Hugging Face's logo

uuuhjb
/
2nodebs8_decouple

Model card Files
xet
 Community
2nodebs8_decouple
File size: 7,340 Bytes
acf49f8
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
output_dir: ./runs/train_libero_v2_512hdim_variant_decouple_2node/decouple_2node_b8_20260605_184739
batch_size: 8
num_workers: 4
prefetch_factor: 6
lr_scheduler_type: constant
learning_rate: 0.0001
num_epochs: 10
max_steps: 80000
log_every: 10
save_every: 5000
state_keep_last_n: 1
weights_keep_last_n: 1000
long_term_save_every: 10000
long_term_save_start: 0
eval_every: 0
eval_num_inference_steps: 10
gradient_accumulation_steps: 1
mixed_precision: bf16
seed: 42
max_grad_norm: 1.0
weight_decay: 0.01
resume: null
compile_mot: false
optimizer_type: adamw8bit
wandb:
  enabled: true
  workspace: null
  project: fastwam_ltx_decouple
  name: decouple_2node_b8_20260605_184739
  group: null
  mode: online
data:
  train:
    _target_: fastwam.datasets.lerobot.robot_video_dataset.RobotVideoDataset
    dataset_dirs:
    - ./data/LIBERO-fastwam/libero_spatial_no_noops_lerobot
    - ./data/LIBERO-fastwam/libero_object_no_noops_lerobot
    - ./data/LIBERO-fastwam/libero_goal_no_noops_lerobot
    - ./data/LIBERO-fastwam/libero_10_no_noops_lerobot
    shape_meta:
      images:
      - key: image
        raw_shape:
        - 3
        - 512
        - 512
        shape:
        - 3
        - 224
        - 224
      - key: wrist_image
        raw_shape:
        - 3
        - 512
        - 512
        shape:
        - 3
        - 224
        - 224
      action:
      - key: default
        raw_shape: 7
        shape: 7
      state:
      - key: default
        raw_shape: 8
        shape: 8
    num_frames: 33
    global_sample_stride: 1
    action_video_freq_ratio: 4
    video_size:
    - 224
    - 448
    camera_key: null
    val_set_proportion: 0
    is_training_set: true
    skip_padding_as_possible: false
    concat_multi_camera: horizontal
    processor:
      _target_: fastwam.datasets.lerobot.processors.fastwam_processor.FastWAMProcessor
      shape_meta:
        images:
        - key: image
          raw_shape:
          - 3
          - 512
          - 512
          shape:
          - 3
          - 224
          - 224
        - key: wrist_image
          raw_shape:
          - 3
          - 512
          - 512
          shape:
          - 3
          - 224
          - 224
        action:
        - key: default
          raw_shape: 7
          shape: 7
        state:
        - key: default
          raw_shape: 8
          shape: 8
      num_obs_steps: 33
      num_output_cameras: 2
      action_output_dim: 7
      proprio_output_dim: 8
      delta_action_dim_mask:
        default:
        - true
        - true
        - true
        - true
        - true
        - true
        - false
      action_state_transforms: null
      use_stepwise_action_norm: false
      norm_default_mode: min/max
      norm_exception_mode: null
      action_state_merger:
        _target_: fastwam.datasets.lerobot.transforms.action_state_merger.ConcatLeftAlign
      train_transforms:
      - _target_: fastwam.datasets.lerobot.transforms.image.ToTensor
      - _target_: torchvision.transforms.Resize
        size:
        - 224
        - 224
      val_transforms:
      - _target_: fastwam.datasets.lerobot.transforms.image.ToTensor
      - _target_: torchvision.transforms.Resize
        size:
        - 224
        - 224
    text_embedding_cache_dir: ./data/text_embeds_cache/libero
    text_cache_slug: ltx23_gemma3_12b_v2connector
    context_len: 128
    joint_latent_cache_dir: ./data/joint_latents/libero_ratio4_nf33
  val:
    _target_: fastwam.datasets.lerobot.robot_video_dataset.RobotVideoDataset
    dataset_dirs:
    - ./data/LIBERO-fastwam/libero_spatial_no_noops_lerobot
    - ./data/LIBERO-fastwam/libero_object_no_noops_lerobot
    - ./data/LIBERO-fastwam/libero_goal_no_noops_lerobot
    - ./data/LIBERO-fastwam/libero_10_no_noops_lerobot
    shape_meta:
      images:
      - key: image
        raw_shape:
        - 3
        - 512
        - 512
        shape:
        - 3
        - 224
        - 224
      - key: wrist_image
        raw_shape:
        - 3
        - 512
        - 512
        shape:
        - 3
        - 224
        - 224
      action:
      - key: default
        raw_shape: 7
        shape: 7
      state:
      - key: default
        raw_shape: 8
        shape: 8
    num_frames: 33
    global_sample_stride: 1
    action_video_freq_ratio: 4
    video_size:
    - 224
    - 448
    camera_key: null
    val_set_proportion: 0
    is_training_set: false
    skip_padding_as_possible: false
    concat_multi_camera: horizontal
    processor:
      _target_: fastwam.datasets.lerobot.processors.fastwam_processor.FastWAMProcessor
      shape_meta:
        images:
        - key: image
          raw_shape:
          - 3
          - 512
          - 512
          shape:
          - 3
          - 224
          - 224
        - key: wrist_image
          raw_shape:
          - 3
          - 512
          - 512
          shape:
          - 3
          - 224
          - 224
        action:
        - key: default
          raw_shape: 7
          shape: 7
        state:
        - key: default
          raw_shape: 8
          shape: 8
      num_obs_steps: 33
      num_output_cameras: 2
      action_output_dim: 7
      proprio_output_dim: 8
      delta_action_dim_mask:
        default:
        - true
        - true
        - true
        - true
        - true
        - true
        - false
      action_state_transforms: null
      use_stepwise_action_norm: false
      norm_default_mode: min/max
      norm_exception_mode: null
      action_state_merger:
        _target_: fastwam.datasets.lerobot.transforms.action_state_merger.ConcatLeftAlign
      train_transforms:
      - _target_: fastwam.datasets.lerobot.transforms.image.ToTensor
      - _target_: torchvision.transforms.Resize
        size:
        - 224
        - 224
      val_transforms:
      - _target_: fastwam.datasets.lerobot.transforms.image.ToTensor
      - _target_: torchvision.transforms.Resize
        size:
        - 224
        - 224
    text_embedding_cache_dir: ./data/text_embeds_cache/libero
    text_cache_slug: ltx23_gemma3_12b_v2connector
    context_len: 128
model:
  _target_: fastwam.runtime.create_fastwam
  ckpt_path: checkpoints/Lightricks/LTX-2.3/ltx-2.3-22b-dev.safetensors
  gemma_path: checkpoints/google/gemma-3-12b-it-qat-q4_0-unquantized
  load_text_encoder: false
  attach_gemma_to_text_encoder: false
  proprio_dim: 8
  mot_checkpoint_mixed_attn: false
  action_dit_pretrained_path: checkpoints/preprocessed/ltx_action_dit_backbone.pt
  skip_dit_load_from_pretrain: false
  video_dit_config:
    text_dim: 4096
    use_gradient_checkpointing: false
    action_dim: 7
  action_dit_config:
    action_dim: 7
    hidden_dim: 512
    num_heads: 32
    attn_head_dim: 128
    num_layers: 48
    text_dim: 4096
    eps: 1.0e-06
    cross_attention_adaln: false
    use_gradient_checkpointing: false
  video_scheduler:
    type: ltx2
    min_shift: 0.95
    max_shift: 2.05
    min_tokens: 1024
    max_tokens: 4096
    infer_shift: 2.05
    num_train_timesteps: 1000
    train_shift: 5.0
    sigma_floor: 0.0
  action_scheduler:
    type: wan
    train_shift: 5.0
    infer_shift: 5.0
    num_train_timesteps: 1000
    sigma_floor: 0.0
  loss:
    lambda_video: 0.1
    lambda_action: 1.0
  action_only_train: false
  mot_attn_decouple_frac: 0.25
video_expert_lr: 1.0e-05
action_expert_lr: 0.0001