zx1239856 commited on
Commit
d8f2631
·
verified ·
1 Parent(s): d2ab24e

Upload checkpoint/config.yaml with huggingface_hub

Browse files
Files changed (1) hide show
  1. checkpoint/config.yaml +386 -0
checkpoint/config.yaml ADDED
@@ -0,0 +1,386 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ CUDNN_BENCHMARK: false
2
+ DATALOADER:
3
+ ASPECT_RATIO_GROUPING: true
4
+ FILTER_EMPTY_ANNOTATIONS: true
5
+ NUM_WORKERS: 8
6
+ REPEAT_SQRT: true
7
+ REPEAT_THRESHOLD: 0.0
8
+ SAMPLER_TRAIN: TrainingSampler
9
+ DATASETS:
10
+ PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
11
+ PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
12
+ PROPOSAL_FILES_TEST: []
13
+ PROPOSAL_FILES_TRAIN: []
14
+ TEST:
15
+ - front3d_pifu_test
16
+ TRAIN:
17
+ - front3d_pifu_train
18
+ GLOBAL:
19
+ EMPTY_CACHE_BEFORE_STEP: false
20
+ HACK: 1.0
21
+ INPUT:
22
+ COLOR_AUG_SSD: true
23
+ CROP:
24
+ ENABLED: false
25
+ SINGLE_CATEGORY_MAX_AREA: 1.0
26
+ SIZE:
27
+ - 484
28
+ - 484
29
+ TYPE: absolute
30
+ DATASET_MAPPER_NAME: front3d_pifu
31
+ FORMAT: RGB
32
+ MASK_FORMAT: polygon
33
+ MAX_SIZE_TEST: 648
34
+ MAX_SIZE_TRAIN: 648
35
+ MIN_SIZE_TEST: 484
36
+ MIN_SIZE_TRAIN:
37
+ - 484
38
+ MIN_SIZE_TRAIN_SAMPLING: choice
39
+ PSEUDO_DEPTH:
40
+ DEPTH_PATH: ""
41
+ ENABLED: false
42
+ RANDOM_DEPTH_NOISE: true
43
+ RANDOM_FLIP: none
44
+ RANDOM_INST_MASK:
45
+ ENABLED: true
46
+ MASK_PATH: datasets/front3d_pifu/data/instance_mask
47
+ PROB: 0.5
48
+ SIZE_DIVISIBILITY: -1
49
+ MODEL:
50
+ ANCHOR_GENERATOR:
51
+ ANGLES:
52
+ - - -90
53
+ - 0
54
+ - 90
55
+ ASPECT_RATIOS:
56
+ - - 0.5
57
+ - 1.0
58
+ - 2.0
59
+ NAME: DefaultAnchorGenerator
60
+ OFFSET: 0.0
61
+ SIZES:
62
+ - - 32
63
+ - 64
64
+ - 128
65
+ - 256
66
+ - 512
67
+ BACKBONE:
68
+ FREEZE_AT: 0
69
+ NAME: build_dinov2_backbone
70
+ DEVICE: cuda
71
+ DINO:
72
+ MODEL_NAME: dinov2_vits14_reg
73
+ REPO_NAME: facebookresearch/dinov2
74
+ FPN:
75
+ FUSE_TYPE: sum
76
+ IN_FEATURES: []
77
+ NORM: ''
78
+ OUT_CHANNELS: 256
79
+ KEYPOINT_ON: false
80
+ LOAD_PROPOSALS: false
81
+ MASK_ON: false
82
+ META_ARCHITECTURE: DepR
83
+ MODEL_2D:
84
+ FEATURE_DIM: 384
85
+ MIN_INSTANCE_PIXELS: 200
86
+ NO_FPN: true
87
+ SIZE_DIVISIBILITY: 14
88
+ TRAIN_NUM_SAMPLES: 80000
89
+ PANOPTIC_FPN:
90
+ COMBINE:
91
+ ENABLED: true
92
+ INSTANCES_CONFIDENCE_THRESH: 0.5
93
+ OVERLAP_THRESH: 0.5
94
+ STUFF_AREA_LIMIT: 4096
95
+ INSTANCE_LOSS_WEIGHT: 1.0
96
+ PIXEL_MEAN:
97
+ - 123.675
98
+ - 116.28
99
+ - 103.53
100
+ PIXEL_STD:
101
+ - 58.395
102
+ - 57.12
103
+ - 57.375
104
+ PROJECTION: HighResCubeForwardProjection
105
+ PROPOSAL_GENERATOR:
106
+ MIN_SIZE: 0
107
+ NAME: RPN
108
+ RESNETS:
109
+ DEFORM_MODULATED: false
110
+ DEFORM_NUM_GROUPS: 1
111
+ DEFORM_ON_PER_STAGE:
112
+ - false
113
+ - false
114
+ - false
115
+ - false
116
+ DEPTH: 50
117
+ NORM: SyncBN
118
+ NUM_GROUPS: 1
119
+ OUT_FEATURES:
120
+ - res2
121
+ - res3
122
+ - res4
123
+ - res5
124
+ RES2_OUT_CHANNELS: 256
125
+ RES4_DILATION: 1
126
+ RES5_DILATION: 1
127
+ RES5_MULTI_GRID:
128
+ - 1
129
+ - 1
130
+ - 1
131
+ STEM_OUT_CHANNELS: 64
132
+ STEM_TYPE: basic
133
+ STRIDE_IN_1X1: false
134
+ WIDTH_PER_GROUP: 64
135
+ RETINANET:
136
+ BBOX_REG_LOSS_TYPE: smooth_l1
137
+ BBOX_REG_WEIGHTS: &id002
138
+ - 1.0
139
+ - 1.0
140
+ - 1.0
141
+ - 1.0
142
+ FOCAL_LOSS_ALPHA: 0.25
143
+ FOCAL_LOSS_GAMMA: 2.0
144
+ IN_FEATURES:
145
+ - p3
146
+ - p4
147
+ - p5
148
+ - p6
149
+ - p7
150
+ IOU_LABELS:
151
+ - 0
152
+ - -1
153
+ - 1
154
+ IOU_THRESHOLDS:
155
+ - 0.4
156
+ - 0.5
157
+ NMS_THRESH_TEST: 0.5
158
+ NORM: ''
159
+ NUM_CLASSES: 80
160
+ NUM_CONVS: 4
161
+ PRIOR_PROB: 0.01
162
+ SCORE_THRESH_TEST: 0.05
163
+ SMOOTH_L1_LOSS_BETA: 0.1
164
+ TOPK_CANDIDATES_TEST: 1000
165
+ ROI_BOX_CASCADE_HEAD:
166
+ BBOX_REG_WEIGHTS:
167
+ - &id001
168
+ - 10.0
169
+ - 10.0
170
+ - 5.0
171
+ - 5.0
172
+ - - 20.0
173
+ - 20.0
174
+ - 10.0
175
+ - 10.0
176
+ - - 30.0
177
+ - 30.0
178
+ - 15.0
179
+ - 15.0
180
+ IOUS:
181
+ - 0.5
182
+ - 0.6
183
+ - 0.7
184
+ ROI_BOX_HEAD:
185
+ BBOX_REG_LOSS_TYPE: smooth_l1
186
+ BBOX_REG_LOSS_WEIGHT: 1.0
187
+ BBOX_REG_WEIGHTS: *id001
188
+ CLS_AGNOSTIC_BBOX_REG: false
189
+ CONV_DIM: 256
190
+ FC_DIM: 1024
191
+ FED_LOSS_FREQ_WEIGHT_POWER: 0.5
192
+ FED_LOSS_NUM_CLASSES: 50
193
+ NAME: ''
194
+ NORM: ''
195
+ NUM_CONV: 0
196
+ NUM_FC: 0
197
+ POOLER_RESOLUTION: 14
198
+ POOLER_SAMPLING_RATIO: 0
199
+ POOLER_TYPE: ROIAlignV2
200
+ SMOOTH_L1_BETA: 0.0
201
+ TRAIN_ON_PRED_BOXES: false
202
+ USE_FED_LOSS: false
203
+ USE_SIGMOID_CE: false
204
+ ROI_HEADS:
205
+ BATCH_SIZE_PER_IMAGE: 512
206
+ IN_FEATURES:
207
+ - res4
208
+ IOU_LABELS:
209
+ - 0
210
+ - 1
211
+ IOU_THRESHOLDS:
212
+ - 0.5
213
+ NAME: Res5ROIHeads
214
+ NMS_THRESH_TEST: 0.5
215
+ NUM_CLASSES: 80
216
+ POSITIVE_FRACTION: 0.25
217
+ PROPOSAL_APPEND_GT: true
218
+ SCORE_THRESH_TEST: 0.05
219
+ ROI_KEYPOINT_HEAD:
220
+ CONV_DIMS:
221
+ - 512
222
+ - 512
223
+ - 512
224
+ - 512
225
+ - 512
226
+ - 512
227
+ - 512
228
+ - 512
229
+ LOSS_WEIGHT: 1.0
230
+ MIN_KEYPOINTS_PER_IMAGE: 1
231
+ NAME: KRCNNConvDeconvUpsampleHead
232
+ NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: true
233
+ NUM_KEYPOINTS: 17
234
+ POOLER_RESOLUTION: 14
235
+ POOLER_SAMPLING_RATIO: 0
236
+ POOLER_TYPE: ROIAlignV2
237
+ ROI_MASK_HEAD:
238
+ CLS_AGNOSTIC_MASK: false
239
+ CONV_DIM: 256
240
+ NAME: MaskRCNNConvUpsampleHead
241
+ NORM: ''
242
+ NUM_CONV: 0
243
+ POOLER_RESOLUTION: 14
244
+ POOLER_SAMPLING_RATIO: 0
245
+ POOLER_TYPE: ROIAlignV2
246
+ RPN:
247
+ BATCH_SIZE_PER_IMAGE: 256
248
+ BBOX_REG_LOSS_TYPE: smooth_l1
249
+ BBOX_REG_LOSS_WEIGHT: 1.0
250
+ BBOX_REG_WEIGHTS: *id002
251
+ BOUNDARY_THRESH: -1
252
+ CONV_DIMS:
253
+ - -1
254
+ HEAD_NAME: StandardRPNHead
255
+ IN_FEATURES:
256
+ - res4
257
+ IOU_LABELS:
258
+ - 0
259
+ - -1
260
+ - 1
261
+ IOU_THRESHOLDS:
262
+ - 0.3
263
+ - 0.7
264
+ LOSS_WEIGHT: 1.0
265
+ NMS_THRESH: 0.7
266
+ POSITIVE_FRACTION: 0.5
267
+ POST_NMS_TOPK_TEST: 1000
268
+ POST_NMS_TOPK_TRAIN: 2000
269
+ PRE_NMS_TOPK_TEST: 6000
270
+ PRE_NMS_TOPK_TRAIN: 12000
271
+ SMOOTH_L1_BETA: 0.0
272
+ SEM_SEG_HEAD:
273
+ ASPP_CHANNELS: 256
274
+ ASPP_DILATIONS:
275
+ - 6
276
+ - 12
277
+ - 18
278
+ ASPP_DROPOUT: 0.1
279
+ COMMON_STRIDE: 4
280
+ CONVS_DIM: 128
281
+ IGNORE_VALUE: 255
282
+ IN_FEATURES:
283
+ - p2
284
+ - p3
285
+ - p4
286
+ - p5
287
+ LOSS_TYPE: hard_pixel_mining
288
+ LOSS_WEIGHT: 1.0
289
+ NAME: SemSegFPNHead
290
+ NORM: GN
291
+ NUM_CLASSES: 54
292
+ PROJECT_CHANNELS:
293
+ - 48
294
+ PROJECT_FEATURES:
295
+ - res2
296
+ USE_DEPTHWISE_SEPARABLE_CONV: false
297
+ TRIPLANE_DIFFUSION:
298
+ CFG_PROB: 0.0
299
+ EIK_WEIGHT: 0.5
300
+ EMBED_MEAN: 0.9908467
301
+ EMBED_STD: 1.4378415
302
+ ENABLE_ATTN: true
303
+ MSE_WEIGHT: 0.5
304
+ NOR_WEIGHT: 5.0
305
+ PRETRAINED_PATH: configs/unet_c10/unet/config.json
306
+ SCHEDULER_CONFIG: configs/unet_c10/scheduler/scheduler_config.json
307
+ SDF_WEIGHT: 5.0
308
+ SUR_WEIGHT: 10.0
309
+ TRIPLANE_VAE:
310
+ DIMENSIONS: 32
311
+ EXTRUDE_DEPTH: 0.0
312
+ KL_STD: 0.25
313
+ MLP_PATH: checkpoint/mlp.pth
314
+ NUM_HEADS: 16
315
+ PLANE_SHAPE:
316
+ - 3
317
+ - 32
318
+ - 128
319
+ - 128
320
+ PRETRAINED_PATH: checkpoint/vae_decoder.safetensors
321
+ STAT_PATH: checkpoint/vae_stat.pkl
322
+ TRANSFORM_DEPTH: 1
323
+ Z_SHAPE:
324
+ - 2
325
+ - 32
326
+ - 32
327
+ WEIGHTS: ""
328
+ OUTPUT_DIR: ./output/3dproj_attn_dino_c9_augdep_augmask_nocfg
329
+ SEED: -1
330
+ SOLVER:
331
+ AMP:
332
+ ENABLED: false
333
+ BACKBONE_MULTIPLIER: 0.1
334
+ BASE_LR: 0.0001
335
+ BASE_LR_END: 0.0
336
+ BIAS_LR_FACTOR: 1.0
337
+ CHECKPOINT_PERIOD: 5000
338
+ CLIP_GRADIENTS:
339
+ CLIP_TYPE: full_model
340
+ CLIP_VALUE: 0.1
341
+ ENABLED: true
342
+ NORM_TYPE: 2.0
343
+ GAMMA: 0.1
344
+ IMS_PER_BATCH: 12
345
+ LR_SCHEDULER_NAME: WarmupMultiStepLR
346
+ MAX_ITER: 220000
347
+ MOMENTUM: 0.9
348
+ NESTEROV: false
349
+ NUM_DECAYS: 3
350
+ OPTIMIZER: ADAMW
351
+ POLY_LR_CONSTANT_ENDING: 0.0
352
+ POLY_LR_POWER: 0.9
353
+ REFERENCE_WORLD_SIZE: 0
354
+ RESCALE_INTERVAL: false
355
+ STEPS: []
356
+ WARMUP_FACTOR: 1.0
357
+ WARMUP_ITERS: 0
358
+ WARMUP_METHOD: linear
359
+ WEIGHT_DECAY: 0.05
360
+ WEIGHT_DECAY_BIAS: null
361
+ WEIGHT_DECAY_EMBED: 0.0
362
+ WEIGHT_DECAY_NORM: 0.0
363
+ TEST:
364
+ AUG:
365
+ ENABLED: false
366
+ FLIP: true
367
+ MAX_SIZE: 4000
368
+ MIN_SIZES:
369
+ - 400
370
+ - 500
371
+ - 600
372
+ - 700
373
+ - 800
374
+ - 900
375
+ - 1000
376
+ - 1100
377
+ - 1200
378
+ DETECTIONS_PER_IMAGE: 100
379
+ EVAL_PERIOD: 0
380
+ EXPECTED_RESULTS: []
381
+ KEYPOINT_OKS_SIGMAS: []
382
+ PRECISE_BN:
383
+ ENABLED: false
384
+ NUM_ITER: 200
385
+ VERSION: 2
386
+ VIS_PERIOD: 0