-
Notifications
You must be signed in to change notification settings - Fork 86
/
train.py
72 lines (70 loc) · 3.78 KB
/
train.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
import argparse
from parse_config import cfg, cfg_from_file, assert_and_infer_cfg
from utils.util import fix_seed, load_specific_dict
from models.loss import SupConLoss, get_pen_loss
from models.model import SDT_Generator
from utils.logger import set_log
from data_loader.loader import ScriptDataset
import torch
from trainer.trainer import Trainer
def main(opt):
""" load config file into cfg"""
cfg_from_file(opt.cfg_file)
assert_and_infer_cfg()
"""fix the random seed"""
fix_seed(cfg.TRAIN.SEED)
""" prepare log file """
logs = set_log(cfg.OUTPUT_DIR, opt.cfg_file, opt.log_name)
""" set dataset"""
train_dataset = ScriptDataset(
cfg.DATA_LOADER.PATH, cfg.DATA_LOADER.DATASET, cfg.TRAIN.ISTRAIN, cfg.MODEL.NUM_IMGS)
print('number of training images: ', len(train_dataset))
train_loader = torch.utils.data.DataLoader(train_dataset,
batch_size=cfg.TRAIN.IMS_PER_BATCH,
shuffle=True,
drop_last=False,
collate_fn=train_dataset.collate_fn_,
num_workers=cfg.DATA_LOADER.NUM_THREADS)
test_dataset = ScriptDataset(
cfg.DATA_LOADER.PATH, cfg.DATA_LOADER.DATASET, cfg.TEST.ISTRAIN, cfg.MODEL.NUM_IMGS)
test_loader = torch.utils.data.DataLoader(test_dataset,
batch_size=cfg.TRAIN.IMS_PER_BATCH,
shuffle=True,
sampler=None,
drop_last=False,
collate_fn=test_dataset.collate_fn_,
num_workers=cfg.DATA_LOADER.NUM_THREADS)
char_dict = test_dataset.char_dict
""" build model, criterion and optimizer"""
model = SDT_Generator(num_encoder_layers=cfg.MODEL.ENCODER_LAYERS,
num_head_layers= cfg.MODEL.NUM_HEAD_LAYERS,
wri_dec_layers=cfg.MODEL.WRI_DEC_LAYERS,
gly_dec_layers= cfg.MODEL.GLY_DEC_LAYERS).to('cuda')
### load checkpoint
if len(opt.pretrained_model) > 0:
model.load_state_dict(torch.load(opt.pretrained_model))
print('load pretrained model from {}'.format(opt.pretrained_model))
elif len(opt.content_pretrained) > 0:
model_dict = load_specific_dict(model.content_encoder, opt.content_pretrained, "feature_ext")
model.content_encoder.load_state_dict(model_dict)
print('load content pretrained model from {}'.format(opt.content_pretrained))
else:
pass
criterion = dict(NCE=SupConLoss(contrast_mode='all'), PEN=get_pen_loss)
optimizer = torch.optim.Adam(model.parameters(), lr=cfg.SOLVER.BASE_LR)
"""start training iterations"""
trainer = Trainer(model, criterion, optimizer, train_loader, logs, char_dict, test_loader)
trainer.train()
if __name__ == '__main__':
"""Parse input arguments"""
parser = argparse.ArgumentParser()
parser.add_argument('--pretrained_model', default='',
dest='pretrained_model', required=False, help='continue to train model')
parser.add_argument('--content_pretrained', default='model_zoo/position_layer2_dim512_iter138k_test_acc0.9443.pth',
dest='content_pretrained', required=False, help='continue to train content encoder')
parser.add_argument('--cfg', dest='cfg_file', default='configs/CHINESE_CASIA.yml',
help='Config file for training (and optionally testing)')
parser.add_argument('--log', default='debug',
dest='log_name', required=False, help='the filename of log')
opt = parser.parse_args()
main(opt)