-
Notifications
You must be signed in to change notification settings - Fork 8
/
Copy pathmain.py
165 lines (147 loc) · 7.23 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
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
import argparse
import torch
import numpy as np
import torch.nn as nn
import torch.optim as optim
from torch.optim import lr_scheduler
from torch.utils.data import DataLoader
from os import makedirs
from os.path import join, exists
from IPython.core.debugger import Pdb
# from preprocess import preprocess
from dataset import ReviewsDataset
from train import train_model, test_model
from model import HAN
from utils import log
parser = argparse.ArgumentParser()
parser.add_argument('--config', type=str, default='config.yaml')
parser.add_argument('--testfile', type=str, metavar='PATH')
parser.add_argument('--outputfile', type=str, metavar='PATH')
def load_datasets(config, phases, logfile=None):
config = config['data']
log('Loading vocabularies...', logfile)
import pickle
review_vocab = pickle.load(open(join(config['dir'], config['review_vocab']), 'rb'))
if config['review_vocab'] != config['summary_vocab']:
summary_vocab = pickle.load(open(join(config['dir'], config['summary_vocab']), 'rb'))
else:
summary_vocab = review_vocab
log('Loading preprocessed datasets...', logfile)
datafiles = {x: config[x]['jsonfile'] for x in phases}
datasets = {x: ReviewsDataset(data=join(config['dir'], datafiles[x]), review_vocab=review_vocab, summary_vocab=summary_vocab)
for x in phases}
def collate_fn(batch):
reviews = [sample[0] for sample in batch]
summaries = [sample[1] for sample in batch]
targets = torch.LongTensor([sample[2] for sample in batch])
return (reviews, summaries, targets)
if 'weights' not in config or not config['weights']:
dataloaders = {x: DataLoader(datasets[x], batch_size=config[x]['batch_size'], shuffle=True if x == 'train' else False, collate_fn=collate_fn) for x in phases}
else:
if config['weights'] == 'weighted':
samplers = {x: datasets[x].get_sampler() if x == 'train' else None for x in phases}
else:
samplers = {x: datasets[x].get_sampler(np.array(config['weights'])) if x == 'train' else None for x in phases}
dataloaders = {x: DataLoader(datasets[x], batch_size=config[x]['batch_size'], shuffle=False, sampler=samplers[x], collate_fn=collate_fn) for x in phases}
dataset_sizes = {x: len(datasets[x]) for x in phases}
log(dataset_sizes, logfile)
log("review vocab size: {}".format(len(review_vocab.itos)), logfile)
log("summary vocab size: {}".format(len(summary_vocab.itos)), logfile)
return dataloaders, review_vocab, summary_vocab
def build_model(config, review_vocab, summary_vocab, logfile=None):
use_gpu = config['use_gpu']
# Create Model
config['model']['params']['review_vocab_size'] = len(review_vocab)
config['model']['params']['summary_vocab_size'] = len(summary_vocab)
config['model']['params']['use_gpu'] = use_gpu
config = config['model']
model = HAN(**config['params'])
log(model, logfile)
# Copy pretrained word embeddings
model.review_lookup.weight.data.copy_(review_vocab.vectors)
if 'combined_lookup' not in config['params']:
config['params']['combined_lookup'] = False
if config['params']['use_summary'] and not config['params']['combined_lookup']:
model.summary_lookup.weight.data.copy_(summary_vocab.vectors)
if use_gpu:
model = model.cuda()
return model
def reload(config, model, optimizer=None, logfile=None):
save_dir = config['save_dir']
config = config['model']
best_fscore = 0
start_epoch = 0
if 'reload' in config:
reload_path = join(save_dir, config['reload'])
if exists(reload_path):
log("=> loading checkpoint/model found at '{0}'".format(reload_path), logfile)
checkpoint = torch.load(reload_path)
if model.__version__() != checkpoint['model_version']:
log('Model version mismatch: current version={}, checkpoint version={}'
.format(model.__version__(), checkpoint['model_version']))
start_epoch = checkpoint['epoch']
best_fscore = checkpoint['fscore']
model.load_state_dict(checkpoint['state_dict'])
if optimizer is not None:
optimizer.load_state_dict(checkpoint['optimizer'])
else:
log("no checkpoint/model found at '{0}'".format(reload_path), logfile)
return model, optimizer, best_fscore, start_epoch
def main(config):
logfile = join(config['save_dir'], 'log')
log(config, logfile)
if config['mode'] == 'test':
phases = ['test']
else:
phases = ['train', 'val']
dataloaders, review_vocab, summary_vocab = load_datasets(config, phases, logfile)
# Create Model
model = build_model(config, review_vocab, summary_vocab, logfile)
if config['mode'] == 'train':
# Select Optimizer
if config['optim']['class'] == 'sgd':
optimizer = optim.SGD(filter(lambda p: p.requires_grad, model.parameters()),
**config['optim']['params'])
elif config['optim']['class'] == 'rmsprop':
optimizer = optim.RMSprop(filter(lambda p: p.requires_grad, model.parameters()),
**config['optim']['params'])
else:
optimizer = optim.Adam(filter(lambda p: p.requires_grad, model.parameters()),
**config['optim']['params'])
# Reload model from checkpoint if provided
model, optimizer, best_fscore, start_epoch = reload(config, model, optimizer, logfile)
log(optimizer, logfile)
criterion = nn.CrossEntropyLoss()
patience = config['optim']['scheduler']['patience']
factor = config['optim']['scheduler']['factor']
scheduler = lr_scheduler.ReduceLROnPlateau(optimizer, mode='min', patience=patience, factor=factor,
threshold=0.05, threshold_mode='rel', verbose=True)
log(scheduler, logfile)
log("Begin Training...", logfile)
model = train_model(model, dataloaders, criterion, optimizer, scheduler, config['save_dir'],
num_epochs=config['training']['n_epochs'], use_gpu=config['use_gpu'],
best_fscore=best_fscore, start_epoch=start_epoch, logfile=logfile)
elif config['mode'] == 'test':
# Reload model from checkpoint if provided
model, _, _, _ = reload(config, model, logfile=logfile)
log('Testing on {}...'.format(config['data']['test']['jsonfile']))
test_model(model, dataloaders['test'], config['outputfile'], use_gpu=config['use_gpu'], logfile=logfile)
else:
log("Invalid config mode %s !!" % config['mode'], logfile)
if __name__ == '__main__':
global args
args = parser.parse_args()
import yaml
config = yaml.load(open(args.config))
config['use_gpu'] = config['use_gpu'] and torch.cuda.is_available()
# TODO: seeding still not perfect
torch.manual_seed(config['seed'])
torch.cuda.manual_seed(config['seed'])
if args.testfile:
config['data']['test']['jsonfile'] = args.testfile
config['outputfile'] = args.outputfile
config['data']['dir'] = ''
config['save_dir'] = ''
else:
makedirs(config['save_dir'], exist_ok=True)
main(config)