Hugging Face's logo

mondo-robotics
/
dit4dit-model

Model card Files
xet
 Community
TeliMa commited on
Commit
46237ae
·
verified ·
1 Parent(s): c36987a

Upload 3 files

Browse files
Files changed (3) hide show
  1. dit4dit_libero/config.yaml +87 -0
  2. dit4dit_libero/dataset_statistics.json +133 -0
  3. dit4dit_libero/final_model/pytorch_model.pt +3 -0
dit4dit_libero/config.yaml ADDED
@@ -0,0 +1,87 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ datasets:
2
+ vla_data:
3
+ action_video_freq_ratio: 2
4
+ data_mix: libero_all
5
+ data_root_dir: /path/to/libero
6
+ dataset_py: lerobot_datasets
7
+ delete_pause_frame: false
8
+ per_device_batch_size: 4
9
+ video_backend: torchvision_av
10
+ video_delta_indices:
11
+ - 0
12
+ - 1
13
+ - 2
14
+ - 3
15
+ - 4
16
+ - 5
17
+ - 6
18
+ - 7
19
+ - 8
20
+ framework:
21
+ action_model:
22
+ action_dim: 8
23
+ action_horizon: 8
24
+ action_model_type: DiT-B
25
+ add_pos_embed: true
26
+ diffusion_model_cfg:
27
+ cross_attention_dim: 2048
28
+ dropout: 0.2
29
+ final_dropout: true
30
+ interleave_self_attention: true
31
+ norm_type: ada_norm
32
+ num_layers: 16
33
+ output_dim: 2560
34
+ positional_embeddings: null
35
+ future_action_window_size: 7
36
+ hidden_size: 2560
37
+ max_seq_len: 1024
38
+ noise_beta_alpha: 1.5
39
+ noise_beta_beta: 1.0
40
+ noise_s: 0.999
41
+ num_inference_timesteps: 4
42
+ num_timestep_buckets: 1000
43
+ past_action_window_size: 0
44
+ state_dim: 16
45
+ cosmos25:
46
+ base_model: /path/to/Cosmos-Predict2.5-2B
47
+ conditional_frame_timestep: 0.0001
48
+ extract_layer: 17
49
+ flow_matching:
50
+ high_sigma_min: null
51
+ high_sigma_ratio: null
52
+ time_distribution: uniform
53
+ future_loss_type: flow_matching
54
+ future_num_inference_steps: 1
55
+ training: joint
56
+ vl_hidden_dim: 2048
57
+ name: DiT4DiT
58
+ output_dir: ./playground/Checkpoints_libero/dit4dit_libero
59
+ run_id: dit4dit_libero
60
+ run_root_dir: ./playground/Checkpoints_libero
61
+ seed: 42
62
+ trainer:
63
+ eval_interval: 100
64
+ freeze_modules: backbone_interface.extractor.text_encoder,backbone_interface.extractor.vae
65
+ gradient_accumulation_steps: 1
66
+ gradient_clipping: 1.0
67
+ is_resume: false
68
+ learning_rate:
69
+ action_model: 0.0001
70
+ backbone_interface: 0.0001
71
+ base: 3.0e-05
72
+ logging_frequency: 10
73
+ lr_scheduler_type: cosine_with_min_lr
74
+ max_train_steps: 160000
75
+ num_warmup_steps: 10000
76
+ optimizer:
77
+ betas:
78
+ - 0.9
79
+ - 0.95
80
+ eps: 1.0e-08
81
+ weight_decay: 1.0e-08
82
+ repeated_diffusion_steps: 4
83
+ save_interval: 80000
84
+ scheduler_specific_kwargs:
85
+ min_lr: 5.0e-07
86
+ wandb_entity: <your_wandb_entity>
87
+ wandb_project: DiT4DiT_libero
dit4dit_libero/dataset_statistics.json ADDED
@@ -0,0 +1,133 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "franka": {
3
+ "action": {
4
+ "mean": [
5
+ 0.07237596483901143,
6
+ 0.08987006871029735,
7
+ -0.10144743137061596,
8
+ -0.00045383188989944756,
9
+ 0.006273590726777911,
10
+ -0.003878799732774496,
11
+ 0.524486355483532
12
+ ],
13
+ "std": [
14
+ 0.3498823308902479,
15
+ 0.37794140366375184,
16
+ 0.460084266976933,
17
+ 0.0403885784928603,
18
+ 0.06616144248501059,
19
+ 0.07763074391911857,
20
+ 0.4994683356809767
21
+ ],
22
+ "max": [
23
+ 0.9375,
24
+ 0.9375,
25
+ 0.9375,
26
+ 0.3557142913341522,
27
+ 0.375,
28
+ 0.375,
29
+ 1.0
30
+ ],
31
+ "min": [
32
+ -0.9375,
33
+ -0.9375,
34
+ -0.9375,
35
+ -0.2582142949104309,
36
+ -0.375,
37
+ -0.3675000071525574,
38
+ 0.0
39
+ ],
40
+ "q01": [
41
+ -0.8785714507102966,
42
+ -0.8758928775787354,
43
+ -0.9375,
44
+ -0.1510714292526245,
45
+ -0.20678570866584778,
46
+ -0.2742857038974762,
47
+ 0.0
48
+ ],
49
+ "q99": [
50
+ 0.9375,
51
+ 0.9107142686843872,
52
+ 0.9375,
53
+ 0.20357142388820648,
54
+ 0.26357144117355347,
55
+ 0.375,
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
+ -0.04889854742214084,
71
+ 0.03689368185587227,
72
+ 0.7890402488410473,
73
+ 2.9771945476531982,
74
+ -0.1417286954820156,
75
+ -0.11769362539052963,
76
+ 0.026436020154505968,
77
+ -0.02665513101965189
78
+ ],
79
+ "std": [
80
+ 0.10639013941746686,
81
+ 0.15115733130675715,
82
+ 0.38406895599530033,
83
+ 0.3530238395244304,
84
+ 0.8227341427331599,
85
+ 0.32357567121520087,
86
+ 0.014583991652936385,
87
+ 0.014467005007200339
88
+ ],
89
+ "max": [
90
+ 0.21031762659549713,
91
+ 0.39128610491752625,
92
+ 1.3660105466842651,
93
+ 3.6714255809783936,
94
+ 3.560650587081909,
95
+ 1.386339545249939,
96
+ 0.04233968257904053,
97
+ 0.0013633022317662835
98
+ ],
99
+ "min": [
100
+ -0.4828203022480011,
101
+ -0.3255046010017395,
102
+ 0.008128180168569088,
103
+ 0.35277295112609863,
104
+ -3.641430377960205,
105
+ -1.842738389968872,
106
+ -0.0013586411951109767,
107
+ -0.042040832340717316
108
+ ],
109
+ "q01": [
110
+ -0.42401049643754957,
111
+ -0.2838300323486328,
112
+ 0.009925739830359817,
113
+ 1.3085840785503386,
114
+ -2.886677579879761,
115
+ -1.1599004411697387,
116
+ 0.001503719249740243,
117
+ -0.040336399003863335
118
+ ],
119
+ "q99": [
120
+ 0.1530261474847791,
121
+ 0.3629165390133857,
122
+ 1.2910678112506866,
123
+ 3.303542451858519,
124
+ 2.7496529006957933,
125
+ 0.6893712210655194,
126
+ 0.040610933862626555,
127
+ -0.0015016929572448147
128
+ ]
129
+ },
130
+ "num_transitions": 273465,
131
+ "num_trajectories": 1693
132
+ }
133
+ }
dit4dit_libero/final_model/pytorch_model.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:10cdc70029d84519942157ff741fe570a78f161be3fda6ff1e87286b4c7faa29
3
+ size 21283256413