-
Notifications
You must be signed in to change notification settings - Fork 1
/
Copy pathtrainer.py
421 lines (347 loc) · 16.9 KB
/
trainer.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
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
# -*-coding:utf-8 -*-
# --------------------------------------------------------
# Pytorch W2VV++
# Written by Xirong Li & Chaoxi Xu
# Modified by Jie Wang & Fan Hu
# --------------------------------------------------------
import sys
import time
import json
import shutil
import importlib
import os
os.environ['CUDA_LAUNCH_BLOCKING'] = '1'
import numpy as np
import random
from torch.utils.tensorboard import SummaryWriter
import torch.multiprocessing
import torch.distributed as dist
import util
import evaluation
import data_providerneg as data
from common import *
from bigfile import BigFile
from generic_utils import Progbar
from model.model import get_model
from do_trainer import parse_args
from collections import OrderedDict
from pathlib import Path
PYTORCH_PRETRAINED_BERT_CACHE = Path(os.getenv('PYTORCH_PRETRAINED_BERT_CACHE',
Path.home() / '.pytorch_pretrained_bert'))
def load_config(config_path):
module = importlib.import_module(config_path)
return module.config()
def prepare_config(opt, checkToSkip=True,train=True ):
np.random.seed(opt.random_seed)
random.seed(opt.random_seed)
torch.manual_seed(opt.random_seed)
torch.backends.cudnn.deterministic = True
rootpath = opt.rootpath
trainCollection = opt.trainCollection
valCollection = opt.valCollection
if "task3_caption" in opt:
task3_caption_suffix = opt.task3_caption
else:
task3_caption_suffix='no_task3_caption'
if opt.val_set == 'no':
val_set = ''
else:
val_set = opt.val_set
# cuda number
global device
if torch.cuda.is_available() and opt.device != "cpu":
device = torch.device('cuda')
torch.cuda.manual_seed_all(opt.random_seed)
else:
device = torch.device('cpu')
config = load_config('configs.%s' % opt.config_name) # 模型参数文件
model_name = config.model_name
# set the config parm you adjust
if opt.parm_adjust_config != 'None':
config.adjust_parm(opt.parm_adjust_config)
model_path = os.path.join(rootpath, trainCollection, 'w2vvpp_train', valCollection, val_set, opt.config_name,
opt.model_prefix)
if checkToSkip :
if util.checkToSkip(os.path.join(model_path, 'model_best.pth.tar'), opt.overwrite):
sys.exit(0)
if not os.path.exists(model_path):
util.makedirs(model_path)
print(json.dumps(vars(opt), indent=2))
global writer
writer = SummaryWriter(log_dir=model_path, flush_secs=5)
config.cross=False
if not train:
config.float16 = True
return {'config': config}
collections = {'train': trainCollection, 'val': valCollection} # 数据集
capfiles = {'train': '%s.caption.txt', 'val': os.path.join(val_set, '%s.caption.txt')}
if train==False:
collections={}
capfiles_negationset=None
# 标注文件
cap_file_paths = {x: os.path.join(rootpath, collections[x], 'TextData', capfiles[x] % collections[x]) for x in
collections}
# ***************************萌萌哒*****************************
vis_feat_files = {x: None for x in collections}
vis_muti_feat_dicts = {x: None for x in collections}
vis_frame_feat_dicts = {x: None for x in collections}
# 视频帧文件
if config.frame_loader:
frame_id_path_file = {'train': os.path.join(rootpath, trainCollection, 'id.imagepath.txt'),
'val': os.path.join(rootpath, valCollection, 'id.imagepath.txt')
}
else:
frame_id_path_file = {'train': None,
'val': None
}
# ***************************萌萌哒*****************************
origin_vis_feat_files={x: None for x in collections}
if task3_caption_suffix == 'no_task3_caption':
config.task3 = False
capfiles_task3 = {'train': '%s.caption.txt' % ('%s'),
'val': os.path.join(val_set, '%s.caption.txt' % ('%s'))}
cap_file_paths_task3 = {
x: os.path.join(rootpath, collections[x], 'TextData', capfiles_task3[x] % collections[x])
for x in collections}
else:
config.task3=True
capfiles_task3 = {'train': '%s.caption.%s.txt' % ('%s', task3_caption_suffix),
'val': os.path.join(val_set, '%s.caption.%s.txt' % ('%s', task3_caption_suffix))}
cap_file_paths_task3 = {
x: os.path.join(rootpath, collections[x], 'TextData', capfiles_task3[x] % collections[x])
for x in collections}
startepoch=0
model = get_model(model_name, device, config)
if __name__ != '__main__' and torch.cuda.device_count() > 1 and opt.device!='cpu' :
print("GPU 大于1")
# 1） 初始化
torch.distributed.init_process_group(backend='nccl')
# 2） 配置每个进程的gpu
local_rank = torch.distributed.get_rank()
torch.cuda.set_device(local_rank)
device = torch.device("cuda", local_rank)
config.local_rank=local_rank
model.to(device)
model = torch.nn.parallel.DistributedDataParallel(model, device_ids=[local_rank], output_device=local_rank)
model = model.module
else:
model = model.to(device)
model_params = sum(p.numel() for p in model.parameters())
print('params: %.2fM' % (model_params / 1000000.0))
prepared_configs = {'vis_feat_files': vis_feat_files,
'origin_vis_feat_files': origin_vis_feat_files,
'vis_muti_feat_dicts': vis_muti_feat_dicts,
'vis_frame_feat_dicts': vis_frame_feat_dicts,
'frame_id_path_file': frame_id_path_file,
'cap_file_paths': cap_file_paths,
'cap_file_paths_task2': None,
'cap_file_paths_task3': cap_file_paths_task3,
'opt': opt,
'val_set': val_set,
'config': config,
'collections': collections,
'model_path': model_path,
'device': device,
'task3_caption_suffix': task3_caption_suffix,
'model': model,
"startepoch":startepoch
}
return prepared_configs
def main(opt):
prepared_configs = prepare_config(opt)
vis_feat_files = prepared_configs['vis_feat_files']
origin_vis_feat_files=prepared_configs['origin_vis_feat_files']
vis_frame_feat_dicts = prepared_configs['vis_frame_feat_dicts']
frame_id_path_file = prepared_configs['frame_id_path_file']
vis_muti_feat_dicts = prepared_configs['vis_muti_feat_dicts']
cap_file_paths = prepared_configs['cap_file_paths']
cap_file_paths_task2 = prepared_configs['cap_file_paths_task2']
cap_file_paths_task3= prepared_configs['cap_file_paths_task3']
opt = prepared_configs['opt']
config = prepared_configs['config']
collections = prepared_configs['collections']
model_path = prepared_configs['model_path']
model = prepared_configs['model']
device = prepared_configs['device']
val_set = prepared_configs['val_set']
vis_ids = list(
map(str.strip, open(os.path.join(opt.rootpath, opt.trainCollection, 'VideoSets', opt.trainCollection + '.txt'))))
params_parirloader = {x: {'vis_feat_files': vis_feat_files[x], 'capfile': cap_file_paths_task3[x],
'vis_frame_feat_dicts': vis_frame_feat_dicts[x],
'vis_ids': vis_ids,
'sample_type': config.frame_sample_type_train,
'vis_muti_feat_dicts': vis_muti_feat_dicts[x],
'frame_id_path_file': frame_id_path_file[x], 'pin_memory': False,
'batch_size': opt.batch_size, 'num_workers': opt.workers,
'config': config,
'collection': x,
'shuffle': (x == 'train'), 'task3': config.task3,'sampler':None,
'clip_vocab_size':config.clip_opt['vocab_size'],"max_txtlength":config.max_txtlength
} for x in ['train']}
if __name__ != '__main__' and torch.cuda.device_count() > 1:
params_parirloader["train"]['sampler'] = 'NotNone'
params_parirloader["train"]['shuffle'] = False
if config.task3:
data_loaders = {x: data.pair_provider_withneg(params_parirloader[x])
for x in ['train']}
else:
params_parirloader['capfile']=cap_file_paths
data_loaders = {x: data.pair_provider(params_parirloader[x])
for x in ['train']}
vis_ids = list(map(str.strip, open(os.path.join(opt.rootpath, opt.valCollection, 'VideoSets', opt.valCollection + '.txt'))))
vis_loader_val = data.vis_provider({'vis_feat_files': vis_feat_files['val'], 'vis_ids': vis_ids, 'pin_memory': False,
'vis_frame_feat_dicts': vis_frame_feat_dicts['val'],
'sample_type': config.frame_sample_type_test,
'frame_id_path_file': frame_id_path_file['val'],
'batch_size': int(opt.batch_size * 2),
'config': config,
'num_workers': opt.workers})
capfile = os.path.join(opt.rootpath, opt.valCollection, 'TextData', val_set, opt.valCollection+'.caption.txt')
txt_loader_val = data.txt_provider({'capfile': capfile, 'pin_memory': False, 'config': config,
'batch_size': opt.batch_size, 'task3': config.task3,
'capfile_task2': False, "max_txtlength": config.max_txtlength})
# Train the Model
best_perf = 0
no_impr_counter = 0
val_perf_hist_fout = open(os.path.join(model_path, 'val_perf_hist.txt'), 'w')
for epoch in range(opt.num_epochs):
logger.info(json.dumps(vars(opt), indent=2))
print('Epoch[{0} / {1}] LR: {2}'.format(epoch, opt.num_epochs, model.learning_rate))
print('-' * 10)
writer.add_scalar('train/learning_rate', model.learning_rate[0], epoch)
if epoch > 0 and hasattr(model, 'change_raw_global_emb_weight'):
model.change_raw_global_emb_weight()
# train for one epoch
train(model, data_loaders['train'], epoch)
# additional training data
if 'train2' in data_loaders:
train(model, data_loaders['train2'], epoch)
# evaluate on validation set
cur_perf = validate(model ,txt_loader_val, vis_loader_val, epoch, measure=config.measure, metric=opt.metric,
config=config)
model.lr_step(val_value=cur_perf)
print(' * Current perf: {}\n * Best perf: {}\n'.format(cur_perf, best_perf))
val_perf_hist_fout.write('epoch_%d:\nText2Video(%s): %f\n' % (epoch, opt.metric, cur_perf))
val_perf_hist_fout.flush()
# remember best performance and save checkpoint
is_best = cur_perf > best_perf
best_perf = max(cur_perf, best_perf)
save_checkpoint({'epoch': epoch + 1, 'model': model.state_dict(), 'best_perf': best_perf,
'config': config, 'opt': opt,'optimizer':model.optimizer.state_dict()}, is_best, logdir=model_path, only_best=False,
filename='checkpoint_epoch_%s.pth.tar' % epoch)
if is_best:
no_impr_counter = 0
no_impr_counter += 1
if no_impr_counter > 2 or epoch == opt.num_epochs-1:
save_checkpoint({'epoch': epoch + 1, 'model': model.state_dict(), 'best_perf': best_perf,
'config': config, 'opt': opt,'optimizer':model.optimizer.state_dict()}, is_best=False, logdir=model_path, only_best=True,
filename='checkpoint_epoch_%s.pth.tar' % epoch)
print('Early stopping happended or stopped.\n')
print(json.dumps(vars(opt), indent=2))
break
# 测试状态下早停
if __name__ == '__main__' and epoch > 11:
break
val_perf_hist_fout.close()
message = 'best performance on validation:\n Text to video({}): {}'.format(opt.metric, best_perf)
print(message)
with open(os.path.join(model_path, 'val_perf.txt'), 'w') as fout:
fout.write(message)
# if torch.cuda.device_count() > 1:
# dist.destroy_process_group()
def train(model, train_loader, epoch):
# average meters to record the training statistics
batch_time = util.AverageMeter()
data_time = util.AverageMeter()
# switch to train mode
model.train()
progbar = Progbar(len(train_loader.dataset))
end = time.time()
for i, train_data in enumerate(train_loader):
if __name__ == '__main__':
pass
if i > 5:
break
# sys.exit(0)
data_time.update(time.time() - end)
loss_items = model(train_data,epoch)
values = [('batch_time', batch_time.val)]
# print(loss_items)
# print(torch.cuda.is_available())
for key in loss_items.keys():
if isinstance(loss_items[key], torch.Tensor):
loss_items[key] = round(loss_items[key].item(), 4)
values.append((key, loss_items[key]))
progbar.add(len(list(train_data["vis_ids"])), values=values)
# measure elapsed time
batch_time.update(time.time() - end)
end = time.time()
# Record logs in tensorboard
writer.add_scalar('train/Loss', sum(list(loss_items.values())), model.iters)
for key in loss_items.keys():
writer.add_scalar('train/'+key, loss_items[key], model.iters)
print()
def validate(model, txt_loader, vis_loader, epoch, measure='cosine', metric='mir', config=None):
# compute the encoding for all the validation videos and captions
# vis_embs: 200*2048, txt_embs: 200*2048, vis_ids: 200, txt_ids: 200
if __name__ == '__main__':
debug=True
else:
debug=False
txt2vis_sim, txt_ids, vis_ids,labels = model.predict_multi(txt_loader, vis_loader, config.measure)
inds = np.argsort(txt2vis_sim, axis=1)
label_matrix = np.zeros(inds.shape) #
for index in range(inds.shape[0]):
ind = inds[index][::-1]
# print(txt_ids[index])
gt_index = np.in1d(np.array(vis_ids)[ind], labels[index])
label_matrix[index][gt_index] = 1
(r1, r5, r10, medr, meanr, mir, mAP,_,_,r5s,r10s,ranks) = evaluation.eval(label_matrix)
write_metric(r1, r5, r10, medr, meanr, mir, mAP, epoch)
return locals().get(metric, mir)
def write_metric(r1, r5, r10, medr, meanr, mir, mAP, epoch, mode="task1"):
sum_recall = r1 + r5 + r10
print(" * Text to video:")
print(" * r_1_5_10: {}".format([round(r1, 3), round(r5, 3), round(r10, 3)]))
print(" * medr, meanr, mir: {}".format([round(medr, 3), round(meanr, 3), round(mir, 3)]))
print(" * mAP: {}".format(round(mAP, 3)))
print(" * " + '-' * 10)
writer.add_scalar(mode + 'val/r1', r1, epoch)
writer.add_scalar(mode + 'val/r5', r5, epoch)
writer.add_scalar(mode + 'val/r10', r10, epoch)
writer.add_scalar(mode + 'val/medr', medr, epoch)
writer.add_scalar(mode + 'val/meanr', meanr, epoch)
writer.add_scalar(mode + 'val/mir', mir, epoch)
writer.add_scalar(mode + 'val/mAP', mAP, epoch)
def save_checkpoint(state, is_best, filename='checkpoint.pth.tar', only_best=False, logdir=''):
"""
:param state:
:param is_best: 比以前的好，就保存下来
:param filename:
:param only_best: 当结束训练时，only_best=True, 删除 checkpoint.pth.tar 文件，把 model_temp_best.pth.tar 文件 复制成 model_best.pth.tar
:param logdir:
:return:
"""
resfile = os.path.join(logdir, filename)
if is_best:
torch.save(state, resfile)
shutil.copyfile(resfile, os.path.join(logdir, 'model_temp_best.pth.tar'))
if only_best:
if os.path.exists(os.path.join(logdir, 'model_temp_best.pth.tar')):
shutil.copyfile(os.path.join(logdir, 'model_temp_best.pth.tar'), os.path.join(logdir, 'model_best.pth.tar'))
os.remove(os.path.join(logdir, 'model_temp_best.pth.tar'))
if __name__ == '__main__':
if len(sys.argv) == 1:
print()
sys.argv = "trainer.py --device 1 msrvtt10ktrain msrvtt10kval " \
"--rootpath /home/wzy/VisualSearch --batch_size 32 " \
"--train_strategy usual " \
"--config_name CLIP.CLIPEnd2End_adjust " \
"--parm_adjust_config 1 " \
"--val_set no " \
"--save_mean_last 0 " \
"--pretrained_file_path None " \
"--model_prefix runs_9_ --overwrite 1 " \
"--task3_caption no_task3_caption".split(' ')
opt = parse_args() # 这里opt是输入的值，config 才是参数文件中读取的
main(opt)