-
Notifications
You must be signed in to change notification settings - Fork 33
/
main.py
executable file
·204 lines (178 loc) · 9.27 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
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
import sys
import os
import shutil
import torch
from model import NeuralSpeakerModel, AngleLoss
import numpy as np
import argparse
import torch.optim as optim
import torch.nn.functional as F
from datasets import SequenceDataset
# Author: Nanxin Chen
class ScheduledOptim(object):
""" A simple wrapper class for learning rate scheduling """
def __init__(self, optimizer, n_warmup_steps):
self.optimizer = optimizer
self.d_model = 64
self.n_warmup_steps = n_warmup_steps
self.n_current_steps = 0
self.delta = 1
def step(self):
"Step by the inner optimizer"
self.optimizer.step()
def zero_grad(self):
"Zero out the gradients by the inner optimizer"
self.optimizer.zero_grad()
def increase_delta(self):
self.delta *= 2
def update_learning_rate(self):
"Learning rate scheduling per step"
self.n_current_steps += self.delta
new_lr = np.power(self.d_model, -0.5) * np.min([
np.power(self.n_current_steps, -0.5),
np.power(self.n_warmup_steps, -1.5) * self.n_current_steps])
for param_group in self.optimizer.param_groups:
param_group['lr'] = new_lr
return new_lr
def state_dict(self):
ret = {
'd_model': self.d_model,
'n_warmup_steps': self.n_warmup_steps,
'n_current_steps': self.n_current_steps,
'delta': self.delta,
}
ret['optimizer'] = self.optimizer.state_dict()
return ret
def load_state_dict(self, state_dict):
self.d_model = state_dict['d_model']
self.n_warmup_steps = state_dict['n_warmup_steps']
self.n_current_steps = state_dict['n_current_steps']
self.delta = state_dict['delta']
self.optimizer.load_state_dict(state_dict['optimizer'])
if __name__ == "__main__":
parser = argparse.ArgumentParser(description='PyTorch MNIST Example')
parser.add_argument('--train', type=str, help='training scp')
parser.add_argument('--cv', type=str, help='cv scp')
parser.add_argument('--utt2spkid', type=str, help='utt2spkid')
parser.add_argument('--spk_num', type=int, help='number of speakers')
parser.add_argument('--batch-size', type=int, default=128, metavar='N',
help='input batch size for training (default: 64)')
parser.add_argument('--test-batch-size', type=int, default=256, metavar='N',
help='input batch size for testing (default: 1000)')
parser.add_argument('--epochs', type=int, default=50, metavar='N',
help='number of epochs to train (default: 10)')
parser.add_argument('-n_warmup_steps', type=int, default=8000)
parser.add_argument('--no-cuda', action='store_true', default=False,
help='disables CUDA training')
parser.add_argument('--seed', type=int, default=1, metavar='S',
help='random seed (default: 1)')
parser.add_argument('--log-interval', type=int, default=10, metavar='N',
help='how many batches to wait before logging training status')
parser.add_argument('--model', type=str, required=True, help='feature extractor model type')
parser.add_argument('--input-dim', type=int, required=True, help='input feature dimension')
parser.add_argument('--D', type=int, required=True, help='LDE dictionary components')
parser.add_argument('--hidden-dim', type=int, required=True, help='speaker embedding dimension')
parser.add_argument('--pooling', type=str, required=True, help='mean or mean+std')
parser.add_argument('--network-type', type=str, required=True, help='lde or att')
parser.add_argument('--distance-type', type=str, required=True, help='sqr or norm')
parser.add_argument('--asoftmax', required=True, help='True or False')
parser.add_argument('--m', type=int, help='m for A-softmax')
parser.add_argument('--min-chunk-size', type=int, required=True, help='minimum feature map length')
parser.add_argument('--max-chunk-size', type=int, required=True, help='maximum feature map length')
parser.add_argument('--log-dir', type=str, required=True, help='logging directory')
parser.add_argument('--pretrain-model-pth', type=str)
args = parser.parse_args()
use_cuda = not args.no_cuda and torch.cuda.is_available()
print('use cuda is %s' % use_cuda)
torch.manual_seed(args.seed)
device = torch.device("cuda" if use_cuda else "cpu")
tmp = torch.Tensor([2]).to(device)
kwargs = {'num_workers': 2, 'pin_memory': True} if use_cuda else {}
train = SequenceDataset(scp_file=args.train, utt2spkid_file=args.utt2spkid, min_length=args.max_chunk_size)
val = SequenceDataset(scp_file=args.cv, utt2spkid_file=args.utt2spkid, min_length=args.max_chunk_size)
train_loader = torch.utils.data.DataLoader(
train, batch_size=args.batch_size, shuffle=True, **kwargs)
val_loader = torch.utils.data.DataLoader(
val, batch_size=args.test_batch_size, shuffle=False, **kwargs)
model=NeuralSpeakerModel(model=args.model, input_dim=args.input_dim, output_dim=args.spk_num, D=args.D, hidden_dim=args.hidden_dim, \
pooling=args.pooling, network_type=args.network_type, distance_type=args.distance_type, asoftmax=args.asoftmax, m=args.m).to(device)
model_params = sum(p.numel() for p in model.parameters() if p.requires_grad)
print('===> Model total parameter: {}'.format(model_params))
optimizer = ScheduledOptim( # Transformer optimizer
optim.Adam(
filter(lambda p: p.requires_grad, model.parameters()),
betas=(0.9, 0.98), eps=1e-09, weight_decay=1e-4, amsgrad=True),
args.n_warmup_steps)
start_epoch = 1
best = 0
best_epoch = -1
if args.pretrain_model_pth is not None:
if os.path.isfile(args.pretrain_model_pth):
print('loading pre-trained model from %s' % args.pretrain_model_pth)
model_dict = model.state_dict()
checkpoint = torch.load(args.pretrain_model_pth, map_location=lambda storage, loc: storage) # load for cpu
start_epoch = checkpoint['epoch']
best_epoch = start_epoch
best = checkpoint['best_acc']
model.load_state_dict(checkpoint['state_dict'])
optimizer.load_state_dict(checkpoint['optimizer'])
else:
print("===> no checkpoint found at '{}'".format(args.pretrain_model_pth))
exit()
if args.asoftmax == 'True': # angular-softmax
print('training with Angular Softmax')
criterion = AngleLoss()
else:
print('training with Softmax')
criterion = torch.nn.NLLLoss()
# ------------------
# main training loop
# ------------------
for epoch in range(start_epoch, start_epoch+args.epochs):
print('Epoch %d' % epoch)
model.train()
for batch_idx, (data, target) in enumerate(train_loader):
data, target = data.to(device), target.to(device, non_blocking=True).view((-1,))
optimizer.zero_grad()
output = model(data)
loss = criterion(output, target)
loss.backward()
optimizer.step()
lr = optimizer.update_learning_rate()
if batch_idx % args.log_interval == 0:
print('Train Epoch: {} [{}/{} ({:.0f}%)]\tlr:{:.5f}\tLoss: {:.6f}'.format(
epoch, batch_idx * len(data), len(train_loader.dataset),
100. * batch_idx / len(train_loader), lr, loss.item()))
train.update(np.random.randint(args.min_chunk_size, args.max_chunk_size+1)) # 3-8s chunk
del data, target, output, loss
model.eval()
test_loss = 0
correct = 0
with torch.no_grad():
for data, target in val_loader:
data, target = data.to(device), target.to(device, non_blocking=True).view((-1,))
output = model(data)
test_loss += criterion(output, target).item() # sum up batch loss
if args.asoftmax == 'True': # angular-softmax
output = output[0] # 0=cos_theta 1=phi_theta
pred = output.max(1, keepdim=True)[1] # get the index of the max log-probability
correct += pred.eq(target.view_as(pred)).sum().item()
test_loss /= len(val_loader.dataset)
print('\nTest set: Average loss: {:.4f}, Accuracy: {}/{} ({:.0f}%)\n'.format(
test_loss, correct, len(val_loader.dataset),
100. * correct / len(val_loader.dataset)))
if 100. * correct / len(val_loader.dataset) > best:
best = 100. * correct / len(val_loader.dataset)
torch.save({
'epoch': epoch,
'state_dict': model.state_dict(),
'best_acc': best,
'optimizer' : optimizer.state_dict(),
}, args.log_dir + str(epoch) + "_" + str(int(100. * correct / len(val_loader.dataset))) + ".h5")
print("===> save to checkpoint at {}\n".format(args.log_dir + 'model_best.pth.tar'))
shutil.copyfile(args.log_dir + str(epoch) + "_" + str(int(100. * correct / len(val_loader.dataset))) +
".h5", args.log_dir + 'model_best.pth.tar')
best_epoch = epoch
elif epoch - best_epoch > 2:
optimizer.increase_delta()
best_epoch = epoch