/
train.py
186 lines (154 loc) · 7.33 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
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
# -*- coding: utf-8 -*-
import argparse
import glob
import os
import sys
import time
import torch
import torch.multiprocessing as mp
import torch.nn.functional as F
from torch.nn.parallel import DistributedDataParallel as DDP
from torch.utils.data import DataLoader
from torch.utils.data.distributed import DistributedSampler
from torch.utils.tensorboard import SummaryWriter
from utils import hparams as hp
from Models.AttModel import AttModel
from Models.CTCModel import CTCModel
from utils.utils import frame_stacking, log_config, load_model, adjust_learning_rate, fill_variables, adjust_transformer_learning_rate
from Loss.label_smoothing import label_smoothing_loss
from load_waveform import load_waveform_model
import datasets
DEVICE = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
def train_loop(model, optimizer, writer, step, args, hp) -> int:
dataset_train = datasets.get_dataset(hp.train_script, hp, use_spec_aug=hp.use_spec_aug)
if hp.encoder_type == 'Conformer':
train_sampler = datasets.LengthsBatchSampler(dataset_train, hp.batch_size * 1500, hp.lengths_file)
dataloader = DataLoader(dataset_train, batch_sampler=train_sampler, num_workers=2, collate_fn=datasets.collate_fn)
else:
train_sampler = DistributedSampler(dataset_train) if args.n_gpus > 1 else None
dataloader = DataLoader(dataset_train, batch_size=hp.batch_size, shuffle=hp.shuffle, num_workers=2, sampler=train_sampler, collate_fn=datasets.collate_fn, drop_last=True)
optimizer.zero_grad()
for d in dataloader:
step += 1
if hp.encoder_type == 'Conformer':
lr = adjust_transformer_learning_rate(step)
print(f'step = {step}')
print(f'lr = {lr}')
text, mel_input, pos_text, pos_mel, text_lengths, mel_lengths = d
text = text.to(DEVICE, non_blocking=True)
mel_input = mel_input.to(DEVICE, non_blocking=True)
pos_text = pos_text.to(DEVICE, non_blocking=True)
pos_mel = pos_mel.to(DEVICE, non_blocking=True)
text_lengths = text_lengths.to(DEVICE, non_blocking=True)
if hp.frame_stacking > 1 and hp.encoder_type != 'Wave':
mel_input, mel_lengths = frame_stacking(mel_input, mel_lengths, hp.frame_stacking)
predict_ts = model(mel_input, mel_lengths, text, pos_mel)
if hp.decoder_type == 'Attention':
loss = label_smoothing_loss(predict_ts, text, text_lengths, hp.T_norm, hp.B_norm)
elif hp.decoder_type == 'CTC':
predict_ts = F.log_softmax(predict_ts, dim=2).transpose(0, 1)
loss = F.ctc_loss(predict_ts, text, mel_lengths, text_lengths, blank=0)
loss.backward()
if step % hp.accum_grad == 0:
torch.nn.utils.clip_grad_norm_(model.parameters(), hp.clip)
optimizer.step()
optimizer.zero_grad()
loss.detach()
if torch.isnan(loss):
print('loss is nan')
sys.exit(1)
if hp.debug_mode == 'tensorboard':
#writer.add_scalar("Loss/train", loss, step)
print('loss = {}'.format(loss.item()))
else:
print('loss = {}'.format(loss.item()))
sys.stdout.flush()
return step
def train_epoch(model, optimizer, args, hp, start_epoch=0):
writer = None
if hp.load_wav_model:
model = load_waveform_model(model, hp.load_wav_model)
import pdb;pdb.set_trace()
print(f'{hp.load_wav_model} loaded')
dataset_train = datasets.get_dataset(hp.train_script, hp, use_spec_aug=hp.use_spec_aug)
train_sampler = DistributedSampler(dataset_train) if args.n_gpus > 1 else None
dataloader = DataLoader(dataset_train, batch_size=hp.batch_size, shuffle=hp.shuffle, sampler=train_sampler,
num_workers=1, collate_fn=datasets.collate_fn, drop_last=True)
step = len(dataloader) * start_epoch
for epoch in range(start_epoch, hp.max_epoch):
start_time = time.time()
step = train_loop(model, optimizer, writer, step, args, hp)
if (epoch + 1) % hp.save_per_epoch == 0 or (epoch+1) % hp.reset_optimizer_epoch > 10:
torch.save(model.state_dict(), hp.save_dir + "/network.epoch{}".format(epoch + 1))
torch.save(optimizer.state_dict(), hp.save_dir + "/network.optimizer.epoch{}".format(epoch + 1))
if hp.encoder_type != 'Conformer':
adjust_learning_rate(optimizer, epoch + 1, hp)
if (epoch + 1) % hp.reset_optimizer_epoch == 0:
#if (epoch + 1) in hp.reset_optimizer_epoch == 0:
if hp.encoder_type != 'Conformer':
optimizer = torch.optim.Adam(model.parameters(), weight_decay=1e-5)
print("EPOCH {} end".format(epoch + 1))
print(f'elapsed time = {(time.time()-start_time)//60}m')
def init_distributed(rank, n_gpus):
assert torch.cuda.is_available(), "Distributed mode requires CUDA."
torch.cuda.set_device(rank % n_gpus)
os.environ['MASTER_ADDR'] = 'localhost' #dist_config.MASTER_ADDR
os.environ['MASTER_PORT'] = '600010' #dist_config.MASTER_PORT
torch.distributed.init_process_group(
backend='nccl', world_size=n_gpus, rank=rank
)
def cleanup():
torch.distributed.destroy_process_group()
def run_distributed(fn, args, hp):
try:
mp.spawn(fn, args=(args, hp), nprocs=args.n_gpus, join=True)
except:
cleanup()
def run_training(rank, args, hp):
if args.n_gpus > 1:
init_distributed(rank, args.n_gpus)
torch.cuda.set_device(f'cuda:{rank}')
if hp.decoder_type == 'Attention':
model = AttModel(hp)
elif hp.decoder_type == 'CTC':
model = CTCModel(hp)
model = model.to(rank)
if args.n_gpus > 1:
model = DDP(model, device_ids=[rank])
if hp.encoder_type == 'Conformer':
optimizer = torch.optim.Adam(model.parameters(), lr=1e-3, betas=(0.9, 0.98), eps=1e-9)
else:
optimizer = torch.optim.Adam(model.parameters(), lr=hp.lr, weight_decay=1e-5)
load_epoch = 0
if hp.load_checkpoints:
if hp.load_checkpoints_epoch is None:
path_list = glob.glob(os.path.join(hp.load_checkpoints_path, 'network.epoch*'))
for path in path_list:
epoch = int(path.split('.')[-1].replace('epoch', ''))
if epoch > load_epoch:
load_epoch = epoch
else:
load_epoch = hp.load_checkpoints_epoch
print("{} epoch {} load".format(hp.load_checkpoints_path, load_epoch))
model.load_state_dict(load_model(os.path.join(hp.load_checkpoints_path, 'network.epoch{}'.format(load_epoch))))
if hp.is_flatstart:
load_epoch = 0
#pass
else:
optimizer.load_state_dict(torch.load(os.path.join(hp.load_checkpoints_path, 'network.optimizer.epoch{}'.format(load_epoch))))
train_epoch(model, optimizer, args, hp, start_epoch=load_epoch)
if __name__ == "__main__":
parser = argparse.ArgumentParser()
parser.add_argument('--hp_file', metavar='FILE', default='hparams.py')
args = parser.parse_args()
hp.configure(args.hp_file)
fill_variables(hp)
os.makedirs(hp.save_dir, exist_ok=True)
writer = SummaryWriter(f'{hp.save_dir}/logs/{hp.comment}')
log_config(hp)
n_gpus = torch.cuda.device_count()
args.__setattr__('n_gpus', n_gpus)
if n_gpus > 1:
run_distributed(run_training, args, hp)
else:
run_training(0, args, hp)