142341136932373,
145
+ 0.1900009959936142,
146
+ -1.1259523630142212,
147
+ -0.7086443901062012,
148
+ -1.2021900415420532,
149
+ -0.6679710149765015,
150
+ -0.4076652526855469,
151
+ -0.831639289855957,
152
+ -1.734777808189392,
153
+ -0.939987301826477,
154
+ -0.5639535188674927,
155
+ -1.7956628799438477,
156
+ -1.0757752656936646,
157
+ -1.3881421089172363,
158
+ -0.19929488003253937,
159
+ -0.11428708583116531,
160
+ -0.4480314552783966,
161
+ 0.7400000095367432,
162
+ -0.0659722238779068,
163
+ -0.1961783766746521,
164
+ 0.0,
165
+ 0.0
166
+ ],
167
+ "action_max": [
168
+ 0.0,
169
+ 0.699999988079071,
170
+ 0.699999988079071,
171
+ 0.0,
172
+ 0.0,
173
+ 0.0,
174
+ 0.0,
175
+ 2.660191394973399e-19,
176
+ 1.107640500578317e-16,
177
+ 1.107640500578317e-16,
178
+ 1.5,
179
+ 1.5,
180
+ 0.6000000238418579,
181
+ 1.5,
182
+ 0.10599426180124283,
183
+ 0.8050272464752197,
184
+ 1.5549964904785156,
185
+ 0.7429194450378418,
186
+ 1.2684427499771118,
187
+ 0.30270305275917053,
188
+ 1.1032674312591553,
189
+ 0.19180816411972046,
190
+ -0.1900009959936142,
191
+ 0.5824279189109802,
192
+ 0.9400193095207214,
193
+ 1.6776031255722046,
194
+ 1.2094172239303589,
195
+ 0.4964087903499603,
196
+ 0.35385459661483765,
197
+ 0.5199990272521973,
198
+ 0.7828513383865356,
199
+ 0.7400000095367432,
200
+ 0.5,
201
+ 0.5,
202
+ 0.0,
203
+ 0.0
204
+ ],
205
+ "state_min": [
206
+ -0.5538988709449768,
207
+ -0.038441382348537445,
208
+ -0.00016361191228497773,
209
+ -0.9720256328582764,
210
+ -1.2885046005249023,
211
+ -0.5443171858787537,
212
+ -1.4196548461914062,
213
+ -0.7833675742149353,
214
+ -1.0398303270339966,
215
+ -0.685627281665802,
216
+ -0.007488771341741085,
217
+ -0.0005658457521349192,
218
+ -0.010361753404140472,
219
+ -0.0005848633009009063,
220
+ -0.6532095670700073,
221
+ 0.15542864799499512,
222
+ -1.1354281902313232,
223
+ -0.6519288420677185,
224
+ -1.1867295503616333,
225
+ -0.5740351676940918,
226
+ -0.36576342582702637,
227
+ -0.7506816983222961,
228
+ -1.6270910501480103,
229
+ -0.9108870029449463,
230
+ -0.501366376876831,
231
+ -1.8017230033874512,
232
+ -1.0341318845748901,
233
+ -1.3451646566390991,
234
+ -0.1274876743555069,
235
+ -0.05407978966832161,
236
+ -0.47854718565940857,
237
+ 0.7400000095367432,
238
+ 0.0,
239
+ 0.0,
240
+ 0.0,
241
+ 0.0
242
+ ],
243
+ "state_max": [
244
+ 0.003301811870187521,
245
+ 0.4652666747570038,
246
+ 0.6686679720878601,
247
+ 0.0006876717670820653,
248
+ 0.0005752479191869497,
249
+ 0.0029399837367236614,
250
+ 0.00251425476744771,
251
+ 0.0333738848567009,
252
+ 0.6730948686599731,
253
+ 0.004870148375630379,
254
+ 0.5727766752243042,
255
+ 1.4610393047332764,
256
+ 1.4522792100906372,
257
+ 1.4519342184066772,
258
+ 0.12466324865818024,
259
+ 0.6869632601737976,
260
+ 1.5526267290115356,
261
+ 0.7649725675582886,
262
+ 1.2608386278152466,
263
+ 0.34103095531463623,
264
+ 1.073241114616394,
265
+ 0.20825356245040894,
266
+ -0.1648865044116974,
267
+ 0.5940349698066711,
268
+ 0.9402427077293396,
269
+ 1.6149287223815918,
270
+ 1.4479284286499023,
271
+ 0.47971057891845703,
272
+ 0.18096975982189178,
273
+ 0.5209036469459534,
274
+ 0.8097797632217407,
275
+ 0.7400000095367432,
276
+ 0.0,
277
+ 0.0,
278
+ 0.0,
279
+ 0.0
280
+ ],
281
+ "normalize_state": true,
282
+ "pad_action_dim": 36,
283
+ "pad_state_dim": 36
284
+ }
285
+ },
286
+ "root_dir": "/data/jliu/data",
287
+ "train_repo_ids": [
288
+ "G1WholebodyOpenTrashCanTeleop-v0"
289
+ ],
290
+ "val_repo_ids": [
291
+ "G1WholebodyOpenTrashCanTeleop-v0"
292
+ ]
293
+ },
294
+ "model": {
295
+ "resnet_store_path": null,
296
+ "pretrained_action_header_path": "/hfm/cache/checkpoints/psi0/postpre.1by1.pad36.2601131206.ckpt.he30k",
297
+ "rtc": true,
298
+ "max_delay": 8,
299
+ "action_dim": 36,
300
+ "action_chunk_size": 30,
301
+ "action_exec_horizon": 30,
302
+ "observation_horizon": 1,
303
+ "img_chunk": 1,
304
+ "n_cams": 1,
305
+ "use_obs": "add_token",
306
+ "dropout": 0.1,
307
+ "noise_scheduler": "flow",
308
+ "train_diffusion_steps": 1000,
309
+ "eval_diffusion_steps": 10,
310
+ "share_cam_features": false,
311
+ "early_fusion": false,
312
+ "odim": 36,
313
+ "n_conditions": 0,
314
+ "token_fusion": "concat",
315
+ "loss_w": [
316
+ 0.1,
317
+ 0.2,
318
+ 0.1
319
+ ],
320
+ "time_dim": 256,
321
+ "hidden_dim": 1536,
322
+ "num_blocks": 6,
323
+ "dim_feedforward": 2048,
324
+ "nhead": 24,
325
+ "activation": "gelu",
326
+ "view_feature_dim": 2048,
327
+ "use_film": false,
328
+ "combined_temb": false,
329
+ "use_dit": false,
330
+ "weight_decay": 0.01,
331
+ "model_name_or_path": "/hfm/cache/checkpoints/psi0/pre.fast.1by1.2601091803.ckpt.ego200k.he30k",
332
+ "vlm_ckpt_step": null,
333
+ "tune_vlm": false,
334
+ "tune_mm_llm": false,
335
+ "tune_mm_vision": false,
336
+ "tune_mm_mlp": false,
337
+ "gradient_checkpointing": true,
338
+ "lang_backbone_lr": 0.00001,
339
+ "mm_projector_lr": 0.00001,
340
+ "vision_tower_lr": 1e-6,
341
+ "optim": "adamw_torch",
342
+ "model_max_length": 4096,
343
+ "data_flatten": true,
344
+ "data_packing": true,
345
+ "max_pixels": 451584,
346
+ "min_pixels": 12544
347
+ }
348
+ }
psi0/simple-checkpoints/g1wholebodypushofficechairteleop-v0.simple.flow1000.cosine.lr1.0e-04.b128.gpus8.2605291520/argv.txt ADDED
@@ -0,0 +1,53 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ scripts/train.py
2
+ finetune_simple_psi0_config
3
+ --seed=292285
4
+ --exp=g1wholebodypushofficechairteleop-v0
5
+ --train.name=finetune
6
+ --train.data_parallel=ddp
7
+ --train.mixed_precision=bf16
8
+ --train.train_batch_size=16
9
+ --train.max_checkpoints_to_keep=5
10
+ --train.gradient_accumulation_steps=1
11
+ --train.learning_rate=1e-4
12
+ --train.max_training_steps=40000
13
+ --train.warmup_ratio=None
14
+ --train.warmup_steps=1000
15
+ --train.checkpointing_steps=10000
16
+ --train.validation_steps=500
17
+ --train.val_num_batches=20
18
+ --train.max_grad_norm=1.0
19
+ --train.lr_scheduler_type=cosine
20
+ --train.lr_scheduler_kwargs.weight_decay=1e-6
21
+ --train.lr_scheduler_kwargs.betas 0.95 0.999
22
+ --log.report_to=wandb
23
+ --data.root_dir=/data/jliu/data
24
+ --data.train-repo-ids=G1WholebodyPushOfficeChairTeleop-v0
25
+ --data.transform.repack.pad-action-dim=36
26
+ --data.transform.repack.pad-state-dim=36
27
+ --data.transform.field.stat-path=meta/stats_psi0.json
28
+ --data.transform.field.stat-action-key=action
29
+ --data.transform.field.stat-state-key=states
30
+ --data.transform.field.action_norm_type=bounds
31
+ --data.transform.field.no-use-norm-mask
32
+ --data.transform.field.normalize-state
33
+ --data.transform.field.pad-action-dim=36
34
+ --data.transform.field.pad-state-dim=36
35
+ --data.transform.model.img-aug
36
+ --data.transform.model.resize.size 180 320
37
+ --data.transform.model.center_crop.size 180 320
38
+ --model.model_name_or_path=/hfm/cache/checkpoints/psi0/pre.fast.1by1.2601091803.ckpt.ego200k.he30k
39
+ --model.pretrained-action-header-path=/hfm/cache/checkpoints/psi0/postpre.1by1.pad36.2601131206.ckpt.he30k
40
+ --model.noise-scheduler=flow
41
+ --model.train-diffusion-steps=1000
42
+ --model.n_conditions=0
43
+ --model.action-chunk-size=30
44
+ --model.action-dim=36
45
+ --model.action-exec-horizon=30
46
+ --model.observation-horizon=1
47
+ --model.odim=36
48
+ --model.view_feature_dim=2048
49
+ --model.no-tune-vlm
50
+ --model.no-use_film
51
+ --model.no-combined_temb
52
+ --model.rtc
53
+ --model.max-delay=8
psi0/simple-checkpoints/g1wholebodypushofficechairteleop-v0.simple.flow1000.cosine.lr1.0e-04.b128.gpus8.2605291520/checkpoints/ckpt_40000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ffda4172d813da22d8fe7cb32d0d52c2ab77042da1b1edd11a16bfcc05113b73
3
+ size 6253648840
psi0/simple-checkpoints/g1wholebodypushofficechairteleop-v0.simple.flow1000.cosine.lr1.0e-04.b128.gpus8.2605291520/envs.txt ADDED
@@ -0,0 +1,17 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ OMP_NUM_THREADS=32
2
+ HF_HOME=/data/cache
3
+ TORCH_HOME=/data/cache
4
+ HF_TOKEN=hf_...TiKa
5
+ HF_LEROBOT_HOME=/data/data/lerobot
6
+ WE_HOME=Not Set
7
+ DATA_HOME=/data/data
8
+ UV_CACHE_DIR=/data/cache
9
+ WANDB_API_KEY=90e...5c06
10
+ PROTOCOL_BUFFERS_PYTHON_IMPLEMENTATION=python
11
+ CUDA_VISIBLE_DEVICES=0,1,2,3,4,5,6,7
12
+ WORLD_SIZE=8
13
+ LOCAL_WORLD_SIZE=8
14
+ RANK=0
15
+ LOCAL_RANK=0
16
+ MASTER_ADDR=127.0.0.1
17
+ MASTER_PORT=29508
psi0/simple-checkpoints/g1wholebodypushofficechairteleop-v0.simple.flow1000.cosine.lr1.0e-04.b128.gpus8.2605291520/run_config.json ADDED
@@ -0,0 +1,348 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "exp": "g1wholebodypushofficechairteleop-v0",
3
+ "seed": 292285,
4
+ "auto_tag_run": false,
5
+ "eval": false,
6
+ "debug": false,
7
+ "timestamp": "2605291520",
8
+ "log": {
9
+ "logging_dir": "logs",
10
+ "report_to": "wandb",
11
+ "log_freq": 100
12
+ },
13
+ "wandb": {
14
+ "project": "psi",
15
+ "entity": "jliu530-soochow-university",
16
+ "group": "finetune",
17
+ "id": "8zqzxaew",
18
+ "name": "g1wholebodypushofficechairteleop-v0.simple.flow1000.cosine.lr1.0e-04.b128.gpus8.2605291520",
19
+ "resume": "allow"
20
+ },
21
+ "train": {
22
+ "num_workers": 8,
23
+ "overfit_single_batch": false,
24
+ "name": "finetune",
25
+ "resume_from_checkpoint": null,
26
+ "skip_resumed_steps": false,
27
+ "hf_token": ".hf_token",
28
+ "lora": false,
29
+ "output_dir": ".runs",
30
+ "gradient_accumulation_steps": 1,
31
+ "mixed_precision": "bf16",
32
+ "max_grad_norm": 1.0,
33
+ "optimizer_foreach": null,
34
+ "train_batch_size": 16,
35
+ "val_batch_size": 16,
36
+ "val_num_batches": 20,
37
+ "checkpointing_steps": 10000,
38
+ "max_checkpoints_to_keep": 5,
39
+ "validation_steps": 500,
40
+ "learning_rate": 0.0001,
41
+ "lr_scheduler_type": "cosine",
42
+ "lr_scheduler_kwargs": {
43
+ "betas": [
44
+ 0.95,
45
+ 0.999
46
+ ],
47
+ "weight_decay": 1e-6,
48
+ "eps": 1e-8
49
+ },
50
+ "scheduler_specific_kwargs": {},
51
+ "data_parallel": "ddp",
52
+ "sharding_strategy": "full-shard",
53
+ "deepspeed_config": "/data/jliu/psi/src/InternVLA/config/deepseeds/zero3.json",
54
+ "enable_gradient_checkpointing": true,
55
+ "enable_mixed_precision_training": true,
56
+ "reduce_in_full_precision": true,
57
+ "max_training_steps": 40000,
58
+ "num_train_epochs": null,
59
+ "warmup_steps": 1000,
60
+ "warmup_ratio": null
61
+ },
62
+ "data": {
63
+ "transform": {
64
+ "repack": {
65
+ "dataset_name": "simple",
66
+ "num_past_frames": 0,
67
+ "action_chunk_size": 30,
68
+ "pad_action_dim": 36,
69
+ "pad_state_dim": 36
70
+ },
71
+ "model": {
72
+ "resize": {
73
+ "size": [
74
+ 180,
75
+ 320
76
+ ]
77
+ },
78
+ "center_crop": {
79
+ "size": [
80
+ 180,
81
+ 320
82
+ ]
83
+ },
84
+ "color_jitter": {
85
+ "brightness": 0.2,
86
+ "contrast": [
87
+ 0.8,
88
+ 1.2
89
+ ],
90
+ "saturation": [
91
+ 0.8,
92
+ 1.2
93
+ ],
94
+ "hue": 0.05
95
+ },
96
+ "gaussian_noise": {
97
+ "mean": 0.0,
98
+ "std": 3.0,
99
+ "prob_skip": 0.1
100
+ },
101
+ "img_aug": true,
102
+ "adaptive_resize": false,
103
+ "img_sizes": {
104
+ "egodex": [
105
+ 270,
106
+ 480
107
+ ],
108
+ "he": [
109
+ 240,
110
+ 320
111
+ ]
112
+ }
113
+ },
114
+ "field": {
115
+ "stat_path": "meta/stats_psi0.json",
116
+ "action_norm_type": "bounds",
117
+ "stat_action_key": "action",
118
+ "stat_state_key": "states",
119
+ "use_norm_mask": false,
120
+ "action_norm_masks": [
121
+ true,
122
+ true,
123
+ true,
124
+ true,
125
+ true,
126
+ true,
127
+ false
128
+ ],
129
+ "action_min": [
130
+ -0.5,
131
+ -1.0982101121996204e-16,
132
+ -1.0982101121996204e-16,
133
+ -0.6000000238418579,
134
+ -1.5,
135
+ -1.5,
136
+ -1.5,
137
+ -0.5,
138
+ -0.699999988079071,
139
+ -0.699999988079071,
140
+ -2.202738163743472e-16,
141
+ -2.202738163743472e-16,
142
+ -2.39902178276291e-19,
143
+ -2.202738163743472e-16,
144
+ -0.4062299132347107,
145
+ 0.1900009959936142,
146
+ -0.7789390087127686,
147
+ -1.047199010848999,
148
+ -0.5836427807807922,
149
+ -0.9250030517578125,
150
+ -0.7429955005645752,
151
+ -0.6403748989105225,
152
+ -0.9335375428199768,
153
+ -0.958408772945404,
154
+ -0.9901157021522522,
155
+ -1.2390432357788086,
156
+ -0.9699789881706238,
157
+ -0.8816301226615906,
158
+ -0.15432624518871307,
159
+ -0.22572921216487885,
160
+ -0.3857472538948059,
161
+ 0.7400000095367432,
162
+ -0.4913194477558136,
163
+ -0.4218577742576599,
164
+ -0.5306519865989685,
165
+ 0.0
166
+ ],
167
+ "action_max": [
168
+ 0.0,
169
+ 0.699999988079071,
170
+ 0.699999988079071,
171
+ 0.0,
172
+ 2.1940919425854756e-16,
173
+ 2.1940919425854756e-16,
174
+ 2.1940919425854756e-16,
175
+ 2.39902178276291e-19,
176
+ 1.0987941678865319e-16,
177
+ 1.0987941678865319e-16,
178
+ 1.5,
179
+ 1.5,
180
+ 0.6000000238418579,
181
+ 1.5,
182
+ 0.13131169974803925,
183
+ 0.4889023005962372,
184
+ 0.5719403028488159,
185
+ 0.350136399269104,
186
+ 1.186266303062439,
187
+ 0.4306658208370209,
188
+ 0.7729097604751587,
189
+ 0.2789570093154907,
190
+ -0.1900009959936142,
191
+ 0.5411633849143982,
192
+ 0.2890302240848541,
193
+ 1.4183895587921143,
194
+ 0.8457256555557251,
195
+ 1.6144285202026367,
196
+ 0.08479604125022888,
197
+ 0.12928254902362823,
198
+ 0.3816736936569214,
199
+ 0.7400000095367432,
200
+ 0.5,
201
+ 0.5,
202
+ 0.0,
203
+ 0.3266666531562805
204
+ ],
205
+ "state_min": [
206
+ -0.7087668776512146,
207
+ -0.244631826877594,
208
+ -0.003763168351724744,
209
+ -1.465746521949768,
210
+ -1.4567368030548096,
211
+ -0.5708609819412231,
212
+ -1.4978446960449219,
213
+ -0.7669860124588013,
214
+ -0.9128449559211731,
215
+ -0.7286460399627686,
216
+ -0.0023730271495878696,
217
+ -0.002414399990811944,
218
+ -0.004012954421341419,
219
+ -0.0013138725189492106,
220
+ -0.3674890100955963,
221
+ 0.15613150596618652,
222
+ -0.838959813117981,
223
+ -1.0045318603515625,
224
+ -0.5777662396430969,
225
+ -0.9401854872703552,
226
+ -0.7369406223297119,
227
+ -0.5856585502624512,
228
+ -0.8684520721435547,
229
+ -0.9462044835090637,
230
+ -0.9504987001419067,
231
+ -1.217966914176941,
232
+ -0.9431585073471069,
233
+ -0.9075380563735962,
234
+ -0.14174017310142517,
235
+ -0.1582687944173813,
236
+ -0.31057488918304443,
237
+ 0.7400000095367432,
238
+ 0.0,
239
+ 0.0,
240
+ 0.0,
241
+ 0.0
242
+ ],
243
+ "state_max": [
244
+ 0.0037003702018409967,
245
+ 0.7117500901222229,
246
+ 0.7108933329582214,
247
+ 0.0036252259742468596,
248
+ 0.002007015747949481,
249
+ 0.0016742885345593095,
250
+ 0.0008630083757452667,
251
+ 0.014325465075671673,
252
+ 0.7278810143470764,
253
+ 0.0025119679048657417,
254
+ 0.5710287690162659,
255
+ 1.5086101293563843,
256
+ 1.4543108940124512,
257
+ 1.550452709197998,
258
+ 0.12096124142408371,
259
+ 0.4521321654319763,
260
+ 0.5511183738708496,
261
+ 0.34373360872268677,
262
+ 1.1812156438827515,
263
+ 0.46068504452705383,
264
+ 0.7682955265045166,
265
+ 0.27862611413002014,
266
+ -0.1587948203086853,
267
+ 0.4062543511390686,
268
+ 0.35437384247779846,
269
+ 1.4015804529190063,
270
+ 0.852665901184082,
271
+ 1.6158336400985718,
272
+ 0.19795437157154083,
273
+ 0.19742734730243683,
274
+ 0.4713851511478424,
275
+ 0.7400000095367432,
276
+ 0.0,
277
+ 0.0,
278
+ 0.0,
279
+ 0.0
280
+ ],
281
+ "normalize_state": true,
282
+ "pad_action_dim": 36,
283
+ "pad_state_dim": 36
284
+ }
285
+ },
286
+ "root_dir": "/data/jliu/data",
287
+ "train_repo_ids": [
288
+ "G1WholebodyPushOfficeChairTeleop-v0"
289
+ ],
290
+ "val_repo_ids": [
291
+ "G1WholebodyPushOfficeChairTeleop-v0"
292
+ ]
293
+ },
294
+ "model": {
295
+ "resnet_store_path": null,
296
+ "pretrained_action_header_path": "/hfm/cache/checkpoints/psi0/postpre.1by1.pad36.2601131206.ckpt.he30k",
297
+ "rtc": true,
298
+ "max_delay": 8,
299
+ "action_dim": 36,
300
+ "action_chunk_size": 30,
301
+ "action_exec_horizon": 30,
302
+ "observation_horizon": 1,
303
+ "img_chunk": 1,
304
+ "n_cams": 1,
305
+ "use_obs": "add_token",
306
+ "dropout": 0.1,
307
+ "noise_scheduler": "flow",
308
+ "train_diffusion_steps": 1000,
309
+ "eval_diffusion_steps": 10,
310
+ "share_cam_features": false,
311
+ "early_fusion": false,
312
+ "odim": 36,
313
+ "n_conditions": 0,
314
+ "token_fusion": "concat",
315
+ "loss_w": [
316
+ 0.1,
317
+ 0.2,
318
+ 0.1
319
+ ],
320
+ "time_dim": 256,
321
+ "hidden_dim": 1536,
322
+ "num_blocks": 6,
323
+ "dim_feedforward": 2048,
324
+ "nhead": 24,
325
+ "activation": "gelu",
326
+ "view_feature_dim": 2048,
327
+ "use_film": false,
328
+ "combined_temb": false,
329
+ "use_dit": false,
330
+ "weight_decay": 0.01,
331
+ "model_name_or_path": "/hfm/cache/checkpoints/psi0/pre.fast.1by1.2601091803.ckpt.ego200k.he30k",
332
+ "vlm_ckpt_step": null,
333
+ "tune_vlm": false,
334
+ "tune_mm_llm": false,
335
+ "tune_mm_vision": false,
336
+ "tune_mm_mlp": false,
337
+ "gradient_checkpointing": true,
338
+ "lang_backbone_lr": 0.00001,
339
+ "mm_projector_lr": 0.00001,
340
+ "vision_tower_lr": 1e-6,
341
+ "optim": "adamw_torch",
342
+ "model_max_length": 4096,
343
+ "data_flatten": true,
344
+ "data_packing": true,
345
+ "max_pixels": 451584,
346
+ "min_pixels": 12544
347
+ }
348
+ }