Hugging Face's logo

cheikh025
/
ASR

Model card Files
xet
 Community
cheikh025 commited on
Commit
1f55cef
·
verified ·
1 Parent(s): 6bff6b7

Upload promoted/0014_continued_from_exp0013_dirresume/config.yaml with huggingface_hub

Browse files
Files changed (1) hide show
  1. promoted/0014_continued_from_exp0013_dirresume/config.yaml +200 -0
promoted/0014_continued_from_exp0013_dirresume/config.yaml ADDED
@@ -0,0 +1,200 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ output_dir: ./runs/reweighted_libero90_finetune/exp0014_continued_dirresume
2
+ batch_size: 4
3
+ num_workers: 8
4
+ lr_scheduler_type: cosine
5
+ learning_rate: 3.0e-05
6
+ num_epochs: 1
7
+ max_steps: 14000
8
+ log_every: 10
9
+ save_every: 500
10
+ eval_every: 200
11
+ eval_num_inference_steps: 10
12
+ gradient_accumulation_steps: 4
13
+ mixed_precision: bf16
14
+ seed: 42
15
+ max_grad_norm: 1.0
16
+ weight_decay: 0.01
17
+ resume: ./runs/reweighted_libero90_finetune/exp0013_continued_fixed/checkpoints/state/step_008000
18
+ wandb:
19
+ enabled: false
20
+ workspace: null
21
+ project: fast-wam
22
+ name: exp0014_continued_from_exp0013_dirresume
23
+ group: null
24
+ mode: online
25
+ data:
26
+ train:
27
+ _target_: fastwam.datasets.lerobot.robot_video_dataset.RobotVideoDataset
28
+ dataset_dirs:
29
+ - ./data/libero_mujoco3.3.2/libero_spatial_no_noops_lerobot
30
+ - ./data/libero_mujoco3.3.2/libero_object_no_noops_lerobot
31
+ - ./data/libero_mujoco3.3.2/libero_goal_no_noops_lerobot
32
+ - ./data/libero_mujoco3.3.2/libero_goal_no_noops_lerobot
33
+ - ./data/libero_mujoco3.3.2/libero_goal_no_noops_lerobot
34
+ - ./data/libero_mujoco3.3.2/libero_goal_no_noops_lerobot
35
+ - ./data/libero_mujoco3.3.2/libero_goal_no_noops_lerobot
36
+ - ./data/libero_mujoco3.3.2/libero_10_no_noops_lerobot
37
+ - ./data/libero_mujoco3.3.2/libero_10_no_noops_lerobot
38
+ - ./data/libero_mujoco3.3.2/libero_10_no_noops_lerobot
39
+ - ./data/libero_mujoco3.3.2/libero_10_no_noops_lerobot
40
+ - ./data/libero_mujoco3.3.2/libero_10_no_noops_lerobot
41
+ - ./data/libero_mujoco3.3.2/libero_90_no_noops_lerobot
42
+ shape_meta:
43
+ images:
44
+ - key: image
45
+ raw_shape:
46
+ - 3
47
+ - 512
48
+ - 512
49
+ shape:
50
+ - 3
51
+ - 224
52
+ - 224
53
+ - key: wrist_image
54
+ raw_shape:
55
+ - 3
56
+ - 512
57
+ - 512
58
+ shape:
59
+ - 3
60
+ - 224
61
+ - 224
62
+ action:
63
+ - key: default
64
+ raw_shape: 7
65
+ shape: 7
66
+ state:
67
+ - key: default
68
+ raw_shape: 8
69
+ shape: 8
70
+ num_frames: 33
71
+ global_sample_stride: 1
72
+ action_video_freq_ratio: 4
73
+ video_size:
74
+ - 224
75
+ - 448
76
+ camera_key: null
77
+ val_set_proportion: 0.0
78
+ is_training_set: true
79
+ skip_padding_as_possible: false
80
+ concat_multi_camera: horizontal
81
+ processor:
82
+ _target_: fastwam.datasets.lerobot.processors.fastwam_processor.FastWAMProcessor
83
+ shape_meta:
84
+ images:
85
+ - key: image
86
+ raw_shape:
87
+ - 3
88
+ - 512
89
+ - 512
90
+ shape:
91
+ - 3
92
+ - 224
93
+ - 224
94
+ - key: wrist_image
95
+ raw_shape:
96
+ - 3
97
+ - 512
98
+ - 512
99
+ shape:
100
+ - 3
101
+ - 224
102
+ - 224
103
+ action:
104
+ - key: default
105
+ raw_shape: 7
106
+ shape: 7
107
+ state:
108
+ - key: default
109
+ raw_shape: 8
110
+ shape: 8
111
+ num_obs_steps: 33
112
+ num_output_cameras: 2
113
+ action_output_dim: 7
114
+ proprio_output_dim: 8
115
+ delta_action_dim_mask:
116
+ default:
117
+ - true
118
+ - true
119
+ - true
120
+ - true
121
+ - true
122
+ - true
123
+ - false
124
+ action_state_transforms: null
125
+ use_stepwise_action_norm: false
126
+ norm_default_mode: min/max
127
+ norm_exception_mode: null
128
+ action_state_merger:
129
+ _target_: fastwam.datasets.lerobot.transforms.action_state_merger.ConcatLeftAlign
130
+ train_transforms:
131
+ - _target_: fastwam.datasets.lerobot.transforms.image.ToTensor
132
+ - _target_: torchvision.transforms.Resize
133
+ size:
134
+ - 224
135
+ - 224
136
+ val_transforms:
137
+ - _target_: fastwam.datasets.lerobot.transforms.image.ToTensor
138
+ - _target_: torchvision.transforms.Resize
139
+ size:
140
+ - 224
141
+ - 224
142
+ text_embedding_cache_dir: ./data/text_embeds_cache/libero
143
+ context_len: 128
144
+ model:
145
+ _target_: fastwam.runtime.create_fastwam
146
+ model_id: Wan-AI/Wan2.2-TI2V-5B
147
+ tokenizer_model_id: Wan-AI/Wan2.1-T2V-1.3B
148
+ tokenizer_max_len: 128
149
+ load_text_encoder: false
150
+ proprio_dim: 8
151
+ redirect_common_files: true
152
+ mot_checkpoint_mixed_attn: false
153
+ action_dit_pretrained_path: checkpoints/ActionDiT_linear_interp_Wan22_alphascale_1024hdim.pt
154
+ skip_dit_load_from_pretrain: false
155
+ video_dit_config:
156
+ has_image_input: false
157
+ patch_size:
158
+ - 1
159
+ - 2
160
+ - 2
161
+ in_dim: 48
162
+ hidden_dim: 3072
163
+ ffn_dim: 14336
164
+ freq_dim: 256
165
+ text_dim: 4096
166
+ out_dim: 48
167
+ num_heads: 24
168
+ attn_head_dim: 128
169
+ num_layers: 30
170
+ eps: 1.0e-06
171
+ seperated_timestep: true
172
+ require_clip_embedding: false
173
+ require_vae_embedding: false
174
+ fuse_vae_embedding_in_latents: true
175
+ use_gradient_checkpointing: false
176
+ video_attention_mask_mode: first_frame_causal
177
+ action_conditioned: false
178
+ action_dim: 7
179
+ action_group_causal_mask_mode: group_diagonal
180
+ action_dit_config:
181
+ action_dim: 7
182
+ hidden_dim: 1024
183
+ ffn_dim: 4096
184
+ num_heads: 24
185
+ attn_head_dim: 128
186
+ num_layers: 30
187
+ text_dim: 4096
188
+ freq_dim: 256
189
+ eps: 1.0e-06
190
+ use_gradient_checkpointing: false
191
+ video_scheduler:
192
+ train_shift: 5.0
193
+ infer_shift: 5.0
194
+ num_train_timesteps: 1000
195
+ action_scheduler:
196
+ train_shift: 5.0
197
+ infer_shift: 5.0
198
+ num_train_timesteps: 1000
199
+ loss:
200
+ lambda_action: 1.0