yslan commited on
Commit
c350a43
·
verified ·
1 Parent(s): 4121404

Add immutable checkpoint backup wan22_5b_lora_video_raymap_camera_5k resolved_run_config

Browse files
stage1_4d_ablations/wan22_5b_lora_video_raymap_camera/step_005000/config.yaml ADDED
@@ -0,0 +1,540 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ output_dir: /scratch/shared/engs2907/logs/vgg_forecast_4d/train/flexpi_four_d_pretraining_wan22_5b_separate_streams_lora_video_and_raymap_camera_20261101/run
2
+ gpu_model: H200
3
+ batch_size: 8
4
+ num_workers: 8
5
+ learning_rate: 5.0e-05
6
+ num_epochs: 20
7
+ max_steps: 10000
8
+ log_every: 10
9
+ save_every: 1000
10
+ eval_every: 0
11
+ eval_num_inference_steps: 25
12
+ keep_last_n_states: 1
13
+ keep_last_n_weights: 5
14
+ eval_video: false
15
+ share_pca_basis: true
16
+ gradient_accumulation_steps: 4
17
+ mixed_precision: bf16
18
+ seed: 42
19
+ max_grad_norm: 1.0
20
+ weight_decay: 0.01
21
+ resume: null
22
+ resume_reanchor_lr_schedule: false
23
+ resume_config_source: current
24
+ pretrained_ckpt: null
25
+ pretrained_ckpt_strict_shape: false
26
+ wandb:
27
+ enabled: true
28
+ workspace: null
29
+ project: omni4d-robotwin
30
+ name: flexpi_four_d_pretraining_wan22_5b_separate_streams_lora_video_and_raymap_camera
31
+ group: flexpi_four_d_pretraining_wan22_5b_separate_streams_lora_camera_interface
32
+ mode: offline
33
+ data:
34
+ train:
35
+ _target_: four_d_windows.FourDPretrainingWindows
36
+ sources:
37
+ - name: four_d_pretraining_relaunch_syn4d_sources
38
+ kind: track4d_source_config
39
+ config: train/mot/wan_video_depth_vae_from_video_four_d_pretraining_relaunch_rendered_depth_sources_21_frames_512
40
+ weight: 15.2
41
+ left_out_sources:
42
+ - omniworld_game
43
+ - tartanair
44
+ - mvs_synth
45
+ - name: four_d_pretraining_relaunch_omniworld_game_source
46
+ kind: track4d_source_config
47
+ config: train/mot/wan_video_depth_vae_from_video_four_d_pretraining_relaunch_rendered_depth_sources_21_frames_512
48
+ weight: 8.75
49
+ left_out_sources:
50
+ - syn4d_v1
51
+ - syn4d_v2_mixed
52
+ - syn4d_v2_mixed_no_sim
53
+ - tartanair
54
+ - mvs_synth
55
+ - name: four_d_pretraining_relaunch_synthverse_source
56
+ kind: track4d_source_config
57
+ config: train/mot/wan_video_depth_vae_from_video_four_d_pretraining_relaunch_synthverse_dynamic_replica_vkitti_midair_spring_sources_without_midair_quality_masks_21_frames_512
58
+ weight: 11.05
59
+ left_out_sources:
60
+ - dynamic_replica
61
+ - vkitti
62
+ - midair
63
+ - spring
64
+ - name: four_d_pretraining_relaunch_vkitti_source
65
+ kind: track4d_source_config
66
+ config: train/mot/wan_video_depth_vae_from_video_four_d_pretraining_relaunch_synthverse_dynamic_replica_vkitti_midair_spring_sources_without_midair_quality_masks_21_frames_512
67
+ weight: 6.0
68
+ left_out_sources:
69
+ - synthverse_depth
70
+ - dynamic_replica
71
+ - midair
72
+ - spring
73
+ - name: four_d_pretraining_relaunch_tartanair_sky_then_far_percentile_source
74
+ kind: track4d_source_config
75
+ config: train/mot/wan_video_depth_vae_from_video_four_d_pretraining_relaunch_sky_and_far_percentile_depth_validity_rules_21_frames_512
76
+ weight: 6.0
77
+ left_out_sources:
78
+ - dynamic_replica
79
+ - spring
80
+ - midair
81
+ - mvs_synth
82
+ - tartanground
83
+ - paralleldomain4d
84
+ - name: four_d_pretraining_relaunch_dynamic_replica_stored_depth_source
85
+ kind: track4d_source_config
86
+ config: train/mot/wan_video_depth_vae_from_video_four_d_pretraining_relaunch_sky_and_far_percentile_depth_validity_rules_21_frames_512
87
+ weight: 5.0
88
+ left_out_sources:
89
+ - tartanair
90
+ - spring
91
+ - midair
92
+ - mvs_synth
93
+ - tartanground
94
+ - paralleldomain4d
95
+ - name: four_d_pretraining_relaunch_spring_stored_depth_source
96
+ kind: track4d_source_config
97
+ config: train/mot/wan_video_depth_vae_from_video_four_d_pretraining_relaunch_sky_and_far_percentile_depth_validity_rules_21_frames_512
98
+ weight: 8.0
99
+ left_out_sources:
100
+ - tartanair
101
+ - dynamic_replica
102
+ - midair
103
+ - mvs_synth
104
+ - tartanground
105
+ - paralleldomain4d
106
+ - name: four_d_pretraining_relaunch_midair_within_200_meters_source
107
+ kind: track4d_source_config
108
+ config: train/mot/wan_video_depth_vae_from_video_four_d_pretraining_relaunch_sky_and_far_percentile_depth_validity_rules_21_frames_512
109
+ weight: 3.0
110
+ left_out_sources:
111
+ - tartanair
112
+ - dynamic_replica
113
+ - spring
114
+ - mvs_synth
115
+ - tartanground
116
+ - paralleldomain4d
117
+ - name: four_d_pretraining_relaunch_mvs_synth_1000_then_far_percentile_source
118
+ kind: track4d_source_config
119
+ config: train/mot/wan_video_depth_vae_from_video_four_d_pretraining_relaunch_sky_and_far_percentile_depth_validity_rules_21_frames_512
120
+ weight: 3.0
121
+ left_out_sources:
122
+ - tartanair
123
+ - dynamic_replica
124
+ - spring
125
+ - midair
126
+ - tartanground
127
+ - paralleldomain4d
128
+ - name: four_d_pretraining_relaunch_tartanground_sky_then_far_percentile_source
129
+ kind: track4d_source_config
130
+ config: train/mot/wan_video_depth_vae_from_video_four_d_pretraining_relaunch_sky_and_far_percentile_depth_validity_rules_21_frames_512
131
+ weight: 3.0
132
+ left_out_sources:
133
+ - tartanair
134
+ - dynamic_replica
135
+ - spring
136
+ - midair
137
+ - mvs_synth
138
+ - paralleldomain4d
139
+ - name: four_d_pretraining_relaunch_paralleldomain4d_sky_then_far_percentile_source
140
+ kind: track4d_source_config
141
+ config: train/mot/wan_video_depth_vae_from_video_four_d_pretraining_relaunch_sky_and_far_percentile_depth_validity_rules_21_frames_512
142
+ weight: 10.0
143
+ left_out_sources:
144
+ - tartanair
145
+ - dynamic_replica
146
+ - spring
147
+ - midair
148
+ - mvs_synth
149
+ - tartanground
150
+ frames: 9
151
+ raster_hw:
152
+ - 288
153
+ - 512
154
+ raymap_camera:
155
+ origin_fraction: 0.3333333333333333
156
+ origin_range: 1.0
157
+ decode_margin: 16
158
+ origin_box_grid: 1
159
+ pointmap_log_factor: 0.4
160
+ pointmap_hole_fill:
161
+ synthverse_depth: far_clip
162
+ syn4d: far_clip
163
+ omniworld_game: far_clip
164
+ dynamic_replica: nearest_labelled_depth_of_3x3_neighbours
165
+ tartanair: far_clip
166
+ dl3dv: far_clip
167
+ scannet: telea_log_depth
168
+ virtual_kitti2: far_clip
169
+ midair: far_clip
170
+ pointodyssey: far_clip
171
+ mvs_synth: far_clip
172
+ wildrgbd: telea_log_depth
173
+ hypersim: far_clip
174
+ spring: far_clip
175
+ kubric49: far_clip
176
+ kubric81: far_clip
177
+ droid_stereo_depth: telea_log_depth
178
+ tartanground: far_clip
179
+ paralleldomain4d: far_clip
180
+ pointmap_gauge: window_labelled_median_depth
181
+ geometry_image: genception_log_depth
182
+ context_len: 512
183
+ batch_size: 8
184
+ stream_seed: 42
185
+ epoch_windows: 32000000
186
+ val_set_proportion: 0.0
187
+ resume_state: null
188
+ gradient_accumulation_steps: 4
189
+ loader_workers: 8
190
+ loader_opencv_threads: 1
191
+ window_captions:
192
+ caption_dropout: 0.1
193
+ missing_caption: empty_prompt
194
+ caches:
195
+ syn4d_v1_gemini_holo4d:
196
+ embeddings: /scratch/shared/engs2907/logs/vgg_forecast_4d/cache/wan_umt5_captions/four_d_window_captions_512/syn4d_v1_gemini_holo4d_local_and_global_with_repaired_globals_20261031/captions_t5_emb.safetensors
197
+ meta: /scratch/shared/engs2907/logs/vgg_forecast_4d/cache/wan_umt5_captions/four_d_window_captions_512/syn4d_v1_gemini_holo4d_local_and_global_with_repaired_globals_20261031/captions_t5_emb_meta.json
198
+ syn4d_v2_gemini_holo4d:
199
+ embeddings: /scratch/shared/engs2907/logs/vgg_forecast_4d/cache/wan_umt5_captions/four_d_window_captions_512/syn4d_v2_gemini_holo4d_20261030/captions_t5_emb.safetensors
200
+ meta: /scratch/shared/engs2907/logs/vgg_forecast_4d/cache/wan_umt5_captions/four_d_window_captions_512/syn4d_v2_gemini_holo4d_20261030/captions_t5_emb_meta.json
201
+ tartanair_gemini_holo4d:
202
+ embeddings: /scratch/shared/engs2907/logs/vgg_forecast_4d/cache/wan_umt5_captions/four_d_window_captions_512/tartanair_gemini_holo4d/captions_t5_emb.safetensors
203
+ meta: /scratch/shared/engs2907/logs/vgg_forecast_4d/cache/wan_umt5_captions/four_d_window_captions_512/tartanair_gemini_holo4d/captions_t5_emb_meta.json
204
+ mvs_synth_gemini_holo4d:
205
+ embeddings: /scratch/shared/engs2907/logs/vgg_forecast_4d/cache/wan_umt5_captions/four_d_window_captions_512/mvs_synth_gemini_holo4d/captions_t5_emb.safetensors
206
+ meta: /scratch/shared/engs2907/logs/vgg_forecast_4d/cache/wan_umt5_captions/four_d_window_captions_512/mvs_synth_gemini_holo4d/captions_t5_emb_meta.json
207
+ omniworld_game_gemini_local_official_text_chunks_and_gemini_global:
208
+ embeddings: /scratch/shared/engs2907/logs/vgg_forecast_4d/cache/wan_umt5_captions/four_d_window_captions_512/omniworld_game_gemini_local_official_text_chunks_and_gemini_global_20261030b/captions_t5_emb.safetensors
209
+ meta: /scratch/shared/engs2907/logs/vgg_forecast_4d/cache/wan_umt5_captions/four_d_window_captions_512/omniworld_game_gemini_local_official_text_chunks_and_gemini_global_20261030b/captions_t5_emb_meta.json
210
+ synthverse_gemini_holo4d:
211
+ embeddings: /scratch/shared/engs2907/logs/vgg_forecast_4d/cache/wan_umt5_captions/four_d_window_captions_512/synthverse_gemini_holo4d_local96_and_global_20261031/captions_t5_emb.safetensors
212
+ meta: /scratch/shared/engs2907/logs/vgg_forecast_4d/cache/wan_umt5_captions/four_d_window_captions_512/synthverse_gemini_holo4d_local96_and_global_20261031/captions_t5_emb_meta.json
213
+ dynamic_replica_gemini_holo4d:
214
+ embeddings: /scratch/shared/engs2907/logs/vgg_forecast_4d/cache/wan_umt5_captions/four_d_window_captions_512/dynamic_replica_gemini_holo4d/captions_t5_emb.safetensors
215
+ meta: /scratch/shared/engs2907/logs/vgg_forecast_4d/cache/wan_umt5_captions/four_d_window_captions_512/dynamic_replica_gemini_holo4d/captions_t5_emb_meta.json
216
+ vkitti_gemini_holo4d:
217
+ embeddings: /scratch/shared/engs2907/logs/vgg_forecast_4d/cache/wan_umt5_captions/four_d_window_captions_512/vkitti_gemini_holo4d/captions_t5_emb.safetensors
218
+ meta: /scratch/shared/engs2907/logs/vgg_forecast_4d/cache/wan_umt5_captions/four_d_window_captions_512/vkitti_gemini_holo4d/captions_t5_emb_meta.json
219
+ midair_gemini_holo4d:
220
+ embeddings: /scratch/shared/engs2907/logs/vgg_forecast_4d/cache/wan_umt5_captions/four_d_window_captions_512/midair_gemini_holo4d_20261030/captions_t5_emb.safetensors
221
+ meta: /scratch/shared/engs2907/logs/vgg_forecast_4d/cache/wan_umt5_captions/four_d_window_captions_512/midair_gemini_holo4d_20261030/captions_t5_emb_meta.json
222
+ spring_gemini_holo4d:
223
+ embeddings: /scratch/shared/engs2907/logs/vgg_forecast_4d/cache/wan_umt5_captions/four_d_window_captions_512/spring_gemini_holo4d/captions_t5_emb.safetensors
224
+ meta: /scratch/shared/engs2907/logs/vgg_forecast_4d/cache/wan_umt5_captions/four_d_window_captions_512/spring_gemini_holo4d/captions_t5_emb_meta.json
225
+ tartanground_gemini_holo4d:
226
+ embeddings: /scratch/shared/engs2907/logs/vgg_forecast_4d/cache/wan_umt5_captions/four_d_window_captions_512/tartanground_gemini_holo4d_20261030/captions_t5_emb.safetensors
227
+ meta: /scratch/shared/engs2907/logs/vgg_forecast_4d/cache/wan_umt5_captions/four_d_window_captions_512/tartanground_gemini_holo4d_20261030/captions_t5_emb_meta.json
228
+ paralleldomain4d_gemini_holo4d:
229
+ embeddings: /scratch/shared/engs2907/logs/vgg_forecast_4d/cache/wan_umt5_captions/four_d_window_captions_512/paralleldomain4d_gemini_holo4d_20261030/captions_t5_emb.safetensors
230
+ meta: /scratch/shared/engs2907/logs/vgg_forecast_4d/cache/wan_umt5_captions/four_d_window_captions_512/paralleldomain4d_gemini_holo4d_20261030/captions_t5_emb_meta.json
231
+ stanzas:
232
+ syn4d_v1:
233
+ product: syn4d_v1_gemini_holo4d
234
+ join: local_covering_then_global
235
+ syn4d_v2_mixed:
236
+ product: syn4d_v2_gemini_holo4d
237
+ join: local_covering_then_global
238
+ syn4d_v2_mixed_no_sim:
239
+ product: syn4d_v2_gemini_holo4d
240
+ join: local_covering_then_global
241
+ tartanair:
242
+ product: tartanair_gemini_holo4d
243
+ join: local_covering
244
+ mvs_synth:
245
+ product: mvs_synth_gemini_holo4d
246
+ join: local_covering
247
+ omniworld_game:
248
+ product: omniworld_game_gemini_local_official_text_chunks_and_gemini_global
249
+ join: local_covering_then_second_local_covering_then_global
250
+ pointodyssey_test: none
251
+ kubric49_validation: none
252
+ wildrgbd_test: none
253
+ vkitti_test:
254
+ product: vkitti_gemini_holo4d
255
+ join: local_covering_then_global
256
+ droid_held_out: none
257
+ synthverse_depth:
258
+ product: synthverse_gemini_holo4d
259
+ join: local_covering_then_global
260
+ dynamic_replica:
261
+ product: dynamic_replica_gemini_holo4d
262
+ join: global
263
+ vkitti:
264
+ product: vkitti_gemini_holo4d
265
+ join: local_covering_then_global
266
+ midair:
267
+ product: midair_gemini_holo4d
268
+ join: local_covering_then_global
269
+ spring:
270
+ product: spring_gemini_holo4d
271
+ join: global
272
+ tartanground:
273
+ product: tartanground_gemini_holo4d
274
+ join: local_covering_then_global
275
+ paralleldomain4d:
276
+ product: paralleldomain4d_gemini_holo4d
277
+ join: global
278
+ paralleldomain4d_validation:
279
+ product: paralleldomain4d_gemini_holo4d
280
+ join: global
281
+ training_scene_exclusions:
282
+ reason: rendered depth passes through glass (raw EXR depth through glass panes
283
+ at 190-260 m), owner 10-29; evidence /scratch/shared/engs2907/logs/vgg_forecast_4d/tools/syn4d_glass_depth_check_20261029;
284
+ scifiroom_bald added 10-30 from /scratch/shared/engs2907/logs/vgg_forecast_4d/tools/depth_clipping_audit_20261030
285
+ stanzas:
286
+ syn4d_v1:
287
+ - bigoffice_v1
288
+ - lab_bald
289
+ - office_bald
290
+ - scifiroom_bald
291
+ syn4d_v2_mixed:
292
+ - citynight
293
+ - mall
294
+ - metro
295
+ syn4d_v2_mixed_no_sim:
296
+ - citynight
297
+ - mall
298
+ - metro
299
+ model:
300
+ _target_: flexpi_track_video.create_flexpi_track_video
301
+ model_id: Wan-AI/Wan2.2-TI2V-5B
302
+ tokenizer_model_id: Wan-AI/Wan2.1-T2V-1.3B
303
+ tokenizer_max_len: 128
304
+ load_text_encoder: false
305
+ proprio_dim: null
306
+ redirect_common_files: true
307
+ mot_checkpoint_mixed_attn: true
308
+ action_dit_pretrained_path: null
309
+ skip_dit_load_from_pretrain: false
310
+ dino_dim: 768
311
+ dino_model_name: vit_base_patch16_dinov3.lvd1689m
312
+ composite_layout: tshape_robotwin_384x320_uniform_dino_on_video_grid
313
+ dino_cam_patches: null
314
+ dino_cam_regions: null
315
+ dino_temporal_stride: 1
316
+ dino_stride_keep_far: true
317
+ dino_pred_x0: true
318
+ dino_pool_mode: avg
319
+ freeze_dino_encoder: true
320
+ dino_pixel_unshuffle: 2
321
+ enable_pointmap: true
322
+ pointmap_max_depth_m: 2.0
323
+ pointmap_norm_bounds:
324
+ x:
325
+ - -0.5
326
+ - 0.5
327
+ 'y':
328
+ - -0.5
329
+ - 0.5
330
+ z:
331
+ - 0.0
332
+ - 1.5
333
+ pointmap_depth_vis_mode: turbo
334
+ joint_video: true
335
+ joint_dino: true
336
+ joint_pointmap: true
337
+ flex_joint:
338
+ enabled: true
339
+ p_present_video: 0.5
340
+ p_present_dino: 0.5
341
+ p_present_pointmap: 0.5
342
+ p_jv: 0.5
343
+ p_jd: 0.5
344
+ p_jp: 0.5
345
+ cross_modal_predict_video: true
346
+ cross_modal_predict_dino: true
347
+ cross_modal_predict_pointmap: true
348
+ video_dit_config:
349
+ has_image_input: false
350
+ patch_size:
351
+ - 1
352
+ - 2
353
+ - 2
354
+ in_dim: 48
355
+ hidden_dim: 3072
356
+ ffn_dim: 14336
357
+ freq_dim: 256
358
+ text_dim: 4096
359
+ out_dim: 48
360
+ num_heads: 24
361
+ attn_head_dim: 128
362
+ num_layers: 30
363
+ eps: 1.0e-06
364
+ seperated_timestep: true
365
+ require_clip_embedding: false
366
+ require_vae_embedding: false
367
+ fuse_vae_embedding_in_latents: true
368
+ use_gradient_checkpointing: true
369
+ video_attention_mask_mode: first_frame_causal
370
+ action_conditioned: false
371
+ action_dim: 14
372
+ action_group_causal_mask_mode: group_diagonal
373
+ action_dit_config: null
374
+ video_scheduler:
375
+ train_shift: 6.0
376
+ infer_shift: 6.0
377
+ num_train_timesteps: 1000
378
+ action_scheduler:
379
+ train_shift: 1.0
380
+ infer_shift: 1.0
381
+ num_train_timesteps: 1000
382
+ dino_scheduler:
383
+ train_shift: 6.0
384
+ infer_shift: 6.0
385
+ num_train_timesteps: 1000
386
+ pointmap_scheduler:
387
+ train_shift: 6.0
388
+ infer_shift: 6.0
389
+ num_train_timesteps: 1000
390
+ loss:
391
+ lambda_action: 1.0
392
+ lambda_dino: 1.0
393
+ lambda_pointmap: 1.0
394
+ lambda_video: 1.0
395
+ hbridge:
396
+ enabled: true
397
+ bottom_ratio: 0.25
398
+ top_ratio: 0.25
399
+ track_video:
400
+ attention: visual_stream
401
+ sparse_action: true
402
+ train_shift: 6.0
403
+ infer_shift: 6.0
404
+ loss_lambda: 1.0
405
+ action_edge: two_way
406
+ flex:
407
+ p_present: 1.0
408
+ p_joint: 0.5
409
+ cross_modal_predict: true
410
+ noise_level_coupling: shared_uniform
411
+ raymap_camera:
412
+ origin_fraction: 0.3333333333333333
413
+ origin_range: 1.0
414
+ decode_margin: 16
415
+ origin_box_grid: 1
416
+ train_shift: 6.0
417
+ infer_shift: 6.0
418
+ loss_lambda: 1.0
419
+ four_d_pretraining:
420
+ geometry_streams: generated
421
+ dino_stream: pure_noise
422
+ video_loss_lambda: 1.0
423
+ modality_forcing:
424
+ video:
425
+ p_present: 0.5
426
+ p_joint: 0.5
427
+ cross_modal: true
428
+ dino:
429
+ p_present: 0.5
430
+ p_joint: 0.5
431
+ cross_modal: true
432
+ pointmap:
433
+ p_present: 0.5
434
+ p_joint: 0.5
435
+ cross_modal: true
436
+ camera:
437
+ p_present: 0.5
438
+ p_joint: 0.5
439
+ cross_modal: true
440
+ training_mode_shares:
441
+ flexible_joint: 1.0
442
+ v2d: 0.0
443
+ v2c: 0.0
444
+ c2v_from_first_frame: 0.0
445
+ c2vd_from_first_frame: 0.0
446
+ c2v: 0.0
447
+ c2vd: 0.0
448
+ vd2c: 0.0
449
+ vc2d: 0.0
450
+ v2dc: 0.0
451
+ dc2v_from_first_frame: 0.0
452
+ d2vc_from_first_frame: 0.0
453
+ d2v_from_first_frame: 0.0
454
+ dc2v: 0.0
455
+ d2vc: 0.0
456
+ d2v: 0.0
457
+ d2c: 0.0
458
+ c2d: 0.0
459
+ v2d_dino_generated: 0.0
460
+ v2c_dino_generated: 0.0
461
+ v2dc_dino_generated: 0.0
462
+ draw_rules:
463
+ first_frame_anchor: false
464
+ absent:
465
+ video: 0.0
466
+ dino: 0.0
467
+ pointmap: 0.0
468
+ camera: 0.0
469
+ whole_given:
470
+ video: 0.3333333333333333
471
+ dino: 0.3333333333333333
472
+ pointmap: 0.3333333333333333
473
+ camera: 0.3333333333333333
474
+ pointmap_log_factor: 0.4
475
+ trained_geometry_streams:
476
+ - camera
477
+ channel_concat: null
478
+ stage_one_without_action_expert: true
479
+ stream_only_end_blocks:
480
+ bottom: 7
481
+ top: 7
482
+ minimum_learning_rate: 1.0e-06
483
+ training_scope: the 5B stage-1 camera-interface comparison's arm with the video and
484
+ the RayMap camera image (the depth image and DINO absent), separate token streams,
485
+ LoRA rank 128, one shared generated noise level, global batch 32 on one B200, 5,000
486
+ updates of the 10,000-update schedule
487
+ native_control_fps: 50
488
+ action_dimension: 14
489
+ initialization: native_wan_actiondit
490
+ track_video_encoding:
491
+ record_root: /scratch/shared/engs2907/logs/vgg_forecast_4d/data/robotwin_robot_records_all_tasks_7bf528a2_20261023
492
+ embodiment_root: /scratch/shared/engs2907/checkouts/robotwin_objects_b5230d8b_266f3aad/RoboTwin/assets/embodiments/aloha-agilex
493
+ surface_points: 8192
494
+ min_surface_points_per_body: 16
495
+ query_count:
496
+ - 64
497
+ - 256
498
+ camera_query_quota:
499
+ - 224
500
+ - 16
501
+ - 16
502
+ min_visible_queries: 16
503
+ query_draw: per_read
504
+ query_seed: 0
505
+ visibility_margin_m: 0.01
506
+ pixel_center: 0.0
507
+ colouring: frame0_head_xyz
508
+ xyz_normalization: quantile_0.02_0.98
509
+ hues: null
510
+ neighbour_px: null
511
+ colour_floor: 0.2
512
+ min_gap_px:
513
+ - 8
514
+ - 6
515
+ - 6
516
+ wrist_dense_links:
517
+ - fl_link7
518
+ - fl_link8
519
+ - fr_link7
520
+ - fr_link8
521
+ wrist_dense_points: 16384
522
+ disc_radius_px: 2
523
+ occluded_style: ring
524
+ ring_radius_px: 3
525
+ ring_width_px: 1
526
+ deepspeed_gradient_clipping: true
527
+ warmup_steps: 500
528
+ loader_prefetch_factor: 2
529
+ loader_opencv_threads: 1
530
+ weights_milestone_every: 10000
531
+ fp32_milestone_every: null
532
+ train:
533
+ trainable_parameters:
534
+ mode: lora
535
+ lora:
536
+ rank: 128
537
+ alpha: 128
538
+ target_modules:
539
+ - all_block_linears
540
+ stop_after_updates: 5000