-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathtrain.py
185 lines (155 loc) · 7.84 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
# Copyright (C) 2021. Huawei Technologies Co., Ltd. All rights reserved.
# This program is free software; you can redistribute it and/or modify
# it under the terms of the MIT License.
# This program is distributed in the hope that it will be useful,
# but WITHOUT ANY WARRANTY; without even the implied warranty of
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
# MIT License for more details.
import numpy as np
from tqdm import tqdm
from accelerate import Accelerator
import torch
from torch.utils.data import DataLoader
# from torch.utils.tensorboard import SummaryWriter
from tensorboardX import SummaryWriter
import params
from model.tts import MyGradTTS
from data import TextMelDataset, TextMelBatchCollate
from utils import plot_tensor, save_plot
from text.symbols import symbols
from data_preprocess.core import make_loaders
train_filelist_path = params.train_filelist_path
valid_filelist_path = params.valid_filelist_path
cmudict_path = params.cmudict_path
add_blank = params.add_blank
log_dir = "logs/0604"
n_epochs = params.n_epochs
batch_size = params.batch_size #16
out_size = params.out_size #172
learning_rate = params.learning_rate
random_seed = params.seed
nsymbols = len(symbols) + 1 if add_blank else len(symbols) #149
n_enc_channels = params.n_enc_channels #192
filter_channels = params.filter_channels #768
filter_channels_dp = params.filter_channels_dp #256
n_enc_layers = params.n_enc_layers #6
enc_kernel = params.enc_kernel #3
enc_dropout = params.enc_dropout #0.1
n_heads = params.n_heads #2
window_size = params.window_size #4
n_feats = params.n_feats #80
n_fft = params.n_fft #1024
sample_rate = params.sample_rate #22050
hop_length = params.hop_length #256
win_length = params.win_length #1024
f_min = params.f_min #0
f_max = params.f_max #8000
dec_dim = params.dec_dim #64
beta_min = params.beta_min #0.05
beta_max = params.beta_max #20.0
pe_scale = params.pe_scale #1000
if __name__ == "__main__":
torch.manual_seed(random_seed)
np.random.seed(random_seed)
accelerator = Accelerator()
print('Initializing logger...')
logger = SummaryWriter(log_dir=log_dir)
# print('Initializing data loaders...')
# train_dataset = TextMelDataset(train_filelist_path, cmudict_path, add_blank,
# n_fft, n_feats, sample_rate, hop_length,
# win_length, f_min, f_max) #y : [80,759],x : [229]
# batch_collate = TextMelBatchCollate()
# loader = DataLoader(dataset=train_dataset, batch_size=batch_size,
# collate_fn=batch_collate, drop_last=True,
# num_workers=4, shuffle=False)
# test_dataset = TextMelDataset(valid_filelist_path, cmudict_path, add_blank,
# n_fft, n_feats, sample_rate, hop_length,
# win_length, f_min, f_max)
loader=make_loaders("train")
test_loader=make_loaders("test")
it=iter(test_loader)
test_batch=next(it)
print('Initializing model...')
model = MyGradTTS(nsymbols, 1, None, n_enc_channels, filter_channels, filter_channels_dp,
n_heads, n_enc_layers, enc_kernel, enc_dropout, window_size,
n_feats, dec_dim, beta_min, beta_max, pe_scale).cuda()
print('Number of encoder + duration predictor parameters: %.2fm' % (model.encoder.nparams/1e6))
print('Number of decoder parameters: %.2fm' % (model.decoder.nparams/1e6))
print('Total parameters: %.2fm' % (model.nparams/1e6))
print('Initializing optimizer...')
# model = torch.nn.DataParallel(model)
optimizer = torch.optim.Adam(params=model.parameters(), lr=learning_rate)
print('Logging test batch...')
# for i, item in enumerate(test_batch):
mel = test_batch[2][0].transpose(0,1)
logger.add_image(f'image_0/ground_truth', plot_tensor(mel),global_step=0, dataformats='HWC')
save_plot(mel, f'{log_dir}/original_{0}.png')
print('Start training...')
iteration = 0
model.load(f"{log_dir}/grad_2_.pt")
loader, model, optimizer = accelerator.prepare(
loader, model, optimizer
)
for epoch in range(1, n_epochs + 1):
model.train()
prior_losses = []
diff_losses = []
with tqdm(loader, total=len(loader)//batch_size) as progress_bar:
for batch_idx, batch in enumerate(progress_bar):
model.zero_grad()
notes, phonemes, y = batch[0], batch[1], batch[2]
prior_loss,diff_loss = model.module.compute_loss(notes, phonemes, y, out_size=out_size)
loss = sum([prior_loss,diff_loss])
accelerator.backward(loss)
enc_grad_norm = torch.nn.utils.clip_grad_norm_(model.module.encoder.parameters(),
max_norm=1)
dec_grad_norm = torch.nn.utils.clip_grad_norm_(model.module.decoder.parameters(),
max_norm=1)
optimizer.step()
logger.add_scalar('training/prior_loss', prior_loss.item(),
global_step=iteration)
logger.add_scalar('training/diffusion_loss', diff_loss.item(),
global_step=iteration)
logger.add_scalar('training/encoder_grad_norm', enc_grad_norm,
global_step=iteration)
logger.add_scalar('training/decoder_grad_norm', dec_grad_norm,
global_step=iteration)
prior_losses.append(prior_loss.item())
diff_losses.append(diff_loss.item())
if batch_idx % 5 == 0:
msg = f'Epoch: {epoch}, iteration: {iteration} | diff_loss: {diff_loss.item()}'
progress_bar.set_description(msg)
iteration += 1
log_msg = 'Epoch %d: ' % (epoch)
log_msg += '| prior loss = %.3f ' % np.mean(prior_losses)
log_msg += '| diffusion loss = %.3f\n' % np.mean(diff_losses)
with open(f'{log_dir}/train.log', 'a') as f:
f.write(log_msg)
if epoch % params.save_every > 0:
continue
model.eval()
print('Synthesis...')
with torch.no_grad():
# for i, item in enumerate(test_batch):
notes, phonemes, y = test_batch[0][0].unsqueeze(0).cuda(), test_batch[1][0].unsqueeze(0).cuda(), test_batch[2][0].unsqueeze(0).cuda()
i=0
# # x = item['x'].to(torch.long).unsqueeze(0).cuda()
# x_lengths = torch.LongTensor([x.shape[-1]]).cuda()
y_enc, y_dec = model.module(notes, phonemes, n_timesteps=50)
logger.add_image(f'image_{i}/generated_enc',
plot_tensor(y_enc.squeeze().cpu()),
global_step=iteration, dataformats='HWC')
logger.add_image(f'image_{i}/generated_dec',
plot_tensor(y_dec.squeeze().cpu()),
global_step=iteration, dataformats='HWC')
# logger.add_image(f'image_{i}/alignment',
# plot_tensor(attn.squeeze().cpu()),
# global_step=iteration, dataformats='HWC')
save_plot(y_enc.squeeze().cpu(), f'{log_dir}/generated_enc_{i}.png')
print(f'{log_dir}/generated_enc_{i}.png')
save_plot(y_dec.squeeze().cpu(), f'{log_dir}/generated_dec_{i}.png')
print(f'{log_dir}/generated_dec_{i}.png')
# save_plot(attn.squeeze().cpu(),
# f'{log_dir}/alignment_{i}.png')
ckpt = model.module.state_dict()
torch.save(ckpt, f=f"{log_dir}/grad_{epoch}.pt")