-
Notifications
You must be signed in to change notification settings - Fork 1
/
config.py
237 lines (199 loc) · 7.42 KB
/
config.py
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
import os
import yaml
from yacs.config import CfgNode as CN
_C = CN()
# -----------------------------------------------------------------------------
# Base settings
# -----------------------------------------------------------------------------
_C.BASE = ['']
_C.DIS = False
_C.WORLD_SIZE = 1
_C.SEED = 1234
_C.AMP = True
_C.EXPERIMENT_ID = ""
_C.SAVE_DIR = "save_pth"
_C.VAL_OUTPUT_PATH = "/home/lwt/code/flare/FLARE22-TwoStagePHTrans/save_results"
_C.COARSE_MODEL_PATH = "/home/lwt/code/flare/FLARE22-TwoStagePHTrans/save_pth/phtrans_c_220813_001000"
_C.FINE_MODEL_PATH = "/home/lwt/code/flare/FLARE22-TwoStagePHTrans/save_pth/phtrans_f_220813_001056"
# -----------------------------------------------------------------------------
# Wandb settings
# -----------------------------------------------------------------------------
_C.WANDB = CN()
_C.WANDB.COARSE_PROJECT = "FLARE2022_COARSE"
_C.WANDB.FINE_PROJECT = "FLARE2022_FINE"
_C.WANDB.TAG = "PHTrans"
_C.WANDB.MODE = "offline"
# -----------------------------------------------------------------------------
# Data settings
# -----------------------------------------------------------------------------
_C.DATASET = CN()
_C.DATASET.WITH_VAL = False
_C.DATASET.TRAIN_UNLABELED_IMAGE_PATH = "/home/lwt/data/flare22/UnlabeledCase"
_C.DATASET.TRAIN_UNLABELED_MASK_PATH = "/home/lwt/data/flare22/Unlabel2000_phtranPre"
_C.DATASET.TRAIN_IMAGE_PATH = "/home/lwt/data/flare22/Training/FLARE22_LabeledCase50/images"
_C.DATASET.TRAIN_MASK_PATH = "/home/lwt/data/flare22/Training/FLARE22_LabeledCase50/labels"
_C.DATASET.VAL_IMAGE_PATH = "/home/lwt/data/flare22/Validation"
_C.DATASET.EXTEND_SIZE = 20
_C.DATASET.IS_NORMALIZATION_DIRECTION = True
_C.DATASET.COARSE = CN()
_C.DATASET.COARSE.PROPRECESS_PATH = "/home/lwt/data_pro/flare22/Training/coarse_646464"
_C.DATASET.COARSE.PROPRECESS_UL_PATH = "/home/lwt/data_pro/flare22/Unlabel2000_coarse_646464"
_C.DATASET.COARSE.NUM_EACH_EPOCH = 512
_C.DATASET.COARSE.SIZE = [64, 64, 64]
_C.DATASET.COARSE.LABEL_CLASSES = 2
_C.DATASET.FINE = CN()
_C.DATASET.FINE.PROPRECESS_PATH = "/home/lwt/data_pro/flare22/Training/fine_96192192"
_C.DATASET.FINE.PROPRECESS_UL_PATH = "/home/lwt/data_pro/flare22/Unlabel2000_fine_96192192"
_C.DATASET.FINE.NUM_EACH_EPOCH = 512
_C.DATASET.FINE.SIZE = [96, 192, 192]
_C.DATASET.FINE.LABEL_CLASSES = 14
_C.DATASET.DA = CN()
_C.DATASET.DA.DO_2D_AUG = True
_C.DATASET.DA.DO_ELASTIC = True
_C.DATASET.DA.DO_SCALING = True
_C.DATASET.DA.DO_ROTATION = True
_C.DATASET.DA.RANDOM_CROP = False
_C.DATASET.DA.DO_GAMMA = True
_C.DATASET.DA.DO_MIRROR = False
_C.DATASET.DA.DO_ADDITIVE_BRIGHTNESS = True
# -----------------------------------------------------------------------------
# Dataloader settings
# -----------------------------------------------------------------------------
_C.DATALOADER = CN()
_C.DATALOADER.BATCH_SIZE = 1
_C.DATALOADER.PIN_MEMORY = True
_C.DATALOADER.NUM_WORKERS = 8
# -----------------------------------------------------------------------------
# Model settings
# -----------------------------------------------------------------------------
_C.MODEL = CN()
_C.MODEL.DEEP_SUPERVISION = True
_C.MODEL.COARSE = CN()
_C.MODEL.COARSE.TYPE = "phtrans"
_C.MODEL.COARSE.BASE_NUM_FEATURES = 16
_C.MODEL.COARSE.NUM_ONLY_CONV_STAGE = 2
_C.MODEL.COARSE.NUM_CONV_PER_STAGE = 2
_C.MODEL.COARSE.FEAT_MAP_MUL_ON_DOWNSCALE = 2
_C.MODEL.COARSE.POOL_OP_KERNEL_SIZES = [
[2, 2, 2], [2, 2, 2], [2, 2, 2], [2, 2, 2]]
_C.MODEL.COARSE.CONV_KERNEL_SIZES = [
[3, 3, 3], [3, 3, 3], [3, 3, 3], [3, 3, 3], [3, 3, 3]]
_C.MODEL.COARSE.DROPOUT_P = 0.1
_C.MODEL.COARSE.MAX_NUM_FEATURES = 200
_C.MODEL.COARSE.DEPTHS = [2, 2, 2, 2]
_C.MODEL.COARSE.NUM_HEADS = [4, 4, 4, 4]
_C.MODEL.COARSE.WINDOW_SIZE = [4, 4, 4]
_C.MODEL.COARSE.MLP_RATIO = 1.
_C.MODEL.COARSE.QKV_BIAS = True
_C.MODEL.COARSE.QK_SCALE = None
_C.MODEL.COARSE.DROP_RATE = 0.
_C.MODEL.COARSE.DROP_PATH_RATE = 0.1
_C.MODEL.FINE = CN()
_C.MODEL.FINE.TYPE = "phtrans"
_C.MODEL.FINE.BASE_NUM_FEATURES = 16
_C.MODEL.FINE.NUM_ONLY_CONV_STAGE = 2
_C.MODEL.FINE.NUM_CONV_PER_STAGE = 2
_C.MODEL.FINE.FEAT_MAP_MUL_ON_DOWNSCALE = 2
_C.MODEL.FINE.POOL_OP_KERNEL_SIZES = [
[2, 2, 2], [2, 2, 2], [2, 2, 2], [2, 2, 2]]
_C.MODEL.FINE.CONV_KERNEL_SIZES = [[3, 3, 3], [
3, 3, 3], [3, 3, 3], [3, 3, 3], [3, 3, 3]]
_C.MODEL.FINE.DROPOUT_P = 0.1
_C.MODEL.FINE.MAX_NUM_FEATURES = 200
_C.MODEL.FINE.DEPTHS = [2, 2, 2, 2]
_C.MODEL.FINE.NUM_HEADS = [4, 4, 4, 4]
_C.MODEL.FINE.WINDOW_SIZE = [3, 4, 4]
_C.MODEL.FINE.MLP_RATIO = 1.
_C.MODEL.FINE.QKV_BIAS = True
_C.MODEL.FINE.QK_SCALE = None
_C.MODEL.FINE.DROP_RATE = 0.
_C.MODEL.FINE.DROP_PATH_RATE = 0.1
# -----------------------------------------------------------------------------
# Training settings
# -----------------------------------------------------------------------------
_C.TRAIN = CN()
_C.TRAIN.DO_BACKPROP = True
_C.TRAIN.VAL_NUM_EPOCHS = 1
_C.TRAIN.SAVE_PERIOD = 1
_C.TRAIN.EPOCHS = 300
_C.TRAIN.WEIGHT_DECAY = 0.01
_C.TRAIN.WARMUP_EPOCHS = 20
_C.TRAIN.BASE_LR = 5e-4
_C.TRAIN.WARMUP_LR = 5e-7
_C.TRAIN.MIN_LR = 5e-6
# LR scheduler
_C.TRAIN.LR_SCHEDULER = CN()
_C.TRAIN.LR_SCHEDULER.NAME = 'cosine'
# Epoch interval to decay LR, used in StepLRScheduler
_C.TRAIN.LR_SCHEDULER.DECAY_EPOCHS = 30
# LR decay rate, used in StepLRScheduler
_C.TRAIN.LR_SCHEDULER.DECAY_RATE = 0.1
# Optimizer
_C.TRAIN.OPTIMIZER = CN()
_C.TRAIN.OPTIMIZER.NAME = 'adamw'
# Optimizer Epsilon
_C.TRAIN.OPTIMIZER.EPS = 1e-8
# Optimizer Betas
_C.TRAIN.OPTIMIZER.BETAS = (0.9, 0.999)
# SGD momentum
_C.TRAIN.OPTIMIZER.MOMENTUM = 0.9
# -----------------------------------------------------------------------------
# Test settings
# -----------------------------------------------------------------------------
_C.VAL = CN()
_C.VAL.IS_POST_PROCESS = True
_C.VAL.IS_WITH_DATALOADER = True
def _update_config_from_file(config, cfg_file):
config.defrost()
with open(cfg_file, 'r') as f:
yaml_cfg = yaml.load(f, Loader=yaml.FullLoader)
for cfg in yaml_cfg.setdefault('BASE', ['']):
if cfg:
_update_config_from_file(
config, os.path.join(os.path.dirname(cfg_file), cfg)
)
print('=> merge config from {}'.format(cfg_file))
config.merge_from_file(cfg_file)
config.freeze()
def update_config(config, args):
if args.cfg is not None:
_update_config_from_file(config, args.cfg)
config.defrost()
if args.opts:
config.merge_from_list(args.opts)
if args.batch_size:
config.DATALOADER.BATCH_SIZE = args.batch_size
if args.tag:
config.WANDB.TAG = args.tag
if args.wandb_mode == "online":
config.WANDB.MODE = args.wandb_mode
if args.world_size:
config.WORLD_SIZE = args.world_size
if args.with_distributed:
config.DIS = True
config.freeze()
def update_val_config(config, args):
if args.cfg is not None:
_update_config_from_file(config, args.cfg)
config.defrost()
if args.opts:
config.merge_from_list(args.opts)
# merge from specific arguments
if args.save_model_path:
config.SAVE_MODEL_PATH = args.save_model_path
if args.data_path:
config.DATASET.VAL_IMAGE_PATH = args.data_path
if args.output_path:
config.VAL_OUTPUT_PATH = args.output_path
config.freeze()
def get_config(args=None):
config = _C.clone()
update_config(config, args)
return config
def get_config_no_args():
config = _C.clone()
return config
def get_val_config(args=None):
config = _C.clone()
update_val_config(config, args)
return config