-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathmain.py
140 lines (105 loc) · 4.81 KB
/
main.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
import os
import subprocess
os.environ['PYTHONIOENCODING'] = 'utf-8'
os.environ['CUDA_VISIBLE_DEVICES'] = '0'
import random
import numpy as np
import torch
import torch.nn as nn
import torch.optim as optim
seed = 42
random.seed(seed)
np.random.seed(seed)
torch.manual_seed(seed)
torch.cuda.manual_seed(seed)
torch.backends.cudnn.deterministic = True
from utils.parser import get_parser
from utils.logger import get_logger
parser = get_parser()
option = parser.parse_args()
root_path = 'result'
logs_folder = os.path.join(root_path, 'logs', option.name)
save_folder = os.path.join(root_path, 'save', option.name)
sample_folder = os.path.join(root_path, 'sample', option.name)
result_folder = os.path.join(root_path, 'result', option.name)
subprocess.run('mkdir -p %s' % logs_folder, shell = True)
subprocess.run('mkdir -p %s' % save_folder, shell = True)
subprocess.run('mkdir -p %s' % sample_folder, shell = True)
subprocess.run('mkdir -p %s' % result_folder, shell = True)
logger = get_logger(option.name, os.path.join(logs_folder, 'main.log'))
from loaders.loader1 import get_loader as get_loader1
from modules.module1 import get_module as get_module1
from utils.misc import train, valid, test, save_checkpoint, load_checkpoint, save_sample, calc_matrix
device = torch.device('cuda') if torch.cuda.is_available() else torch.device('cpu')
logger.info('prepare loader')
vocab, train_loader, valid_loader, test_loader = get_loader1(option)
logger.info('prepare module')
seq2seq = get_module1(option, vocab.size)
seq2seq = seq2seq.to (device)
logger.info('prepare envs')
params_list = list(seq2seq.parameters())
ada_init_lr = option.lr_coverage if option.is_coverage else option.lr
ada_init_ac = option.ada_init_ac
optimizer = optim.Adagrad(params_list, lr = ada_init_lr, initial_accumulator_value = ada_init_ac)
print_interval = 100
check_interval = 5000
assert option.mode in ['train', 'valid', 'test']
if option.mode == 'train':
logger.info('start training!')
if option.ckpt != '' and os.path.isfile(option.ckpt):
start_iter = load_checkpoint(option.ckpt, seq2seq, optimizer)
logger.info('training a old model: %s' % (option.ckpt))
else:
start_iter = 0
logger.info('training a new model')
for count_iter in range(start_iter + 1, option.total_iter):
batch = train_loader.get_batch()
if batch is None:
break
loss, cove_loss = train(batch, seq2seq, device, optimizer, option.is_copy, option.is_coverage, option.cov_loss_wt, option.max_dec_step, option.grad_clip)
if count_iter % print_interval == 0:
logger.info('iter: %d, loss: %f, cove_loss: %f' % (count_iter, loss, cove_loss))
if count_iter % check_interval == 0:
save_checkpoint(os.path.join(save_folder, '%s.ckpt' % str(count_iter)), seq2seq, optimizer, count_iter)
if option.mode == 'valid':
logger.info('start validing!')
if option.ckpt != '' and os.path.isfile(option.ckpt):
final_iter = load_checkpoint(option.ckpt, seq2seq, optimizer)
logger.info('validing a old model: %s' % (option.ckpt))
else:
logger.info('validing a new model, unexpected')
raise Exception('Expect to use a pre-existing model')
count_iter = 0
while True:
count_iter += 1
batch = valid_loader.get_batch()
if batch is None:
break
loss, cove_loss = valid(batch, seq2seq, device, option.is_copy, option.is_coverage, option.cov_loss_wt, option.max_dec_step)
if count_iter % print_interval == 0:
logger.info('iter: %d, loss: %f, cove_loss: %f' % (count_iter, loss, cove_loss))
if option.mode == 'test':
logger.info('start testing!')
if option.ckpt != '' and os.path.isfile(option.ckpt):
final_iter = load_checkpoint(option.ckpt, seq2seq, optimizer)
logger.info('testing a old model: %s' % (option.ckpt))
else:
logger.info('testing a new model, unexpected')
raise Exception('Expect to use a pre-existing model')
sources = []
targets = []
predict = []
count_iter = 0
while True:
count_iter += 1
batch = test_loader.get_batch()
if batch is None:
break
src, trg, pred = test(batch, seq2seq, device, option.is_copy, option.is_coverage, vocab, option.min_dec_step, option.max_dec_step, option.beam_width)
sources.append(src)
targets.append(trg)
predict.append(pred)
if count_iter % print_interval == 0:
logger.info('iter: %d\nsrc: %s\ntrg: %s\npred: %s' % (count_iter, src, trg, pred))
save_sample(result_folder, final_iter, sources, targets, predict)
calc_matrix(result_folder, final_iter, sources, targets, predict)