Hugging Face's logo

nnh-pbbb
/
internpi

Model card Files
xet
 Community
nnh-pbbb commited on
Commit
8d6614d
·
verified ·
1 Parent(s): 2e6d79d

Upload folder using huggingface_hub

Browse files
Files changed (3) hide show
  1. internpi-0127-eep-small/config.json +106 -0
  2. internpi-0127-eep-small/config.yaml +67 -0
  3. internpi-0127-eep-small/dataset_statistics.json +218 -0
internpi-0127-eep-small/config.json ADDED
@@ -0,0 +1,106 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "run_id": "InternVLPI-0127-eep-test",
3
+ "run_root_dir": "/lumos-vePFS/shenzhen/models/lumos_pi",
4
+ "seed": 42,
5
+ "trackers": [
6
+ "jsonl"
7
+ ],
8
+ "wandb_entity": null,
9
+ "wandb_project": null,
10
+ "is_debug": false,
11
+ "framework": {
12
+ "name": "internvlPI",
13
+ "qwenvl": {
14
+ "base_vlm": "/lumos-vePFS/shenzhen/models/InternVL3_5-1B-Instruct",
15
+ "attn_implementation": "eager",
16
+ "vl_hidden_dim": 2048
17
+ },
18
+ "action_model": {
19
+ "add_pos_embed": true,
20
+ "hidden_size": 1024,
21
+ "max_seq_len": 1024,
22
+ "action_model_type": "DiT-B",
23
+ "action_dim": 20,
24
+ "state_dim": 20,
25
+ "future_action_window_size": 15,
26
+ "past_action_window_size": 0,
27
+ "action_horizon": 30,
28
+ "repeated_diffusion_steps": 2,
29
+ "noise_beta_alpha": 1.5,
30
+ "noise_beta_beta": 1.0,
31
+ "noise_s": 0.999,
32
+ "num_timestep_buckets": 1000,
33
+ "num_inference_timesteps": 8,
34
+ "num_target_vision_tokens": 32,
35
+ "use_scaled_noise": true,
36
+ "diffusion_model_cfg": {
37
+ "dropout": 0.2,
38
+ "final_dropout": true,
39
+ "interleave_self_attention": true,
40
+ "norm_type": "ada_norm",
41
+ "num_layers": 28,
42
+ "output_dim": 2560,
43
+ "positional_embeddings": null
44
+ },
45
+ "smoothness_loss_weight": 1
46
+ }
47
+ },
48
+ "datasets": {
49
+ "vla_data": {
50
+ "dataset_py": "lerobot_datasets",
51
+ "data_root_dir": "/lumos-vePFS/shenzhen/data/data_warehouse/data_warehouse_output",
52
+ "data_mix": "train_QwenPI",
53
+ "per_device_batch_size": 4,
54
+ "num_workers": 2,
55
+ "action_type": "abs_ee",
56
+ "image_size": [
57
+ 480,
58
+ 360
59
+ ],
60
+ "video_backend": "pyav"
61
+ }
62
+ },
63
+ "trainer": {
64
+ "epochs": 100,
65
+ "max_train_steps": 35000,
66
+ "num_warmup_steps": 5000,
67
+ "save_interval": 1000,
68
+ "eval_interval": 100000,
69
+ "learning_rate": {
70
+ "base": 1e-05,
71
+ "intern_vl_interface": 1e-05,
72
+ "action_model": 0.0001
73
+ },
74
+ "lr_scheduler_type": "cosine_with_min_lr",
75
+ "scheduler_specific_kwargs": {
76
+ "min_lr": 1e-06
77
+ },
78
+ "freeze_modules": null,
79
+ "loss_scale": {
80
+ "vla": 1.0,
81
+ "vlm": 0.05
82
+ },
83
+ "max_grad_norm": 1.0,
84
+ "warmup_ratio": 0.1,
85
+ "weight_decay": 0.0,
86
+ "logging_frequency": 10,
87
+ "gradient_clipping": 1.0,
88
+ "gradient_accumulation_steps": 2,
89
+ "optimizer": {
90
+ "name": "AdamW",
91
+ "betas": [
92
+ 0.9,
93
+ 0.95
94
+ ],
95
+ "eps": 1e-08,
96
+ "weight_decay": 0
97
+ }
98
+ },
99
+ "is_resume": true,
100
+ "resume_epoch": null,
101
+ "resume_step": 5000,
102
+ "enable_gradient_checkpointing": true,
103
+ "enable_mixed_precision_training": true,
104
+ "report_to": "jsonl",
105
+ "output_dir": "/lumos-vePFS/shenzhen/models/lumos_pi/InternVLPI-0127-eep-test"
106
+ }
internpi-0127-eep-small/config.yaml ADDED
@@ -0,0 +1,67 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ datasets:
2
+ vla_data:
3
+ data_mix: train_QwenPI
4
+ data_root_dir: /lumos-vePFS/shenzhen/data/data_warehouse/data_warehouse_output
5
+ dataset_py: lerobot_datasets
6
+ image_size:
7
+ - 480
8
+ - 360
9
+ per_device_batch_size: 4
10
+ video_backend: pyav
11
+ framework:
12
+ action_model:
13
+ action_dim: 20
14
+ add_pos_embed: true
15
+ diffusion_model_cfg:
16
+ dropout: 0.2
17
+ final_dropout: true
18
+ interleave_self_attention: true
19
+ norm_type: ada_norm
20
+ num_layers: 28
21
+ output_dim: 2560
22
+ positional_embeddings: null
23
+ future_action_window_size: 15
24
+ hidden_dim: 1024
25
+ max_seq_len: 1024
26
+ noise_beta_alpha: 1.5
27
+ noise_beta_beta: 1.0
28
+ noise_s: 0.999
29
+ num_inference_timesteps: 8
30
+ num_target_vision_tokens: 32
31
+ num_timestep_buckets: 1000
32
+ past_action_window_size: 0
33
+ smoothness_loss_weight: 1
34
+ state_dim: 20
35
+ use_scaled_noise: true
36
+ name: internvlPI
37
+ qwenvl:
38
+ attn_implementation: eager
39
+ base_vlm: /lumos-vePFS/shenzhen/models/InternVL3_5-1B-Instruct
40
+ num_vl_layers: 28
41
+ vl_hidden_dim: 1024
42
+ output_dir: /lumos-vePFS/shenzhen/models/lumos_pi/InternVLPI-0127-eep-test
43
+ run_id: InternVLPI-0127-eep-test
44
+ run_root_dir: /lumos-vePFS/shenzhen/models/lumos_pi
45
+ seed: 42
46
+ trainer:
47
+ eval_interval: 100000
48
+ freeze_modules: null
49
+ gradient_accumulation_steps: 2
50
+ gradient_clipping: 1.0
51
+ learning_rate:
52
+ action_model: 0.0001
53
+ base: 1.0e-05
54
+ intern_vl_interface: 1.0e-05
55
+ logging_frequency: 10
56
+ lr_scheduler_type: cosine_with_min_lr
57
+ max_train_steps: 35000
58
+ num_warmup_steps: 5000
59
+ optimizer:
60
+ betas:
61
+ - 0.9
62
+ - 0.95
63
+ eps: 1.0e-08
64
+ weight_decay: 0
65
+ save_interval: 1000
66
+ scheduler_specific_kwargs:
67
+ min_lr: 1.0e-06
internpi-0127-eep-small/dataset_statistics.json ADDED
@@ -0,0 +1,218 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "new_embodiment": {
3
+ "action": {
4
+ "mean": [
5
+ 0.24831613898277283,
6
+ 0.027144895866513252,
7
+ 0.22650116682052612,
8
+ -0.8404249548912048,
9
+ 0.5109900832176208,
10
+ 0.3438127040863037,
11
+ 0.05076254904270172,
12
+ 0.27124547958374023,
13
+ 0.1341637223958969,
14
+ 0.23118293285369873,
15
+ 0.5084952116012573,
16
+ 0.7631425261497498,
17
+ -1.8881397247314453,
18
+ 0.04549280181527138
19
+ ],
20
+ "std": [
21
+ 0.08048900961875916,
22
+ 0.11851683259010315,
23
+ 0.05085735768079758,
24
+ 2.8058412075042725,
25
+ 0.3912830650806427,
26
+ 2.907383441925049,
27
+ 0.02002296969294548,
28
+ 0.10783883184194565,
29
+ 0.09286945313215256,
30
+ 0.05440365895628929,
31
+ 2.881397008895874,
32
+ 0.2646803855895996,
33
+ 1.884290337562561,
34
+ 0.023190317675471306
35
+ ],
36
+ "max": [
37
+ 0.4355350136756897,
38
+ 0.42503198981285095,
39
+ 0.4411650002002716,
40
+ 3.1415927410125732,
41
+ 1.5707963705062866,
42
+ 3.1415927410125732,
43
+ 0.059709999710321426,
44
+ 0.528249979019165,
45
+ 0.4351179897785187,
46
+ 0.4433180093765259,
47
+ 3.1415927410125732,
48
+ 1.5707963705062866,
49
+ 3.1415927410125732,
50
+ 0.05992000177502632
51
+ ],
52
+ "min": [
53
+ 0.042451001703739166,
54
+ -0.21000799536705017,
55
+ 0.11678700149059296,
56
+ -3.1410341262817383,
57
+ -0.282882958650589,
58
+ -3.141016721725464,
59
+ 0.0005600000149570405,
60
+ 0.0038419999182224274,
61
+ -0.10948000103235245,
62
+ 0.09222699701786041,
63
+ -3.1410515308380127,
64
+ 0.1244419738650322,
65
+ -3.1410515308380127,
66
+ 7.000000186963007e-05
67
+ ],
68
+ "q01": [
69
+ 0.05619199946522713,
70
+ -0.17355053395032882,
71
+ 0.13174245327711107,
72
+ -3.133350660800934,
73
+ -0.1359262466430664,
74
+ -3.1331721353530884,
75
+ 0.000699999975040555,
76
+ 0.05602800101041794,
77
+ -0.03544915910810232,
78
+ 0.11746856845915318,
79
+ -3.1342583203315737,
80
+ 0.2646872201561928,
81
+ -3.1262485551834107,
82
+ 0.0002800000074785203
83
+ ],
84
+ "q99": [
85
+ 0.3960674390196795,
86
+ 0.35788134992122633,
87
+ 0.3558399733901022,
88
+ 3.1415927410125732,
89
+ 1.5315977966785421,
90
+ 3.1415927410125732,
91
+ 0.05964000150561333,
92
+ 0.48414370894432057,
93
+ 0.3517844846844672,
94
+ 0.36821531414985653,
95
+ 3.1415927410125732,
96
+ 1.5277097356319418,
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
+ 0.24831613898277283,
120
+ 0.027144895866513252,
121
+ 0.22650116682052612,
122
+ -0.8404249548912048,
123
+ 0.5109900832176208,
124
+ 0.3438127040863037,
125
+ 0.05076254904270172,
126
+ 0.27124547958374023,
127
+ 0.1341637223958969,
128
+ 0.23118293285369873,
129
+ 0.5084952116012573,
130
+ 0.7631425261497498,
131
+ -1.8881397247314453,
132
+ 0.04549280181527138
133
+ ],
134
+ "std": [
135
+ 0.08048900961875916,
136
+ 0.11851683259010315,
137
+ 0.05085735768079758,
138
+ 2.8058412075042725,
139
+ 0.3912830650806427,
140
+ 2.907383441925049,
141
+ 0.02002296969294548,
142
+ 0.10783883184194565,
143
+ 0.09286945313215256,
144
+ 0.05440365895628929,
145
+ 2.881397008895874,
146
+ 0.2646803855895996,
147
+ 1.884290337562561,
148
+ 0.023190317675471306
149
+ ],
150
+ "max": [
151
+ 0.4355350136756897,
152
+ 0.42503198981285095,
153
+ 0.4411650002002716,
154
+ 3.1415927410125732,
155
+ 1.5707963705062866,
156
+ 3.1415927410125732,
157
+ 0.059709999710321426,
158
+ 0.528249979019165,
159
+ 0.4351179897785187,
160
+ 0.4433180093765259,
161
+ 3.1415927410125732,
162
+ 1.5707963705062866,
163
+ 3.1415927410125732,
164
+ 0.05992000177502632
165
+ ],
166
+ "min": [
167
+ 0.042451001703739166,
168
+ -0.21000799536705017,
169
+ 0.11678700149059296,
170
+ -3.1410341262817383,
171
+ -0.282882958650589,
172
+ -3.141016721725464,
173
+ 0.0005600000149570405,
174
+ 0.0038419999182224274,
175
+ -0.10948000103235245,
176
+ 0.09222699701786041,
177
+ -3.1410515308380127,
178
+ 0.1244419738650322,
179
+ -3.1410515308380127,
180
+ 7.000000186963007e-05
181
+ ],
182
+ "q01": [
183
+ 0.05619199946522713,
184
+ -0.17355053395032882,
185
+ 0.13174245327711107,
186
+ -3.133350660800934,
187
+ -0.1359262466430664,
188
+ -3.1331721353530884,
189
+ 0.000699999975040555,
190
+ 0.05602800101041794,
191
+ -0.03544915910810232,
192
+ 0.11746856845915318,
193
+ -3.1342583203315737,
194
+ 0.2646872201561928,
195
+ -3.1262485551834107,
196
+ 0.0002800000074785203
197
+ ],
198
+ "q99": [
199
+ 0.3960674390196795,
200
+ 0.35788134992122633,
201
+ 0.3558399733901022,
202
+ 3.1415927410125732,
203
+ 1.5315977966785421,
204
+ 3.1415927410125732,
205
+ 0.05964000150561333,
206
+ 0.48414370894432057,
207
+ 0.3517844846844672,
208
+ 0.36821531414985653,
209
+ 3.1415927410125732,
210
+ 1.5277097356319418,
211
+ 3.1415927410125732,
212
+ 0.05992000177502632
213
+ ]
214
+ },
215
+ "num_transitions": 34124,
216
+ "num_trajectories": 52
217
+ }
218
+ }