Hugging Face's logo

Chengpengaaa
/
starvla-adapter

Model card Files
xet
 Community
Chengpengaaa commited on
Commit
8ed511e
·
verified ·
1 Parent(s): 07f1842

Upload folder using huggingface_hub

Browse files
Files changed (4) hide show
  1. internvl-0123v/checkpoints/steps_85000_pytorch_model.pt +3 -0
  2. internvl-0123v/config.json +88 -0
  3. internvl-0123v/config.yaml +48 -0
  4. internvl-0123v/dataset_statistics.json +218 -0
internvl-0123v/checkpoints/steps_85000_pytorch_model.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5e2c2dfc6ad2b6e1443392b50ca164e613b982d5d68f8afdeda2fbbf7ab95f17
3
+ size 2620088697
internvl-0123v/config.json ADDED
@@ -0,0 +1,88 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "run_id": "internvlAdapter-0123v-eep",
3
+ "run_root_dir": "/root/jinchengp/LumosVLA/results",
4
+ "seed": 42,
5
+ "trackers": [
6
+ "jsonl",
7
+ "wandb"
8
+ ],
9
+ "wandb_entity": "xspadex-nanyang-technological-university-singapore",
10
+ "wandb_project": "internvl-0123v-eep",
11
+ "is_debug": false,
12
+ "framework": {
13
+ "name": "internvlAdapter",
14
+ "qwenvl": {
15
+ "base_vlm": "/root/jinchengp/LumosVLA/base_vl/internvl",
16
+ "attn_implementation": "flash_attention_2",
17
+ "vl_hidden_dim": 2048
18
+ },
19
+ "action_model": {
20
+ "action_model_type": "VLA-Adapter",
21
+ "hidden_dim": 2048,
22
+ "action_dim": 20,
23
+ "state_dim": 20,
24
+ "action_query_num": 64,
25
+ "use_pro_version": true,
26
+ "use_proprio": true,
27
+ "phase": "Training",
28
+ "num_actions_chunk": 30
29
+ },
30
+ "framework_py": "internvlAdapter"
31
+ },
32
+ "datasets": {
33
+ "vla_data": {
34
+ "dataset_py": "lerobot_datasets",
35
+ "data_root_dir": "/lumos-vePFS/shenzhen/data/data_warehouse/data_warehouse_output",
36
+ "data_mix": "train_internAdapter",
37
+ "per_device_batch_size": 32,
38
+ "num_workers": 4,
39
+ "action_type": "abs_ee",
40
+ "image_size": 256,
41
+ "video_backend": "pyav"
42
+ }
43
+ },
44
+ "trainer": {
45
+ "epochs": 100,
46
+ "max_train_steps": 85000,
47
+ "num_warmup_steps": 1000,
48
+ "save_interval": 5000,
49
+ "eval_interval": 100000,
50
+ "learning_rate": {
51
+ "base": 4e-06,
52
+ "intern_vl_interface": 4e-06,
53
+ "action_model": 1e-05
54
+ },
55
+ "lr_scheduler_type": "cosine_with_min_lr",
56
+ "scheduler_specific_kwargs": {
57
+ "min_lr": 5e-07
58
+ },
59
+ "freeze_modules": "intern_vl_interface.model.language_model",
60
+ "loss_scale": {
61
+ "vla": 1.0,
62
+ "vlm": 0.1
63
+ },
64
+ "max_grad_norm": 1.0,
65
+ "warmup_ratio": 0.1,
66
+ "weight_decay": 0.0,
67
+ "logging_frequency": 10,
68
+ "gradient_clipping": 1.0,
69
+ "gradient_accumulation_steps": 1,
70
+ "optimizer": {
71
+ "name": "AdamW",
72
+ "betas": [
73
+ 0.9,
74
+ 0.95
75
+ ],
76
+ "eps": 1e-08,
77
+ "weight_decay": 1e-08
78
+ }
79
+ },
80
+ "is_resume": false,
81
+ "resume_epoch": null,
82
+ "resume_step": null,
83
+ "enable_gradient_checkpointing": true,
84
+ "enable_mixed_precision_training": true,
85
+ "report_to": "wandb",
86
+ "wandb_run_name": "v1",
87
+ "output_dir": "/root/jinchengp/LumosVLA/results/internvlAdapter-0123v-eep"
88
+ }
internvl-0123v/config.yaml ADDED
@@ -0,0 +1,48 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ datasets:
2
+ vla_data:
3
+ data_mix: train_internAdapter
4
+ data_root_dir: /lumos-vePFS/shenzhen/data/data_warehouse/data_warehouse_output
5
+ dataset_py: lerobot_datasets
6
+ image_size: 256
7
+ per_device_batch_size: 32
8
+ video_backend: pyav
9
+ framework:
10
+ action_model:
11
+ action_dim: 20
12
+ action_query_num: 64
13
+ hidden_dim: 1024
14
+ num_actions_chunk: 30
15
+ phase: Training
16
+ state_dim: 20
17
+ use_pro_version: true
18
+ use_proprio: true
19
+ name: internvlAdapter
20
+ qwenvl:
21
+ base_vlm: /root/jinchengp/LumosVLA/base_vl/internvl
22
+ vl_hidden_dim: 1024
23
+ output_dir: /root/jinchengp/LumosVLA/results/internvlAdapter-0123v-eep
24
+ run_id: internvlAdapter-0123v-eep
25
+ run_root_dir: /root/jinchengp/LumosVLA/results
26
+ seed: 42
27
+ trainer:
28
+ eval_interval: 100000
29
+ freeze_modules: intern_vl_interface.model.language_model
30
+ gradient_accumulation_steps: 1
31
+ gradient_clipping: 1.0
32
+ learning_rate:
33
+ action_model: 1.0e-05
34
+ base: 4.0e-06
35
+ intern_vl_interface: 4.0e-06
36
+ logging_frequency: 10
37
+ lr_scheduler_type: cosine_with_min_lr
38
+ max_train_steps: 85000
39
+ num_warmup_steps: 1000
40
+ optimizer:
41
+ betas:
42
+ - 0.9
43
+ - 0.95
44
+ eps: 1.0e-08
45
+ weight_decay: 1.0e-08
46
+ save_interval: 5000
47
+ scheduler_specific_kwargs:
48
+ min_lr: 5.0e-07
internvl-0123v/dataset_statistics.json ADDED
@@ -0,0 +1,218 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "new_embodiment": {
3
+ "action": {
4
+ "mean": [
5
+ 0.23995271011402736,
6
+ -0.06203787953065906,
7
+ 0.20739578024337169,
8
+ -1.3512385056207057,
9
+ 0.4042915521483673,
10
+ 0.9000311023310614,
11
+ 0.039615929911011157,
12
+ 0.24132249935677183,
13
+ 0.09746678467643889,
14
+ 0.21104620397090915,
15
+ 1.2572790937204115,
16
+ 0.5353348255157471,
17
+ -1.7396099034108616,
18
+ 0.03882587279536224
19
+ ],
20
+ "std": [
21
+ 0.08994903988381915,
22
+ 0.12128905753062412,
23
+ 0.07054434849407303,
24
+ 2.5913290385321748,
25
+ 0.35082715580144624,
26
+ 2.74726566787674,
27
+ 0.026358133686756022,
28
+ 0.09577926520263415,
29
+ 0.07878744378135882,
30
+ 0.05945691873630537,
31
+ 2.6677117240484702,
32
+ 0.3140146100237212,
33
+ 2.200234565003498,
34
+ 0.026658042007117622
35
+ ],
36
+ "max": [
37
+ 0.5393329858779907,
38
+ 0.48442599177360535,
39
+ 0.6720470190048218,
40
+ 3.1415927410125732,
41
+ 1.5707963705062866,
42
+ 3.1415927410125732,
43
+ 0.06384000182151794,
44
+ 0.5695070028305054,
45
+ 0.45939499139785767,
46
+ 0.6456500291824341,
47
+ 3.1415927410125732,
48
+ 1.5707963705062866,
49
+ 3.1415927410125732,
50
+ 0.060339998453855515
51
+ ],
52
+ "min": [
53
+ -0.10788500308990479,
54
+ -0.5191159844398499,
55
+ 0.05765699967741966,
56
+ -3.1410515308380127,
57
+ -1.0231120586395264,
58
+ -3.1410515308380127,
59
+ -0.0017500000540167093,
60
+ -0.03119499981403351,
61
+ -0.26581600308418274,
62
+ 0.07246199995279312,
63
+ -3.1410515308380127,
64
+ -0.6304129362106323,
65
+ -3.1410515308380127,
66
+ 0.0
67
+ ],
68
+ "q01": [
69
+ 0.03489213060587645,
70
+ -0.36912800788879396,
71
+ 0.08692353814840317,
72
+ -3.137159585952759,
73
+ -0.20070483744144438,
74
+ -3.137392144203186,
75
+ 0.0,
76
+ 0.030881709717214108,
77
+ -0.15914200246334076,
78
+ 0.10170399993658066,
79
+ -3.138171911239624,
80
+ -0.11282174646854401,
81
+ -3.1362587141990663,
82
+ 7.000000186963007e-05
83
+ ],
84
+ "q99": [
85
+ 0.46800449550151807,
86
+ 0.41213971734046934,
87
+ 0.43853099942207385,
88
+ 3.1415927410125732,
89
+ 1.5365502309799193,
90
+ 3.1415927410125732,
91
+ 0.06019999831914902,
92
+ 0.5246558457612991,
93
+ 0.3517844846844672,
94
+ 0.4130769228935239,
95
+ 3.1415927410125732,
96
+ 1.5435859155654907,
97
+ 3.1415927410125732,
98
+ 0.05992000177502632
99
+ ],
100
+ "mask": [
101
+ true,
102
+ true,
103
+ true,
104
+ true,
105
+ true,
106
+ true,
107
+ false,
108
+ true,
109
+ true,
110
+ true,
111
+ true,
112
+ true,
113
+ true,
114
+ false
115
+ ]
116
+ },
117
+ "state": {
118
+ "mean": [
119
+ 0.23995271011402736,
120
+ -0.06203787953065906,
121
+ 0.20739578024337169,
122
+ -1.3512385056207057,
123
+ 0.4042915521483673,
124
+ 0.9000311023310614,
125
+ 0.039615929911011157,
126
+ 0.24132249935677183,
127
+ 0.09746678467643889,
128
+ 0.21104620397090915,
129
+ 1.2572790937204115,
130
+ 0.5353348255157471,
131
+ -1.7396099034108616,
132
+ 0.03882587279536224
133
+ ],
134
+ "std": [
135
+ 0.08994903988381915,
136
+ 0.12128905753062412,
137
+ 0.07054434849407303,
138
+ 2.5913290385321748,
139
+ 0.35082715580144624,
140
+ 2.74726566787674,
141
+ 0.026358133686756022,
142
+ 0.09577926520263415,
143
+ 0.07878744378135882,
144
+ 0.05945691873630537,
145
+ 2.6677117240484702,
146
+ 0.3140146100237212,
147
+ 2.200234565003498,
148
+ 0.026658042007117622
149
+ ],
150
+ "max": [
151
+ 0.5393329858779907,
152
+ 0.48442599177360535,
153
+ 0.6720470190048218,
154
+ 3.1415927410125732,
155
+ 1.5707963705062866,
156
+ 3.1415927410125732,
157
+ 0.06384000182151794,
158
+ 0.5695070028305054,
159
+ 0.45939499139785767,
160
+ 0.6456500291824341,
161
+ 3.1415927410125732,
162
+ 1.5707963705062866,
163
+ 3.1415927410125732,
164
+ 0.060339998453855515
165
+ ],
166
+ "min": [
167
+ -0.10788500308990479,
168
+ -0.5191159844398499,
169
+ 0.05765699967741966,
170
+ -3.1410515308380127,
171
+ -1.0231120586395264,
172
+ -3.1410515308380127,
173
+ -0.0017500000540167093,
174
+ -0.03119499981403351,
175
+ -0.26581600308418274,
176
+ 0.07246199995279312,
177
+ -3.1410515308380127,
178
+ -0.6304129362106323,
179
+ -3.1410515308380127,
180
+ 0.0
181
+ ],
182
+ "q01": [
183
+ 0.03489213060587645,
184
+ -0.36912800788879396,
185
+ 0.08692353814840317,
186
+ -3.137159585952759,
187
+ -0.20070483744144438,
188
+ -3.137392144203186,
189
+ 0.0,
190
+ 0.030881709717214108,
191
+ -0.15914200246334076,
192
+ 0.10170399993658066,
193
+ -3.138171911239624,
194
+ -0.11282174646854401,
195
+ -3.1362587141990663,
196
+ 7.000000186963007e-05
197
+ ],
198
+ "q99": [
199
+ 0.46800449550151807,
200
+ 0.41213971734046934,
201
+ 0.43853099942207385,
202
+ 3.1415927410125732,
203
+ 1.5365502309799193,
204
+ 3.1415927410125732,
205
+ 0.06019999831914902,
206
+ 0.5246558457612991,
207
+ 0.3517844846844672,
208
+ 0.4130769228935239,
209
+ 3.1415927410125732,
210
+ 1.5435859155654907,
211
+ 3.1415927410125732,
212
+ 0.05992000177502632
213
+ ]
214
+ },
215
+ "num_transitions": 1645771,
216
+ "num_trajectories": 1570
217
+ }
218
+ }