Hugging Face's logo

Chengpengaaa
/
LFM

Model card Files
xet
 Community
Chengpengaaa commited on
Commit
f54316d
·
verified ·
1 Parent(s): 0ad8ac7

Upload folder using huggingface_hub

Browse files
Files changed (4) hide show
  1. LFMPI-0206-N1-v2/checkpoints/steps_31000_pytorch_model.pt +3 -0
  2. LFMPI-0206-N1-v2/config.json +103 -0
  3. LFMPI-0206-N1-v2/config.yaml +60 -0
  4. LFMPI-0206-N1-v2/dataset_statistics.json +218 -0
LFMPI-0206-N1-v2/checkpoints/steps_31000_pytorch_model.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3c498d3cb4f177e4ee673c42c52df92144776ec359c8940e1c6586f3fb3f2dd2
3
+ size 5559321071
LFMPI-0206-N1-v2/config.json ADDED
@@ -0,0 +1,103 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "run_id": "LFMPI-0206-N1-v2",
3
+ "run_root_dir": "/lumos-vePFS/shenzhen/results",
4
+ "seed": 42,
5
+ "trackers": [
6
+ "jsonl",
7
+ "wandb"
8
+ ],
9
+ "wandb_entity": "wshatake669",
10
+ "wandb_project": "lfm-pi-experiment",
11
+ "is_debug": false,
12
+ "framework": {
13
+ "name": "LFMPI",
14
+ "lfm_vl": {
15
+ "base_vlm": "/lumos-vePFS/shenzhen/models/LFM2.5-VL-1.6B",
16
+ "attn_implementation": "sdpa"
17
+ },
18
+ "action_model": {
19
+ "action_model_type": "DiT-B",
20
+ "add_pos_embed": true,
21
+ "hidden_size": 2048,
22
+ "max_seq_len": 1024,
23
+ "action_dim": 20,
24
+ "state_dim": 20,
25
+ "future_action_window_size": 30,
26
+ "repeated_diffusion_steps": 2,
27
+ "past_action_window_size": 0,
28
+ "noise_beta_alpha": 1.5,
29
+ "noise_beta_beta": 1.0,
30
+ "noise_s": 0.999,
31
+ "num_timestep_buckets": 1000,
32
+ "num_inference_timesteps": 4,
33
+ "num_target_vision_tokens": 32,
34
+ "use_scaled_noise": true,
35
+ "diffusion_model_cfg": {
36
+ "num_layers": 17,
37
+ "dropout": 0.2,
38
+ "interleave_self_attention": true,
39
+ "norm_type": "ada_norm"
40
+ }
41
+ }
42
+ },
43
+ "datasets": {
44
+ "vla_data": {
45
+ "dataset_py": "lerobot_datasets",
46
+ "data_root_dir": "/lumos-vePFS/shenzhen/data/data_warehouse/data_warehouse_output",
47
+ "data_mix": "train_lfm_pi_0205",
48
+ "per_device_batch_size": 48,
49
+ "num_workers": 2,
50
+ "action_type": "abs_ee",
51
+ "image_size": [
52
+ 224,
53
+ 224
54
+ ],
55
+ "video_backend": "pyav"
56
+ }
57
+ },
58
+ "trainer": {
59
+ "epochs": 100,
60
+ "max_train_steps": 80000,
61
+ "num_warmup_steps": 5000,
62
+ "save_interval": 1000,
63
+ "eval_interval": 1000000,
64
+ "repeated_diffusion_steps": 2,
65
+ "learning_rate": {
66
+ "base": 1e-05,
67
+ "intern_vl_interface": 1e-05,
68
+ "action_model": 0.0001
69
+ },
70
+ "lr_scheduler_type": "cosine_with_min_lr",
71
+ "scheduler_specific_kwargs": {
72
+ "min_lr": 1e-06
73
+ },
74
+ "freeze_modules": "lfm_vl_interface.model.model.language_model",
75
+ "loss_scale": {
76
+ "vla": 1.0,
77
+ "vlm": 0.05
78
+ },
79
+ "max_grad_norm": 1.0,
80
+ "warmup_ratio": 0.1,
81
+ "weight_decay": 0.0,
82
+ "logging_frequency": 10,
83
+ "gradient_clipping": 1.0,
84
+ "gradient_accumulation_steps": 1,
85
+ "optimizer": {
86
+ "name": "AdamW",
87
+ "betas": [
88
+ 0.9,
89
+ 0.95
90
+ ],
91
+ "eps": 1e-08,
92
+ "weight_decay": 0.0
93
+ }
94
+ },
95
+ "is_resume": false,
96
+ "resume_epoch": null,
97
+ "resume_step": null,
98
+ "enable_gradient_checkpointing": true,
99
+ "enable_mixed_precision_training": true,
100
+ "report_to": "wandb",
101
+ "wandb_run_name": "v1",
102
+ "output_dir": "/lumos-vePFS/shenzhen/results/LFMPI-0206-N1-v2"
103
+ }
LFMPI-0206-N1-v2/config.yaml ADDED
@@ -0,0 +1,60 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ datasets:
2
+ vla_data:
3
+ data_mix: train_lfm_pi_0205
4
+ data_root_dir: /lumos-vePFS/shenzhen/data/data_warehouse/data_warehouse_output
5
+ dataset_py: lerobot_datasets
6
+ num_workers: 2
7
+ per_device_batch_size: 48
8
+ framework:
9
+ action_model:
10
+ action_dim: 20
11
+ add_pos_embed: true
12
+ diffusion_model_cfg:
13
+ dropout: 0.2
14
+ interleave_self_attention: true
15
+ norm_type: ada_norm
16
+ num_layers: 17
17
+ future_action_window_size: 30
18
+ max_seq_len: 1024
19
+ noise_beta_alpha: 1.5
20
+ noise_beta_beta: 1.0
21
+ noise_s: 0.999
22
+ num_inference_timesteps: 4
23
+ num_target_vision_tokens: 32
24
+ num_timestep_buckets: 1000
25
+ past_action_window_size: 0
26
+ state_dim: 20
27
+ use_scaled_noise: true
28
+ lfm_vl:
29
+ base_vlm: /lumos-vePFS/shenzhen/models/LFM2.5-VL-1.6B
30
+ name: LFMPI
31
+ qwenvl:
32
+ num_vl_layers: 16
33
+ vl_hidden_dim: 2048
34
+ output_dir: /lumos-vePFS/shenzhen/results/LFMPI-0206-N1-v2
35
+ run_id: LFMPI-0206-N1-v2
36
+ run_root_dir: /lumos-vePFS/shenzhen/results
37
+ seed: 42
38
+ trainer:
39
+ eval_interval: 1000000
40
+ freeze_modules: lfm_vl_interface.model.model.language_model
41
+ gradient_accumulation_steps: 1
42
+ gradient_clipping: 1.0
43
+ learning_rate:
44
+ action_model: 0.0001
45
+ base: 1.0e-05
46
+ intern_vl_interface: 1.0e-05
47
+ logging_frequency: 10
48
+ lr_scheduler_type: cosine_with_min_lr
49
+ max_train_steps: 80000
50
+ num_warmup_steps: 5000
51
+ optimizer:
52
+ betas:
53
+ - 0.9
54
+ - 0.95
55
+ eps: 1.0e-08
56
+ weight_decay: 0.0
57
+ repeated_diffusion_steps: 2
58
+ save_interval: 1000
59
+ scheduler_specific_kwargs:
60
+ min_lr: 1.0e-06
LFMPI-0206-N1-v2/dataset_statistics.json ADDED
@@ -0,0 +1,218 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "new_embodiment": {
3
+ "action": {
4
+ "mean": [
5
+ 0.2536556995473802,
6
+ -0.07355320891838346,
7
+ 0.19678196962922812,
8
+ -1.2941462621092796,
9
+ 0.400992049369961,
10
+ 1.3994900174438953,
11
+ 0.034343936713412404,
12
+ 0.2324184780009091,
13
+ 0.08854325220454484,
14
+ 0.20849504927173257,
15
+ 1.4624766141641885,
16
+ 0.5071446867659688,
17
+ -1.7145425528287888,
18
+ 0.03544515132671222
19
+ ],
20
+ "std": [
21
+ 0.09994235415543593,
22
+ 0.11036921545415922,
23
+ 0.06864694427641027,
24
+ 2.6015310152230064,
25
+ 0.33020659125118274,
26
+ 2.4404885524065727,
27
+ 0.028107770048515182,
28
+ 0.09007369079509984,
29
+ 0.07930141357553988,
30
+ 0.0562799268900712,
31
+ 2.5360845693916216,
32
+ 0.29748079788203885,
33
+ 2.224083566284048,
34
+ 0.02813262430858898
35
+ ],
36
+ "max": [
37
+ 0.5931400060653687,
38
+ 0.48442599177360535,
39
+ 0.6967939734458923,
40
+ 3.1415927410125732,
41
+ 1.5707963705062866,
42
+ 3.1415927410125732,
43
+ 0.0810599997639656,
44
+ 0.5580319762229919,
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
+ -1.5302174091339111,
58
+ -3.1410515308380127,
59
+ 0.0,
60
+ -0.02319999970495701,
61
+ -0.3384290039539337,
62
+ 0.06903599947690964,
63
+ -3.1410515308380127,
64
+ -0.7706850171089172,
65
+ -3.1410515308380127,
66
+ 0.0
67
+ ],
68
+ "q01": [
69
+ 0.04410205833613873,
70
+ -0.36912800788879396,
71
+ 0.07184209674596786,
72
+ -3.137246325016022,
73
+ -0.20070483744144438,
74
+ -3.137392144203186,
75
+ 7.000000186963007e-05,
76
+ 0.05372548036277294,
77
+ -0.15914200246334076,
78
+ 0.10170399993658066,
79
+ -3.138237874507904,
80
+ -0.11282174646854401,
81
+ -3.1348206996917725,
82
+ 7.000000186963007e-05
83
+ ],
84
+ "q99": [
85
+ 0.5304324626922607,
86
+ 0.41213971734046934,
87
+ 0.47792605459690085,
88
+ 3.1415927410125732,
89
+ 1.5365502309799193,
90
+ 3.1415927410125732,
91
+ 0.06013000011444092,
92
+ 0.4840700030326843,
93
+ 0.34843780994415285,
94
+ 0.42531779676675785,
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
+ 0.2536556995473802,
120
+ -0.07355320891838346,
121
+ 0.19678196962922812,
122
+ -1.2941462621092796,
123
+ 0.400992049369961,
124
+ 1.3994900174438953,
125
+ 0.034343936713412404,
126
+ 0.2324184780009091,
127
+ 0.08854325220454484,
128
+ 0.20849504927173257,
129
+ 1.4624766141641885,
130
+ 0.5071446867659688,
131
+ -1.7145425528287888,
132
+ 0.03544515132671222
133
+ ],
134
+ "std": [
135
+ 0.09994235415543593,
136
+ 0.11036921545415922,
137
+ 0.06864694427641027,
138
+ 2.6015310152230064,
139
+ 0.33020659125118274,
140
+ 2.4404885524065727,
141
+ 0.028107770048515182,
142
+ 0.09007369079509984,
143
+ 0.07930141357553988,
144
+ 0.0562799268900712,
145
+ 2.5360845693916216,
146
+ 0.29748079788203885,
147
+ 2.224083566284048,
148
+ 0.02813262430858898
149
+ ],
150
+ "max": [
151
+ 0.5931400060653687,
152
+ 0.48442599177360535,
153
+ 0.6967939734458923,
154
+ 3.1415927410125732,
155
+ 1.5707963705062866,
156
+ 3.1415927410125732,
157
+ 0.0810599997639656,
158
+ 0.5580319762229919,
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
+ -1.5302174091339111,
172
+ -3.1410515308380127,
173
+ 0.0,
174
+ -0.02319999970495701,
175
+ -0.3384290039539337,
176
+ 0.06903599947690964,
177
+ -3.1410515308380127,
178
+ -0.7706850171089172,
179
+ -3.1410515308380127,
180
+ 0.0
181
+ ],
182
+ "q01": [
183
+ 0.04410205833613873,
184
+ -0.36912800788879396,
185
+ 0.07184209674596786,
186
+ -3.137246325016022,
187
+ -0.20070483744144438,
188
+ -3.137392144203186,
189
+ 7.000000186963007e-05,
190
+ 0.05372548036277294,
191
+ -0.15914200246334076,
192
+ 0.10170399993658066,
193
+ -3.138237874507904,
194
+ -0.11282174646854401,
195
+ -3.1348206996917725,
196
+ 7.000000186963007e-05
197
+ ],
198
+ "q99": [
199
+ 0.5304324626922607,
200
+ 0.41213971734046934,
201
+ 0.47792605459690085,
202
+ 3.1415927410125732,
203
+ 1.5365502309799193,
204
+ 3.1415927410125732,
205
+ 0.06013000011444092,
206
+ 0.4840700030326843,
207
+ 0.34843780994415285,
208
+ 0.42531779676675785,
209
+ 3.1415927410125732,
210
+ 1.5435859155654907,
211
+ 3.1415927410125732,
212
+ 0.05992000177502632
213
+ ]
214
+ },
215
+ "num_transitions": 2411168,
216
+ "num_trajectories": 1866
217
+ }
218
+ }