Hugging Face's logo

VLyb
/
FrameSkip-SimplerEnv-WidowX

Robotics
Model card Files
xet
 Community
1
VLyb commited on
Commit
dd23092
·
verified ·
1 Parent(s): 5abb6e4

Upload folder using huggingface_hub

Browse files
Files changed (5) hide show
  1. checkpoints/steps_50000_pytorch_model.pt +3 -0
  2. checkpoints/steps_60000_pytorch_model.pt +3 -0
  3. checkpoints/steps_70000_pytorch_model.pt +3 -0
  4. config.yaml +107 -0
  5. dataset_statistics.json +133 -0
checkpoints/steps_50000_pytorch_model.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e2f8b9838d211d7645eca9aa47d83b7420fe26fca65142f64b3634fa8401f67c
3
+ size 9982353306
checkpoints/steps_60000_pytorch_model.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d24dae91b10ee9214b913029c0c42f5625dd5be8119416dc5d1d6ad5d4513296
3
+ size 9982353306
checkpoints/steps_70000_pytorch_model.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4c9f3630b6ee39cc14ee8a9266f6e9900faa42381e99f9da4318bbac73fd6df9
3
+ size 9982353306
config.yaml ADDED
@@ -0,0 +1,107 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ datasets:
2
+ vla_data:
3
+ data_mix: bridge
4
+ data_root_dir: /mnt/project_ai4edu/lyl/worlkplace/starVLA/playground/Datasets/OXE_LEROBOT_DATASET
5
+ image_size:
6
+ - 224
7
+ - 224
8
+ per_device_batch_size: 16
9
+ video_backend: pyav
10
+ frameskip:
11
+ cache_dir: playground/frameskip/frameskip_cache_vac_cr10-100
12
+ default_compression_ratio: 1.0
13
+ enabled: true
14
+ importance:
15
+ allow_backend_fallback: false
16
+ alpha: 0.6
17
+ beta: 0.2
18
+ device: cuda
19
+ enable_vac: true
20
+ gamma: 0.2
21
+ max_vac_frames: 16
22
+ type: gripper_aware
23
+ vac_beta: 0.2
24
+ video_backend: ffmpeg
25
+ visual_encoder_checkpoint: /mnt/project_ai4edu/share/models/timm/vit_large_patch14_dinov2.lvd142m/pytorch_model.bin
26
+ visual_encoder_name: vit_large_patch14_dinov2
27
+ pruning:
28
+ compression_ratios:
29
+ - 0.1
30
+ - 0.2
31
+ - 0.3
32
+ - 0.4
33
+ - 0.5
34
+ - 0.6
35
+ - 0.7
36
+ - 0.8
37
+ - 0.9
38
+ - 1.0
39
+ used_compression_ratios:
40
+ - 0.2
41
+ - 0.2
42
+ - 0.2
43
+ - 0.2
44
+ - 0.2
45
+ - 1.0
46
+ training:
47
+ dynamic_ratio: true
48
+ ratio_schedule: uniform
49
+ warmup_steps: 5000
50
+ framework:
51
+ action_model:
52
+ action_dim: 7
53
+ action_horizon: 16
54
+ action_model_type: DiT-B
55
+ add_pos_embed: true
56
+ diffusion_model_cfg:
57
+ cross_attention_dim: 2560
58
+ dropout: 0.2
59
+ final_dropout: true
60
+ interleave_self_attention: true
61
+ norm_type: ada_norm
62
+ num_layers: 16
63
+ output_dim: 2560
64
+ positional_embeddings: null
65
+ future_action_window_size: 15
66
+ hidden_size: 1024
67
+ max_seq_len: 1024
68
+ noise_beta_alpha: 1.5
69
+ noise_beta_beta: 1.0
70
+ noise_s: 0.999
71
+ num_inference_timesteps: 4
72
+ num_target_vision_tokens: 32
73
+ num_timestep_buckets: 1000
74
+ past_action_window_size: 0
75
+ state_dim: 7
76
+ name: QwenGR00T
77
+ qwenvl:
78
+ base_vlm: FrameSkip_GR00T_bridge_qwen3/Qwen3-VL-4B-Instruct
79
+ template: qwen3_vl
80
+ run_root_dir: ./results/Checkpoints/FrameSkip/SimplerEnv
81
+ seed: 42
82
+ trainer:
83
+ eval_interval: 1000
84
+ freeze_modules: true
85
+ gradient_accumulation_steps: 1
86
+ gradient_clipping: 1.0
87
+ is_resume: true
88
+ learning_rate:
89
+ action_model: 0.0001
90
+ base: 1.0e-05
91
+ qwen_vl_interface: 1.0e-05
92
+ logging_frequency: 100
93
+ lr_scheduler_type: cosine_with_min_lr
94
+ max_train_steps: 100000
95
+ num_warmup_steps: 2000
96
+ optimizer:
97
+ betas:
98
+ - 0.9
99
+ - 0.95
100
+ eps: 1.0e-08
101
+ weight_decay: 1.0e-08
102
+ repeated_diffusion_steps: 4
103
+ save_interval: 10000
104
+ scheduler_specific_kwargs:
105
+ min_lr: 5.0e-07
106
+ wandb_entity: jinhuiye
107
+ wandb_project: starVLA
dataset_statistics.json ADDED
@@ -0,0 +1,133 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "oxe_bridge": {
3
+ "action": {
4
+ "mean": [
5
+ 0.0002273258287459612,
6
+ 0.00013113691238686442,
7
+ -0.00012639156193472445,
8
+ -0.00014385415124706924,
9
+ -0.00039017299422994256,
10
+ 0.00024080397270154208,
11
+ 0.5765793323516846
12
+ ],
13
+ "std": [
14
+ 0.009770580567419529,
15
+ 0.013695014640688896,
16
+ 0.012675177305936813,
17
+ 0.02845500037074089,
18
+ 0.030521046370267868,
19
+ 0.07739090174436569,
20
+ 0.4966537058353424
21
+ ],
22
+ "max": [
23
+ 0.41691166162490845,
24
+ 0.25864794850349426,
25
+ 0.21218234300613403,
26
+ 3.122201919555664,
27
+ 1.8618112802505493,
28
+ 6.272472858428955,
29
+ 1.0
30
+ ],
31
+ "min": [
32
+ -0.4007510244846344,
33
+ -0.13874775171279907,
34
+ -0.22553899884223938,
35
+ -3.2010786533355713,
36
+ -1.8618112802505493,
37
+ -6.279075622558594,
38
+ 0.0
39
+ ],
40
+ "q01": [
41
+ -0.028752606511116028,
42
+ -0.041702210046350954,
43
+ -0.026096698231995105,
44
+ -0.08052822157740593,
45
+ -0.09249736212193965,
46
+ -0.20738411962985992,
47
+ 0.0
48
+ ],
49
+ "q99": [
50
+ 0.02830690816044803,
51
+ 0.04089860741049051,
52
+ 0.04018005654215808,
53
+ 0.08173405691981314,
54
+ 0.07760896608233431,
55
+ 0.20384809583425495,
56
+ 1.0
57
+ ],
58
+ "mask": [
59
+ true,
60
+ true,
61
+ true,
62
+ true,
63
+ true,
64
+ true,
65
+ false
66
+ ]
67
+ },
68
+ "state": {
69
+ "mean": [
70
+ 0.3094329535961151,
71
+ 0.030725397169589996,
72
+ 0.06444206088781357,
73
+ 0.006489328108727932,
74
+ -0.07720080018043518,
75
+ 0.10767334699630737,
76
+ 0.0,
77
+ 0.7081261277198792
78
+ ],
79
+ "std": [
80
+ 0.06056813895702362,
81
+ 0.09190769493579865,
82
+ 0.05156221613287926,
83
+ 0.13109272718429565,
84
+ 0.16913393139839172,
85
+ 0.5779843926429749,
86
+ 0.0,
87
+ 0.35254883766174316
88
+ ],
89
+ "max": [
90
+ 0.5862360596656799,
91
+ 0.4034728705883026,
92
+ 0.3568263053894043,
93
+ 1.3517684936523438,
94
+ 1.570796251296997,
95
+ 3.141204357147217,
96
+ 0.0,
97
+ 1.1121242046356201
98
+ ],
99
+ "min": [
100
+ -0.04167502000927925,
101
+ -0.3563207685947418,
102
+ -0.15537554025650024,
103
+ -3.141592502593994,
104
+ -1.4992541074752808,
105
+ -3.14153790473938,
106
+ 0.0,
107
+ 0.04637829214334488
108
+ ],
109
+ "q01": [
110
+ 0.17102580681443214,
111
+ -0.16981234937906264,
112
+ -0.05563282176852226,
113
+ -0.36493386059999466,
114
+ -0.5418747025728226,
115
+ -1.3542919230461121,
116
+ 0.0,
117
+ 0.052190229296684265
118
+ ],
119
+ "q99": [
120
+ 0.45322125554084775,
121
+ 0.2354859386384485,
122
+ 0.19489662453532214,
123
+ 0.3779941478371616,
124
+ 0.2756884342432019,
125
+ 1.8500668883323654,
126
+ 0.0,
127
+ 1.0105689764022827
128
+ ]
129
+ },
130
+ "num_transitions": 1305714,
131
+ "num_trajectories": 53192
132
+ }
133
+ }