Hugging Face's logo

nnh-pbbb
/
internpi

Model card Files
xet
Community
nnh-pbbb commited on
Commit
21745f5
·
verified ·
1 Parent(s): 4d5af47

Upload folder using huggingface_hub

Browse files
Files changed (4) hide show
  1. InternVLPI05-0302-eep_2222_127000/checkpoints/steps_127000_pytorch_model.pt +3 -0
  2. InternVLPI05-0302-eep_2222_127000/config.json +104 -0
  3. InternVLPI05-0302-eep_2222_127000/config.yaml +67 -0
  4. InternVLPI05-0302-eep_2222_127000/dataset_statistics.json +218 -0
InternVLPI05-0302-eep_2222_127000/checkpoints/steps_127000_pytorch_model.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:89285d55289afa919301590c28d52db23a6e38fc7594fb687af2d033b8bf2760
3
+ size 2853231794
InternVLPI05-0302-eep_2222_127000/config.json ADDED
@@ -0,0 +1,104 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "run_id": "InternVLPI05-0302-eep_2222",
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_adaRMS",
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_rms",
41
+ "num_layers": 28,
42
+ "output_dim": 2560,
43
+ "positional_embeddings": null
44
+ },
45
+ "smoothness_loss_weight": 1,
46
+ "use_kv_cache_dit": true
47
+ }
48
+ },
49
+ "datasets": {
50
+ "vla_data": {
51
+ "dataset_py": "lerobot_datasets",
52
+ "data_root_dir": "/lumos-vePFS/shenzhen/data/data_warehouse/data_warehouse_output_clean",
53
+ "data_mix": "train_QwenPI",
54
+ "per_device_batch_size": 16,
55
+ "num_workers": 2,
56
+ "action_type": "abs_ee",
57
+ "image_size": 256,
58
+ "video_backend": "pyav"
59
+ }
60
+ },
61
+ "trainer": {
62
+ "epochs": 100,
63
+ "max_train_steps": 200000,
64
+ "num_warmup_steps": 5000,
65
+ "save_interval": 1000,
66
+ "eval_interval": 10000000,
67
+ "learning_rate": {
68
+ "base": 1e-05,
69
+ "intern_vl_interface": 1e-05,
70
+ "action_model": 0.0001
71
+ },
72
+ "lr_scheduler_type": "cosine_with_min_lr",
73
+ "scheduler_specific_kwargs": {
74
+ "min_lr": 1e-06
75
+ },
76
+ "freeze_modules": null,
77
+ "loss_scale": {
78
+ "vla": 1.0,
79
+ "vlm": 0.05
80
+ },
81
+ "max_grad_norm": 1.0,
82
+ "warmup_ratio": 0.1,
83
+ "weight_decay": 0.0,
84
+ "logging_frequency": 10,
85
+ "gradient_clipping": 1.0,
86
+ "gradient_accumulation_steps": 2,
87
+ "optimizer": {
88
+ "name": "AdamW",
89
+ "betas": [
90
+ 0.9,
91
+ 0.95
92
+ ],
93
+ "eps": 1e-08,
94
+ "weight_decay": 0
95
+ }
96
+ },
97
+ "is_resume": true,
98
+ "resume_epoch": null,
99
+ "resume_step": 5000,
100
+ "enable_gradient_checkpointing": true,
101
+ "enable_mixed_precision_training": true,
102
+ "report_to": "jsonl",
103
+ "output_dir": "/lumos-vePFS/shenzhen/models/lumos_pi/InternVLPI05-0302-eep_2222"
104
+ }
InternVLPI05-0302-eep_2222_127000/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_clean
5
+ dataset_py: lerobot_datasets
6
+ image_size: 256
7
+ num_workers: 2
8
+ per_device_batch_size: 16
9
+ framework:
10
+ action_model:
11
+ action_dim: 20
12
+ add_pos_embed: true
13
+ diffusion_model_cfg:
14
+ cross_attention_dim: 1024
15
+ dropout: 0.2
16
+ final_dropout: true
17
+ interleave_self_attention: true
18
+ norm_elementwise_affine: false
19
+ norm_type: ada_rms
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_kv_cache_dit: true
36
+ use_scaled_noise: true
37
+ name: internvlPI_adaRMS
38
+ qwenvl:
39
+ attn_implementation: eager
40
+ base_vlm: /lumos-vePFS/shenzhen/models/InternVL3_5-1B-Instruct
41
+ num_vl_layers: 28
42
+ vl_hidden_dim: 1024
43
+ output_dir: /lumos-vePFS/shenzhen/models/lumos_pi/InternVLPI05-0302-eep_2222
44
+ run_id: InternVLPI05-0302-eep_2222
45
+ run_root_dir: /lumos-vePFS/shenzhen/models/lumos_pi
46
+ seed: 42
47
+ trainer:
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: 200000
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
InternVLPI05-0302-eep_2222_127000/dataset_statistics.json ADDED
@@ -0,0 +1,218 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "new_embodiment": {
3
+ "action": {
4
+ "mean": [
5
+ 0.2627080635678384,
6
+ -0.06905352499339759,
7
+ 0.19377399452271005,
8
+ -1.1311389931267308,
9
+ 0.4074968476929972,
10
+ 1.5050847338091942,
11
+ 0.03278727958639783,
12
+ 0.2358027420216991,
13
+ 0.08325479407944988,
14
+ 0.209077634157673,
15
+ 1.521457871722598,
16
+ 0.5131264127069904,
17
+ -1.5608658809815683,
18
+ 0.03465066600831285
19
+ ],
20
+ "std": [
21
+ 0.10375088171089163,
22
+ 0.10887929577588172,
23
+ 0.06680644047420393,
24
+ 2.68255043659873,
25
+ 0.3260696396232237,
26
+ 2.3507239802774746,
27
+ 0.028485141733389688,
28
+ 0.09299172799800302,
29
+ 0.081275375183015,
30
+ 0.05484981855789096,
31
+ 2.487506748874059,
32
+ 0.2941890671188146,
33
+ 2.3367093690706113,
34
+ 0.02833451591418635
35
+ ],
36
+ "max": [
37
+ 0.5931400060653687,
38
+ 0.48442599177360535,
39
+ 0.6720470190048218,
40
+ 3.1415927410125732,
41
+ 1.5707963705062866,
42
+ 3.1415927410125732,
43
+ 0.0810599997639656,
44
+ 0.5695070028305054,
45
+ 0.4637550115585327,
46
+ 0.6456500291824341,
47
+ 3.1415927410125732,
48
+ 1.5707963705062866,
49
+ 3.1415927410125732,
50
+ 0.0611100010573864
51
+ ],
52
+ "min": [
53
+ -0.10788500308990479,
54
+ -0.5191159844398499,
55
+ 0.040369000285863876,
56
+ -3.1410515308380127,
57
+ -1.0231120586395264,
58
+ -3.1410515308380127,
59
+ -0.0017500000540167093,
60
+ -0.03119499981403351,
61
+ -0.3384290039539337,
62
+ 0.07246199995279312,
63
+ -3.1410515308380127,
64
+ -0.7706850171089172,
65
+ -3.1410515308380127,
66
+ 0.0
67
+ ],
68
+ "q01": [
69
+ 0.03489213060587645,
70
+ -0.36912800788879396,
71
+ 0.07184209674596786,
72
+ -3.137246325016022,
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
+ 0.5304324626922607,
86
+ 0.41213971734046934,
87
+ 0.44856594711542136,
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
+ 0.2627080635678384,
120
+ -0.06905352499339759,
121
+ 0.19377399452271005,
122
+ -1.1311389931267308,
123
+ 0.4074968476929972,
124
+ 1.5050847338091942,
125
+ 0.03278727958639783,
126
+ 0.2358027420216991,
127
+ 0.08325479407944988,
128
+ 0.209077634157673,
129
+ 1.521457871722598,
130
+ 0.5131264127069904,
131
+ -1.5608658809815683,
132
+ 0.03465066600831285
133
+ ],
134
+ "std": [
135
+ 0.10375088171089163,
136
+ 0.10887929577588172,
137
+ 0.06680644047420393,
138
+ 2.68255043659873,
139
+ 0.3260696396232237,
140
+ 2.3507239802774746,
141
+ 0.028485141733389688,
142
+ 0.09299172799800302,
143
+ 0.081275375183015,
144
+ 0.05484981855789096,
145
+ 2.487506748874059,
146
+ 0.2941890671188146,
147
+ 2.3367093690706113,
148
+ 0.02833451591418635
149
+ ],
150
+ "max": [
151
+ 0.5931400060653687,
152
+ 0.48442599177360535,
153
+ 0.6720470190048218,
154
+ 3.1415927410125732,
155
+ 1.5707963705062866,
156
+ 3.1415927410125732,
157
+ 0.0810599997639656,
158
+ 0.5695070028305054,
159
+ 0.4637550115585327,
160
+ 0.6456500291824341,
161
+ 3.1415927410125732,
162
+ 1.5707963705062866,
163
+ 3.1415927410125732,
164
+ 0.0611100010573864
165
+ ],
166
+ "min": [
167
+ -0.10788500308990479,
168
+ -0.5191159844398499,
169
+ 0.040369000285863876,
170
+ -3.1410515308380127,
171
+ -1.0231120586395264,
172
+ -3.1410515308380127,
173
+ -0.0017500000540167093,
174
+ -0.03119499981403351,
175
+ -0.3384290039539337,
176
+ 0.07246199995279312,
177
+ -3.1410515308380127,
178
+ -0.7706850171089172,
179
+ -3.1410515308380127,
180
+ 0.0
181
+ ],
182
+ "q01": [
183
+ 0.03489213060587645,
184
+ -0.36912800788879396,
185
+ 0.07184209674596786,
186
+ -3.137246325016022,
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
+ 0.5304324626922607,
200
+ 0.41213971734046934,
201
+ 0.44856594711542136,
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": 2587963,
216
+ "num_trajectories": 1976
217
+ }
218
+ }