AndrewEstornell commited on
Commit
191a554
·
verified ·
1 Parent(s): 0d5bfcd

Upload online/Cplus_hardregion_beta/online_run_dump_config.yaml with huggingface_hub

Browse files
online/Cplus_hardregion_beta/online_run_dump_config.yaml ADDED
@@ -0,0 +1,781 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ CUDNN_BENCHMARK: false
2
+ DATALOADER:
3
+ ASPECT_RATIO_GROUPING: true
4
+ FILTER_EMPTY_ANNOTATIONS: false
5
+ NUM_WORKERS: 32
6
+ REPEAT_SQRT: true
7
+ REPEAT_THRESHOLD: 0.0
8
+ SAMPLER_TRAIN: TrainingSampler
9
+ DATASETS:
10
+ DATASET_NEED_MAP:
11
+ - false
12
+ DATASET_RATIO:
13
+ - 1.0
14
+ DATASET_TYPE:
15
+ - video_panoptic
16
+ DATASET_TYPE_TEST:
17
+ - video_panoptic
18
+ PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
19
+ PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
20
+ PROPOSAL_FILES_TEST: []
21
+ PROPOSAL_FILES_TRAIN: []
22
+ TEST:
23
+ - panoVSPW_vps_video_val
24
+ TRAIN:
25
+ - panoVSPW_vps_video_train
26
+ DEBUG:
27
+ OUTDIR: debug_vis
28
+ SAVE_EVERY: 1000
29
+ VIS_GRAD: false
30
+ FLOAT32_PRECISION: ''
31
+ GLOBAL:
32
+ HACK: 1.0
33
+ INPUT:
34
+ ADD_IMAGE_BRIGHT_CONTRAST: {}
35
+ ADD_IMAGE_ENHANCE:
36
+ ENABLED: false
37
+ AUGMENTATIONS: []
38
+ COLOR_AUG_SSD: false
39
+ CROP:
40
+ ENABLED: false
41
+ SINGLE_CATEGORY_MAX_AREA: 1.0
42
+ SIZE:
43
+ - 608
44
+ - 608
45
+ TYPE: absolute
46
+ DATASET_MAPPER_NAME: mask_former_semantic
47
+ FORMAT: RGB
48
+ GAUSSIAN_NOISE:
49
+ ENABLED: false
50
+ IMAGE_ENHANCE: null
51
+ IMAGE_SIZE: 1024
52
+ LSJ_AUG:
53
+ ENABLED: false
54
+ IMAGE_SIZE: 1024
55
+ MAX_SCALE: 2.0
56
+ MIN_SCALE: 0.1
57
+ MASK_FORMAT: polygon
58
+ MAX_SCALE: 2.0
59
+ MAX_SIZE_TEST: 1333
60
+ MAX_SIZE_TRAIN: 1920
61
+ MIN_SCALE: 0.1
62
+ MIN_SIZE_TEST: 1280
63
+ MIN_SIZE_TRAIN:
64
+ - 960
65
+ - 1120
66
+ - 1280
67
+ MIN_SIZE_TRAIN_SAMPLING: choice
68
+ PSEUDO:
69
+ AUGMENTATIONS:
70
+ - rotation
71
+ CROP:
72
+ ENABLED: false
73
+ SIZE:
74
+ - 608
75
+ - 608
76
+ TYPE: absolute
77
+ MAX_SIZE_TRAIN: 1333
78
+ MIN_SIZE_TRAIN:
79
+ - 480
80
+ - 512
81
+ - 544
82
+ - 576
83
+ - 608
84
+ - 640
85
+ - 672
86
+ - 704
87
+ - 736
88
+ - 768
89
+ - 800
90
+ MIN_SIZE_TRAIN_SAMPLING: choice_by_clip
91
+ RANDOM_FLIP: flip
92
+ REVERSE_AGU: false
93
+ SAMPLING_FRAME_NUM: 5
94
+ SAMPLING_FRAME_RANGE: 5
95
+ SAMPLING_FRAME_RATIO: 1.0
96
+ SAMPLING_FRAME_SHUFFLE: false
97
+ SIZE_DIVISIBILITY: -1
98
+ MODEL:
99
+ ANCHOR_GENERATOR:
100
+ ANGLES:
101
+ - - -90
102
+ - 0
103
+ - 90
104
+ ASPECT_RATIOS:
105
+ - - 0.5
106
+ - 1.0
107
+ - 2.0
108
+ NAME: DefaultAnchorGenerator
109
+ OFFSET: 0.0
110
+ SIZES:
111
+ - - 32
112
+ - 64
113
+ - 128
114
+ - 256
115
+ - 512
116
+ BACKBONE:
117
+ FREEZE_AT: 0
118
+ NAME: D2VitAdapterDinoV2
119
+ CL_PLUGIN:
120
+ AUX_REID_WEIGHT: 3.0
121
+ BIO_CL: false
122
+ CL_PLUGIN_NAME: CTCLPlugin
123
+ FUSION_LOSS: false
124
+ MOMENTUM_EMBED: true
125
+ NOISE_EMBED: false
126
+ NUM_NEGATIVES: 199
127
+ ONE_DIRECTION: true
128
+ REID_WEIGHT: 2.0
129
+ DEVICE: cuda
130
+ FPN:
131
+ FUSE_TYPE: sum
132
+ IN_FEATURES: []
133
+ NORM: ''
134
+ OUT_CHANNELS: 256
135
+ KEYPOINT_ON: false
136
+ LOAD_PROPOSALS: false
137
+ MASK_FORMER:
138
+ CLASS_WEIGHT: 2.0
139
+ DEC_LAYERS: 10
140
+ DEEP_SUPERVISION: true
141
+ DICE_WEIGHT: 5.0
142
+ DIM_FEEDFORWARD: 2048
143
+ DROPOUT: 0.0
144
+ ENC_LAYERS: 0
145
+ ENFORCE_INPUT_PROJ: false
146
+ EOS_NEG_POS_RATIO: 3
147
+ HIDDEN_DIM: 256
148
+ IMPORTANCE_SAMPLE_RATIO: 0.75
149
+ MASK_WEIGHT: 5.0
150
+ MIN_EOS_KEEP: 10
151
+ NHEADS: 8
152
+ NO_OBJECT_WEIGHT: 0.1
153
+ NUM_OBJECT_QUERIES: 200
154
+ NUM_REID_HEAD_LAYERS: 3
155
+ OVERSAMPLE_RATIO: 3.0
156
+ PRE_NORM: false
157
+ REID_BRANCH: true
158
+ REID_HIDDEN_DIM: 256
159
+ SIZE_DIVISIBILITY: 32
160
+ TEST:
161
+ INSTANCE_ON: false
162
+ MAX_NUM: 20
163
+ OBJECT_MASK_THRESHOLD: 0.8
164
+ OVERLAP_THRESHOLD: 0.8
165
+ PANOPTIC_ON: false
166
+ SEMANTIC_ON: true
167
+ SEM_SEG_POSTPROCESSING_BEFORE_INFERENCE: false
168
+ TASK: vps
169
+ WINDOW_INFERENCE: true
170
+ WINDOW_SIZE: 3
171
+ TRAIN_NUM_POINTS: 12544
172
+ TRANSFORMER_DECODER_NAME: VideoMultiScaleMaskedTransformerDecoder_dvisPlus
173
+ TRANSFORMER_IN_FEATURE: multi_scale_pixel_decoder
174
+ MASK_ON: true
175
+ META_ARCHITECTURE: DVIS_Plus_online
176
+ PANOPTIC_FPN:
177
+ COMBINE:
178
+ ENABLED: true
179
+ INSTANCES_CONFIDENCE_THRESH: 0.5
180
+ OVERLAP_THRESH: 0.5
181
+ STUFF_AREA_LIMIT: 4096
182
+ INSTANCE_LOSS_WEIGHT: 1.0
183
+ PIXEL_MEAN:
184
+ - 123.675
185
+ - 116.28
186
+ - 103.53
187
+ PIXEL_STD:
188
+ - 58.395
189
+ - 57.12
190
+ - 57.375
191
+ PROPOSAL_GENERATOR:
192
+ MIN_SIZE: 0
193
+ NAME: RPN
194
+ REFINER:
195
+ APP_PIXEL_XATTN: false
196
+ APP_PIXEL_XATTN_STRIDE: 4
197
+ APP_POOL_FUSE: false
198
+ DECODER_LAYERS: 6
199
+ USE_CL: true
200
+ RESNETS:
201
+ DEFORM_MODULATED: false
202
+ DEFORM_NUM_GROUPS: 1
203
+ DEFORM_ON_PER_STAGE:
204
+ - false
205
+ - false
206
+ - false
207
+ - false
208
+ DEPTH: 50
209
+ NORM: FrozenBN
210
+ NUM_GROUPS: 1
211
+ OUT_FEATURES:
212
+ - res2
213
+ - res3
214
+ - res4
215
+ - res5
216
+ RES2_OUT_CHANNELS: 256
217
+ RES4_DILATION: 1
218
+ RES5_DILATION: 1
219
+ RES5_MULTI_GRID:
220
+ - 1
221
+ - 1
222
+ - 1
223
+ STEM_OUT_CHANNELS: 64
224
+ STEM_TYPE: basic
225
+ STRIDE_IN_1X1: false
226
+ WIDTH_PER_GROUP: 64
227
+ RETINANET:
228
+ BBOX_REG_LOSS_TYPE: smooth_l1
229
+ BBOX_REG_WEIGHTS: &id002
230
+ - 1.0
231
+ - 1.0
232
+ - 1.0
233
+ - 1.0
234
+ FOCAL_LOSS_ALPHA: 0.25
235
+ FOCAL_LOSS_GAMMA: 2.0
236
+ IN_FEATURES:
237
+ - p3
238
+ - p4
239
+ - p5
240
+ - p6
241
+ - p7
242
+ IOU_LABELS:
243
+ - 0
244
+ - -1
245
+ - 1
246
+ IOU_THRESHOLDS:
247
+ - 0.4
248
+ - 0.5
249
+ NMS_THRESH_TEST: 0.5
250
+ NORM: ''
251
+ NUM_CLASSES: 80
252
+ NUM_CONVS: 4
253
+ PRIOR_PROB: 0.01
254
+ SCORE_THRESH_TEST: 0.05
255
+ SMOOTH_L1_LOSS_BETA: 0.1
256
+ TOPK_CANDIDATES_TEST: 1000
257
+ ROI_BOX_CASCADE_HEAD:
258
+ BBOX_REG_WEIGHTS:
259
+ - &id001
260
+ - 10.0
261
+ - 10.0
262
+ - 5.0
263
+ - 5.0
264
+ - - 20.0
265
+ - 20.0
266
+ - 10.0
267
+ - 10.0
268
+ - - 30.0
269
+ - 30.0
270
+ - 15.0
271
+ - 15.0
272
+ IOUS:
273
+ - 0.5
274
+ - 0.6
275
+ - 0.7
276
+ ROI_BOX_HEAD:
277
+ BBOX_REG_LOSS_TYPE: smooth_l1
278
+ BBOX_REG_LOSS_WEIGHT: 1.0
279
+ BBOX_REG_WEIGHTS: *id001
280
+ CLS_AGNOSTIC_BBOX_REG: false
281
+ CONV_DIM: 256
282
+ FC_DIM: 1024
283
+ FED_LOSS_FREQ_WEIGHT_POWER: 0.5
284
+ FED_LOSS_NUM_CLASSES: 50
285
+ NAME: ''
286
+ NORM: ''
287
+ NUM_CONV: 0
288
+ NUM_FC: 0
289
+ POOLER_RESOLUTION: 14
290
+ POOLER_SAMPLING_RATIO: 0
291
+ POOLER_TYPE: ROIAlignV2
292
+ SMOOTH_L1_BETA: 0.0
293
+ TRAIN_ON_PRED_BOXES: false
294
+ USE_FED_LOSS: false
295
+ USE_SIGMOID_CE: false
296
+ ROI_HEADS:
297
+ BATCH_SIZE_PER_IMAGE: 512
298
+ IN_FEATURES:
299
+ - res4
300
+ IOU_LABELS:
301
+ - 0
302
+ - 1
303
+ IOU_THRESHOLDS:
304
+ - 0.5
305
+ NAME: Res5ROIHeads
306
+ NMS_THRESH_TEST: 0.5
307
+ NUM_CLASSES: 80
308
+ POSITIVE_FRACTION: 0.25
309
+ PROPOSAL_APPEND_GT: true
310
+ SCORE_THRESH_TEST: 0.05
311
+ ROI_KEYPOINT_HEAD:
312
+ CONV_DIMS:
313
+ - 512
314
+ - 512
315
+ - 512
316
+ - 512
317
+ - 512
318
+ - 512
319
+ - 512
320
+ - 512
321
+ LOSS_WEIGHT: 1.0
322
+ MIN_KEYPOINTS_PER_IMAGE: 1
323
+ NAME: KRCNNConvDeconvUpsampleHead
324
+ NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: true
325
+ NUM_KEYPOINTS: 17
326
+ POOLER_RESOLUTION: 14
327
+ POOLER_SAMPLING_RATIO: 0
328
+ POOLER_TYPE: ROIAlignV2
329
+ ROI_MASK_HEAD:
330
+ CLS_AGNOSTIC_MASK: false
331
+ CONV_DIM: 256
332
+ NAME: MaskRCNNConvUpsampleHead
333
+ NORM: ''
334
+ NUM_CONV: 0
335
+ POOLER_RESOLUTION: 14
336
+ POOLER_SAMPLING_RATIO: 0
337
+ POOLER_TYPE: ROIAlignV2
338
+ RPN:
339
+ BATCH_SIZE_PER_IMAGE: 256
340
+ BBOX_REG_LOSS_TYPE: smooth_l1
341
+ BBOX_REG_LOSS_WEIGHT: 1.0
342
+ BBOX_REG_WEIGHTS: *id002
343
+ BOUNDARY_THRESH: -1
344
+ CONV_DIMS:
345
+ - -1
346
+ HEAD_NAME: StandardRPNHead
347
+ IN_FEATURES:
348
+ - res4
349
+ IOU_LABELS:
350
+ - 0
351
+ - -1
352
+ - 1
353
+ IOU_THRESHOLDS:
354
+ - 0.3
355
+ - 0.7
356
+ LOSS_WEIGHT: 1.0
357
+ NMS_THRESH: 0.7
358
+ POSITIVE_FRACTION: 0.5
359
+ POST_NMS_TOPK_TEST: 1000
360
+ POST_NMS_TOPK_TRAIN: 2000
361
+ PRE_NMS_TOPK_TEST: 6000
362
+ PRE_NMS_TOPK_TRAIN: 12000
363
+ SMOOTH_L1_BETA: 0.0
364
+ SEM_SEG_HEAD:
365
+ ASPP_CHANNELS: 256
366
+ ASPP_DILATIONS:
367
+ - 6
368
+ - 12
369
+ - 18
370
+ ASPP_DROPOUT: 0.1
371
+ COMMON_STRIDE: 4
372
+ CONVS_DIM: 256
373
+ DEFORMABLE_TRANSFORMER_ENCODER_IN_FEATURES:
374
+ - res3
375
+ - res4
376
+ - res5
377
+ DEFORMABLE_TRANSFORMER_ENCODER_N_HEADS: 8
378
+ DEFORMABLE_TRANSFORMER_ENCODER_N_POINTS: 4
379
+ IGNORE_VALUE: 255
380
+ IN_FEATURES:
381
+ - res2
382
+ - res3
383
+ - res4
384
+ - res5
385
+ LOSS_TYPE: hard_pixel_mining
386
+ LOSS_WEIGHT: 1.0
387
+ MASK_DIM: 256
388
+ NAME: MaskFormerHead
389
+ NORM: GN
390
+ NUM_CLASSES: 29
391
+ PIXEL_DECODER_NAME: MSDeformAttnPixelDecoder
392
+ PROJECT_CHANNELS:
393
+ - 48
394
+ PROJECT_FEATURES:
395
+ - res2
396
+ RETURN_TRANSFORMER_FEATURE: false
397
+ TRANSFORMER_ENC_LAYERS: 6
398
+ USE_DEPTHWISE_SEPARABLE_CONV: false
399
+ SWIN:
400
+ APE: false
401
+ ATTN_DROP_RATE: 0.0
402
+ DEPTHS:
403
+ - 2
404
+ - 2
405
+ - 6
406
+ - 2
407
+ DROP_PATH_RATE: 0.3
408
+ DROP_RATE: 0.0
409
+ EMBED_DIM: 96
410
+ MLP_RATIO: 4.0
411
+ NUM_HEADS:
412
+ - 3
413
+ - 6
414
+ - 12
415
+ - 24
416
+ OUT_FEATURES:
417
+ - res2
418
+ - res3
419
+ - res4
420
+ - res5
421
+ PATCH_NORM: true
422
+ PATCH_SIZE: 4
423
+ PRETRAIN_IMG_SIZE: 224
424
+ QKV_BIAS: true
425
+ QK_SCALE: null
426
+ USE_CHECKPOINT: false
427
+ WINDOW_SIZE: 7
428
+ TRACKER:
429
+ APP_CL: false
430
+ APP_PIXEL_XATTN: false
431
+ APP_PIXEL_XATTN_STRIDE: 4
432
+ APP_POOL_FUSE: false
433
+ DECODER_LAYERS: 6
434
+ NOISE_MODE: wa
435
+ NOISE_RATIO: 0.5
436
+ USE_CL: false
437
+ VIT_ADAPTER:
438
+ FINETUNE: false
439
+ FINETUNE_INDEXES:
440
+ - 0
441
+ FREEZE_VIT: true
442
+ NAME: vitl
443
+ VIT_WEIGHT: /mnt/hdfs/andrew.estornell/multi_modal/repo_backup/weights/work_dirs/dinov2_vitl14_pretrain_.pth
444
+ WITH_CP: true
445
+ WEIGHTS: /mnt/hdfs/andrew.estornell/multi_modal/DVIS_Plus/new_output_dir/new_waymo_config1/baseline_stage1/model_final.pth
446
+ NUM_FRAMES: 4
447
+ NUM_VIDEOS: 1
448
+ OPT_5_SLIDING_WINDOW_STICH: false
449
+ OPT_ADD_CONSTRAINTS: false
450
+ OPT_ADD_CROP_RESIZE: false
451
+ OPT_ADD_DAUG: add_data_aug
452
+ OPT_ADD_FOCAL_LOSS: false
453
+ OPT_ADD_HARD_OBJECT: false
454
+ OPT_ADD_HORIZONTAL_CONCATENATION: false
455
+ OPT_ADD_HORIZONTAL_MERGE: false
456
+ OPT_ADD_LEARNING_SCHEDULER: false
457
+ OPT_ADD_Multi_Num_Hops: false
458
+ OPT_ADD_OBJECTS_MERGE_MODEL: false
459
+ OPT_ADD_REG_KL: false
460
+ OPT_ADD_SECOND_RES: false
461
+ OPT_ADD_SEMANTIC_CLOSE_LABELS: false
462
+ OPT_ADD_STUFF_ERROR_ALL: false
463
+ OPT_ALPHA2_DICE_WEIGHT: 0.1
464
+ OPT_ALPHA2_MASK_WEIGHT: 0.1
465
+ OPT_ALPHA_DICE_WEIGHT: 1.0
466
+ OPT_ALPHA_MASK_WEIGHT: 0.1
467
+ OPT_AQ_EVAL_AT_START: true
468
+ OPT_AQ_EVAL_CACHE: /opt/tiger/multi_modal/downloads/valcache
469
+ OPT_AQ_EVAL_EVERY: 200
470
+ OPT_AQ_EVAL_MAXCLIPS: 120
471
+ OPT_ATTN_MASK_METHOD: attn_mask
472
+ OPT_BACKBONE_INFERENCE_FRAME: 3
473
+ OPT_BETA2_DICE_WEIGHT: 0.1
474
+ OPT_BETA2_MASK_WEIGHT: 0.1
475
+ OPT_BETA_CE_ENABLED: false
476
+ OPT_BETA_CE_METHOD: 1
477
+ OPT_BETA_CE_WEIGHT: 0.0
478
+ OPT_BETA_DICE_WEIGHT: 3.0
479
+ OPT_BETA_MASK_WEIGHT: 1.0
480
+ OPT_BND_R: 1
481
+ OPT_CAMERA: null
482
+ OPT_CB_AGGRESSIVE_EXCLUDED_WEIGHT: 0.001
483
+ OPT_CB_AGGRESSIVE_KEEP_CLASSES: []
484
+ OPT_CB_LOSS2_ALPHA_DICE: false
485
+ OPT_CB_LOSS2_ALPHA_MASK: false
486
+ OPT_CB_LOSS2_BETA_DICE: false
487
+ OPT_CB_LOSS2_BETA_MASK: false
488
+ OPT_CHECK_COMPUTATION_TIME: false
489
+ OPT_CHUNK_NUMBER: null
490
+ OPT_CLASS_BALANCED_CE_BETA: 0.0
491
+ OPT_CLASS_BALANCED_MASK_DICE: false
492
+ OPT_CLASS_COUNTS_JSON: ''
493
+ OPT_CONCATENATE_ALL_CAMERAS: null
494
+ OPT_CONSTRAINT_WEIGHT: 1.0
495
+ OPT_CUSTOM_KL: false
496
+ OPT_CUSTOM_TRAINER: false
497
+ OPT_DATASET: waymo_train_all_5_views
498
+ OPT_DATA_MAPPER: null
499
+ OPT_DATA_MAPPER_VISION: null
500
+ OPT_DEBUG_FILENAME: null
501
+ OPT_DEBUG_FILE_MULTIPLE: false
502
+ OPT_DEBUG_MODE: false
503
+ OPT_DEBUG_SAVE_PATH: null
504
+ OPT_DEBUG_TRAINING: false
505
+ OPT_EASY_REGION_DISTILL: false
506
+ OPT_EASY_REGION_DISTILL_METHOD: 1
507
+ OPT_EASY_REGION_DISTILL_WEIGHT: 1.0
508
+ OPT_EMA_TEACHER: false
509
+ OPT_EMA_TEACHER_DECAY: 0.9999
510
+ OPT_ENLARGE_INPUT: false
511
+ OPT_EP_WEIGHT: 1.0
512
+ OPT_EVALUATION: false
513
+ OPT_EVAL_ITERATION: 1000
514
+ OPT_EVAL_ONLY: false
515
+ OPT_EVAL_PRINT_PER_CLASS_IOU: false
516
+ OPT_Ensemble_Method: null
517
+ OPT_Ensemble_Temporal: null
518
+ OPT_FILENAMES: null
519
+ OPT_FILE_NAME: null
520
+ OPT_FILTER_ABSENT_OBJECTS: false
521
+ OPT_FILT_ONLY: false
522
+ OPT_FIX_BOUNDARY: false
523
+ OPT_FIX_DIMENSION: false
524
+ OPT_FLAG_ABSENT_IDS: true
525
+ OPT_FREEZE_EXCEPT_CLASS_EMBED: false
526
+ OPT_FULL_IMAGE: false
527
+ OPT_Fine_Tune_IoU: false
528
+ OPT_Full_Threshold: 0.9
529
+ OPT_GA: true
530
+ OPT_GATE_MODE: null
531
+ OPT_GATE_WEIGHT: 0.8
532
+ OPT_GA_STEPS: 32
533
+ OPT_GET_INPUT_OUTPUT: false
534
+ OPT_GET_TARGET_LABEL_VIDEO_ID: false
535
+ OPT_GTWP_WEIGHT: 0.5
536
+ OPT_HARD_REGION_CLASS_AWARE: false
537
+ OPT_HARD_REGION_LABELS:
538
+ - 3
539
+ - 4
540
+ - 5
541
+ OPT_HARD_REGION_LABEL_SUBSET: false
542
+ OPT_HARD_REGION_SEMANTIC: false
543
+ OPT_HARD_REGION_SRC: matched
544
+ OPT_HARD_REGION_START_STEP: 0
545
+ OPT_HC_NUM_POINTS_DIVIDER: 5
546
+ OPT_HRP_GAMMA: 1.0
547
+ OPT_Horizontal_INTEREST_STUFF_LABELS: null
548
+ OPT_Horizontal_INTEREST_THING_LABELS: null
549
+ OPT_Horizontal_View: false
550
+ OPT_IF_ADD_COLOR_JITTER: false
551
+ OPT_IF_ADD_RANDOM_CROP: true
552
+ OPT_IF_ADD_RANDOM_FLIP: true
553
+ OPT_INFERENCE: false
554
+ OPT_INFERENCE_ALL_VIEW_CAMERAS: false
555
+ OPT_INFERENCE_SIDE_VIEWS: false
556
+ OPT_INFERENCE_SIDE_VIEWS_PERSON: false
557
+ OPT_INFERENCE_STAGE: null
558
+ OPT_INF_FULL: false
559
+ OPT_INF_HORZONTAL: false
560
+ OPT_INPUT_ADD_SEM_RGB: false
561
+ OPT_INTEREST_STUFF_LABELS: null
562
+ OPT_INTEREST_THING_LABELS: null
563
+ OPT_IS_TEST: false
564
+ OPT_Input_Method: null
565
+ OPT_KD1_ANNEAL: false
566
+ OPT_KD1_CE_WEIGHT: 2.0
567
+ OPT_KD1_DICE_WEIGHT: 5.0
568
+ OPT_KD1_ENABLED: false
569
+ OPT_KD1_EVAL_HARDREGION: false
570
+ OPT_KD1_EVAL_MAX_VIDEOS: 0
571
+ OPT_KD1_EVAL_PERIOD: 0
572
+ OPT_KD1_EVAL_VIEW: front
573
+ OPT_KD1_GATE: false
574
+ OPT_KD1_GATE_IOU: 0.5
575
+ OPT_KD1_HARD_REGION: false
576
+ OPT_KD1_HR_ALPHA_DICE_W: 1.0
577
+ OPT_KD1_HR_ALPHA_MASK_W: 0.1
578
+ OPT_KD1_HR_ANNEAL: false
579
+ OPT_KD1_HR_BETA_DICE_W: 3.0
580
+ OPT_KD1_HR_BETA_MASK_W: 1.0
581
+ OPT_KD1_HR_CLASS_AWARE: true
582
+ OPT_KD1_HR_NUM_POINTS: 0
583
+ OPT_KD1_LAMBDA: 1.0
584
+ OPT_KD1_MASK_WEIGHT: 5.0
585
+ OPT_KD1_MATCH: student_teacher
586
+ OPT_KD1_STAGED: false
587
+ OPT_KD1_STAGED_B1: 0.33
588
+ OPT_KD1_STAGED_B2: 0.66
589
+ OPT_KD1_T: 2.0
590
+ OPT_KD1_TEACHER2_CONFIG: null
591
+ OPT_KD1_TEACHER2_WEIGHTS: null
592
+ OPT_KD1_TEACHER_CONFIG: null
593
+ OPT_KD1_TEACHER_WEIGHTS: null
594
+ OPT_KD1_TWO_TEACHER_MODE: none
595
+ OPT_LABEL_GROUP: null
596
+ OPT_LABEL_INDENTIFICATION: false
597
+ OPT_LABEL_PAIR_TRAIN_METHOD: null
598
+ OPT_LAMBDA: 1.0
599
+ OPT_LISTEN_REFERENCE: false
600
+ OPT_LOAD_RARE_CLASS_LIST: false
601
+ OPT_LOCALIZATION: null
602
+ OPT_LOSS_DEF: 1
603
+ OPT_LOSS_SCALER: 1.0
604
+ OPT_LOW_THRESHOLD: null
605
+ OPT_LVIS_REPEAT_THRESH: 0.0
606
+ OPT_MASK_KL_TOL_EPS: 0.1
607
+ OPT_MASK_KL_WEIGHT: 0.1
608
+ OPT_MERGE_RES_FULL: false
609
+ OPT_MERGE_RES_FULL_HORIZONTAL: false
610
+ OPT_METHOD: vanilla
611
+ OPT_MULTI_HORIZON: false
612
+ OPT_Merge_Edge_Objects: false
613
+ OPT_NBCE_KL: false
614
+ OPT_NBCE_KL_REGION: all
615
+ OPT_NBCE_KL_WEIGHT: 1.0
616
+ OPT_NIGHT: false
617
+ OPT_NUM_FILES: null
618
+ OPT_NUM_HOPS: 20
619
+ OPT_NUM_VIDEOS: null
620
+ OPT_ONLY_CONSTRAINTS: false
621
+ OPT_ONLY_FULL: false
622
+ OPT_ONLY_MASK_KL: false
623
+ OPT_ONLY_RESIDUAL: false
624
+ OPT_ONLY_SAVE_MINI_VIDEOS: false
625
+ OPT_PATCH_NUMBER: 1
626
+ OPT_PINPOINT_PATCHES: false
627
+ OPT_PPRIME_AUX_METHOD: plain
628
+ OPT_PPRIME_AUX_TEACHER: false
629
+ OPT_PPRIME_AUX_WEIGHT: 1.0
630
+ OPT_PPRIME_EVAL_CACHE: /opt/tiger/multi_modal/downloads/valcache
631
+ OPT_PPRIME_EVAL_EVERY: 0
632
+ OPT_PPRIME_EVAL_LOSS: true
633
+ OPT_PPRIME_EVAL_LOSS_CLIPS: 12
634
+ OPT_PPRIME_EVAL_MAXCLIPS: 60
635
+ OPT_PPRIME_FREEZE_GLUE: true
636
+ OPT_PPRIME_FREEZE_UPSTREAM: false
637
+ OPT_PROJECT_GRAD: false
638
+ OPT_QUERY_DISTILL_LOSS: false
639
+ OPT_QUERY_DISTILL_WEIGHT: 1.0
640
+ OPT_RARE_CLASS: false
641
+ OPT_RARE_CLASS_BOOST: 1.0
642
+ OPT_RARE_CLASS_BOOST_LIST: []
643
+ OPT_RARE_STUFF_GROUP: false
644
+ OPT_RARE_TARGET_LABEL: null
645
+ OPT_REGION_EVAL: false
646
+ OPT_REGION_EVAL_MAX_VIDEOS: 0
647
+ OPT_REGION_EVAL_PERIOD: 1000
648
+ OPT_REGION_EVAL_VIEW: front
649
+ OPT_REGION_WEIGHTED_ONLY: false
650
+ OPT_REMOVE_OTHER_CLASSES: false
651
+ OPT_REMOVE_OTHER_LABELS: false
652
+ OPT_REPLACE_BCE_WITH_FOCAL_LOSS: false
653
+ OPT_RES_STRATEGY: 0
654
+ OPT_REUSE_TEACHER_MATCHING: false
655
+ OPT_RW_EASY_WEIGHT: 0.3
656
+ OPT_RW_HARD_WEIGHT: 0.7
657
+ OPT_R_W: 2
658
+ OPT_Residual_Input: full
659
+ OPT_Residual_Num_Hops: 10
660
+ OPT_Residual_Threshold: 0.5
661
+ OPT_SAVE_LABEL_SCORE: false
662
+ OPT_SAVE_LONG_VIEW: false
663
+ OPT_SAVE_OUTPUT: false
664
+ OPT_SAVE_PRED_SCORE: false
665
+ OPT_SAVE_RES_PATH: null
666
+ OPT_SCALE_NUMBER: null
667
+ OPT_SEMANTIC_CLOSE_LABEL_GROUP: null
668
+ OPT_SEM_DISTILL: false
669
+ OPT_SEM_DISTILL_MODE: option1
670
+ OPT_SEM_DISTILL_T: 1.0
671
+ OPT_SEM_DISTILL_WEIGHT: 1.0
672
+ OPT_SIDE_CAMERA_VIEWS: false
673
+ OPT_SKIP_DIQ_IN_KEYREF: false
674
+ OPT_START_FRAME_NUMBER: null
675
+ OPT_STUDENT_EMA: false
676
+ OPT_STUDENT_EMA_DECAY: 0.999
677
+ OPT_STUFF: false
678
+ OPT_Several_Patches: false
679
+ OPT_SmallPatch_LABELS: null
680
+ OPT_TAIL_FRAMES: false
681
+ OPT_TARGET_LABEL: null
682
+ OPT_TARGET_LABEL_TRAIN_STAGE: null
683
+ OPT_TEACHER2_CONFIG: null
684
+ OPT_TEACHER2_WEIGHTS: null
685
+ OPT_TEACHER_CONFIG: /mnt/hdfs/andrew.estornell/multi_modal/DVIS_Plus/new_output_dir/new_waymo_config1/full_online/config.yaml
686
+ OPT_TEACHER_HARD_REFRESH_PERIOD: 0
687
+ OPT_TEACHER_WEIGHTS: /mnt/hdfs/andrew.estornell/multi_modal/DVIS_Plus/new_output_dir/new_waymo_config1/full_online/model_final.pth
688
+ OPT_TEMPERATURE: 1.0
689
+ OPT_TEST_TARGET: null
690
+ OPT_THRESHOLD: 0.8
691
+ OPT_TIME_STEP: 1
692
+ OPT_TRACKER_BETA_HEAD: true
693
+ OPT_TRACKER_HR_CLASS_AWARE: false
694
+ OPT_TRACKER_KD: true
695
+ OPT_TRACKING_VIS_PATH: null
696
+ OPT_TRAIN_LOSS: vanilla
697
+ OPT_TRAIN_MULTI_RES: false
698
+ OPT_TRAIN_STAGE: null
699
+ OPT_TRAIN_STAGE_2: false
700
+ OPT_TRAIN_STRATEGY: 0
701
+ OPT_TWO_POLICY_PARTITION: false
702
+ OPT_Train_Single_Target_Label: false
703
+ OPT_USE_FINAL_TEACHER_HARD_REGION_FOR_ALL_LAYERS: false
704
+ OPT_USE_KDCB_CRITERION: false
705
+ OPT_VALIDATION_VIDEO_ID: null
706
+ OPT_VAL_DATASET: waymo_train_5_views
707
+ OPT_VANILLA: false
708
+ OPT_VERY_SMALL_PATCHES: false
709
+ OPT_VERY_TINY_PATCHES: false
710
+ OPT_VIDEO_ID_str: null
711
+ OPT_VIEW: null
712
+ OPT_View_Merge_Method: null
713
+ OPT_evaluate_with_KD: false
714
+ OUTPUT_DIR: /mnt/hdfs/andrew.estornell/multi_modal/DVIS_Plus/new_output_dir/new_waymo_config1/full_online_ga32_hardregion_beta
715
+ SEED: 42
716
+ SOLVER:
717
+ AMP:
718
+ ENABLED: true
719
+ BACKBONE_MULTIPLIER: 0.1
720
+ BASE_LR: 0.000283
721
+ BASE_LR_END: 0.0
722
+ BIAS_LR_FACTOR: 1.0
723
+ CHECKPOINT_PERIOD: 10000
724
+ CLIP_GRADIENTS:
725
+ CLIP_TYPE: full_model
726
+ CLIP_VALUE: 0.01
727
+ ENABLED: true
728
+ NORM_TYPE: 2.0
729
+ GAMMA: 0.1
730
+ IMS_PER_BATCH: 8
731
+ LR_SCHEDULER_NAME: WarmupMultiStepLR
732
+ MAX_ITER: 106656
733
+ MOMENTUM: 0.9
734
+ NESTEROV: false
735
+ NUM_DECAYS: 3
736
+ OPTIMIZER: ADAMW
737
+ POLY_LR_CONSTANT_ENDING: 0.0
738
+ POLY_LR_POWER: 0.9
739
+ REFERENCE_WORLD_SIZE: 0
740
+ RESCALE_INTERVAL: false
741
+ STEPS:
742
+ - 74656
743
+ - 96000
744
+ WARMUP_FACTOR: 0.001
745
+ WARMUP_ITERS: 12000
746
+ WARMUP_METHOD: linear
747
+ WEIGHT_DECAY: 0.05
748
+ WEIGHT_DECAY_BIAS: null
749
+ WEIGHT_DECAY_EMBED: 0.0
750
+ WEIGHT_DECAY_NORM: 0.0
751
+ TEST:
752
+ AUG:
753
+ ENABLED: false
754
+ FLIP: true
755
+ MAX_SIZE: 4000
756
+ MIN_SIZES:
757
+ - 400
758
+ - 500
759
+ - 600
760
+ - 700
761
+ - 800
762
+ - 900
763
+ - 1000
764
+ - 1100
765
+ - 1200
766
+ DETECTIONS_PER_IMAGE: 100
767
+ EVAL_PERIOD: 0
768
+ EXPECTED_RESULTS: []
769
+ KEYPOINT_OKS_SIGMAS: []
770
+ PRECISE_BN:
771
+ ENABLED: false
772
+ NUM_ITER: 200
773
+ VERSION: 2
774
+ VIS_PERIOD: 0
775
+ enable_evaluation: false
776
+ evaluate_at_iter0: false
777
+ evaluate_opt_step: 0
778
+ evaluation_batch_size: 4
779
+ evaluation_seed: 0
780
+ evaluation_train_clips: 10
781
+ evaluation_val_clips: 10