Hugging Face's logo

ilyassmoummad
/
ProtoCLR

Feature Extraction
PyTorch
Bioacoustics
Model card Files Community
ProtoCLR
File size: 4,844 Bytes 
803c68c
 
 
 
 
 
 
 
bedbb17
 
803c68c
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
 
from __future__ import absolute_import
from __future__ import division
from __future__ import print_function

import os.path as op
import yaml
from yacs.config import CfgNode as CN

#from config import comm
import comm

_C = CN()

_C.BASE = ['']
_C.NAME = ''
_C.DATA_DIR = ''
_C.DIST_BACKEND = 'nccl'
_C.GPUS = (0,)
# _C.LOG_DIR = ''
_C.MULTIPROCESSING_DISTRIBUTED = True
_C.OUTPUT_DIR = ''
_C.PIN_MEMORY = True
_C.PRINT_FREQ = 20
_C.RANK = 0
_C.VERBOSE = True
_C.WORKERS = 4
_C.MODEL_SUMMARY = False

_C.AMP = CN()
_C.AMP.ENABLED = False
_C.AMP.MEMORY_FORMAT = 'nchw'

# Cudnn related params
_C.CUDNN = CN()
_C.CUDNN.BENCHMARK = True
_C.CUDNN.DETERMINISTIC = False
_C.CUDNN.ENABLED = True

# common params for NETWORK
_C.MODEL = CN()
_C.MODEL.NAME = 'cls_hrnet'
_C.MODEL.INIT_WEIGHTS = True
_C.MODEL.PRETRAINED = ''
_C.MODEL.PRETRAINED_LAYERS = ['*']
_C.MODEL.NUM_CLASSES = 1000
_C.MODEL.SPEC = CN(new_allowed=True)

_C.LOSS = CN(new_allowed=True)
_C.LOSS.LABEL_SMOOTHING = 0.0
_C.LOSS.LOSS = 'softmax'

# DATASET related params
_C.DATASET = CN()
_C.DATASET.ROOT = ''
_C.DATASET.DATASET = 'imagenet'
_C.DATASET.TRAIN_SET = 'train'
_C.DATASET.TEST_SET = 'val'
_C.DATASET.DATA_FORMAT = 'jpg'
_C.DATASET.LABELMAP = ''
_C.DATASET.TRAIN_TSV_LIST = []
_C.DATASET.TEST_TSV_LIST = []
_C.DATASET.SAMPLER = 'default'

_C.DATASET.TARGET_SIZE = -1

# training data augmentation
_C.INPUT = CN()
_C.INPUT.MEAN = [0.485, 0.456, 0.406]
_C.INPUT.STD = [0.229, 0.224, 0.225]

# data augmentation
_C.AUG = CN()
_C.AUG.SCALE = (0.08, 1.0)
_C.AUG.RATIO = (3.0/4.0, 4.0/3.0)
_C.AUG.COLOR_JITTER = [0.4, 0.4, 0.4, 0.1, 0.0]
_C.AUG.GRAY_SCALE = 0.0
_C.AUG.GAUSSIAN_BLUR = 0.0
_C.AUG.DROPBLOCK_LAYERS = [3, 4]
_C.AUG.DROPBLOCK_KEEP_PROB = 1.0
_C.AUG.DROPBLOCK_BLOCK_SIZE = 7
_C.AUG.MIXUP_PROB = 0.0
_C.AUG.MIXUP = 0.0
_C.AUG.MIXCUT = 0.0
_C.AUG.MIXCUT_MINMAX = []
_C.AUG.MIXUP_SWITCH_PROB = 0.5
_C.AUG.MIXUP_MODE = 'batch'
_C.AUG.MIXCUT_AND_MIXUP = False
_C.AUG.INTERPOLATION = 2
_C.AUG.TIMM_AUG = CN(new_allowed=True)
_C.AUG.TIMM_AUG.USE_LOADER = False
_C.AUG.TIMM_AUG.USE_TRANSFORM = False

# train
_C.TRAIN = CN()

_C.TRAIN.AUTO_RESUME = True
_C.TRAIN.CHECKPOINT = ''
_C.TRAIN.LR_SCHEDULER = CN(new_allowed=True)
_C.TRAIN.SCALE_LR = True
_C.TRAIN.LR = 0.001

_C.TRAIN.OPTIMIZER = 'sgd'
_C.TRAIN.OPTIMIZER_ARGS = CN(new_allowed=True)
_C.TRAIN.MOMENTUM = 0.9
_C.TRAIN.WD = 0.0001
_C.TRAIN.WITHOUT_WD_LIST = []
_C.TRAIN.NESTEROV = True
# for adam
_C.TRAIN.GAMMA1 = 0.99
_C.TRAIN.GAMMA2 = 0.0

_C.TRAIN.BEGIN_EPOCH = 0
_C.TRAIN.END_EPOCH = 100

_C.TRAIN.IMAGE_SIZE = [224, 224]  # width * height, ex: 192 * 256
_C.TRAIN.BATCH_SIZE_PER_GPU = 32
_C.TRAIN.SHUFFLE = True

_C.TRAIN.EVAL_BEGIN_EPOCH = 0

_C.TRAIN.DETECT_ANOMALY = False

_C.TRAIN.CLIP_GRAD_NORM = 0.0
_C.TRAIN.SAVE_ALL_MODELS = False

# testing
_C.TEST = CN()

# size of images for each device
_C.TEST.BATCH_SIZE_PER_GPU = 32
_C.TEST.CENTER_CROP = True
_C.TEST.IMAGE_SIZE = [224, 224]  # width * height, ex: 192 * 256
_C.TEST.INTERPOLATION = 2
_C.TEST.MODEL_FILE = ''
_C.TEST.REAL_LABELS = False
_C.TEST.VALID_LABELS = ''

_C.FINETUNE = CN()
_C.FINETUNE.FINETUNE = False
_C.FINETUNE.USE_TRAIN_AUG = False
_C.FINETUNE.BASE_LR = 0.003
_C.FINETUNE.BATCH_SIZE = 512
_C.FINETUNE.EVAL_EVERY = 3000
_C.FINETUNE.TRAIN_MODE = True
# _C.FINETUNE.MODEL_FILE = ''
_C.FINETUNE.FROZEN_LAYERS = []
_C.FINETUNE.LR_SCHEDULER = CN(new_allowed=True)
_C.FINETUNE.LR_SCHEDULER.DECAY_TYPE = 'step'

# debug
_C.DEBUG = CN()
_C.DEBUG.DEBUG = False


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, op.join(op.dirname(cfg_file), cfg)
            )
    print('=> merge config from {}'.format(cfg_file))
    config.merge_from_file(cfg_file)
    config.freeze()


def update_config(config, args):
    _update_config_from_file(config, args.cfg)

    config.defrost()
    config.merge_from_list(args.opts)
    if config.TRAIN.SCALE_LR:
        config.TRAIN.LR *= comm.world_size
    file_name, _ = op.splitext(op.basename(args.cfg))
    config.NAME = file_name + config.NAME
    config.RANK = comm.rank

    if 'timm' == config.TRAIN.LR_SCHEDULER.METHOD:
        config.TRAIN.LR_SCHEDULER.ARGS.epochs = config.TRAIN.END_EPOCH

    if 'timm' == config.TRAIN.OPTIMIZER:
        config.TRAIN.OPTIMIZER_ARGS.lr = config.TRAIN.LR

    aug = config.AUG
    if aug.MIXUP > 0.0 or aug.MIXCUT > 0.0 or aug.MIXCUT_MINMAX:
        aug.MIXUP_PROB = 1.0
    config.freeze()


def save_config(cfg, path):
    if comm.is_main_process():
        with open(path, 'w') as f:
            f.write(cfg.dump())


if __name__ == '__main__':
    import sys
    with open(sys.argv[1], 'w') as f:
        print(_C, file=f)