forked from NVIDIA/radtts
-
Notifications
You must be signed in to change notification settings - Fork 0
/
inference_waveglowe.py
195 lines (164 loc) · 8.26 KB
/
inference_waveglowe.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
# SPDX-FileCopyrightText: Copyright (c) 2022 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
# SPDX-License-Identifier: MIT
#
# Permission is hereby granted, free of charge, to any person obtaining a
# copy of this software and associated documentation files (the "Software"),
# to deal in the Software without restriction, including without limitation
# the rights to use, copy, modify, merge, publish, distribute, sublicense,
# and/or sell copies of the Software, and to permit persons to whom the
# Software is furnished to do so, subject to the following conditions:
#
# The above copyright notice and this permission notice shall be included in
# all copies or substantial portions of the Software.
#
# THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
# IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
# FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
# THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
# LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
# FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
# DEALINGS IN THE SOFTWARE.
import sys
import matplotlib
matplotlib.use("Agg")
import matplotlib.pylab as plt
import argparse
import os
import json
import numpy as np
import torch
from torch.cuda import amp
from scipy.io.wavfile import write
from radtts import RADTTS
from data import Data
from common import update_params
from glow import WaveGlow
from denoiser import Denoiser
def lines_to_list(filename):
"""
Takes a text file of filenames and makes a list of filenames
"""
with open(filename, encoding='utf-8') as f:
files = f.readlines()
files = [f.rstrip() for f in files]
return files
def load_vocoder(vocoder_path, to_cuda=True):
vocoder = torch.load(vocoder_path)['model']
denoiser = Denoiser(vocoder)
for k in vocoder.convinv:
k.float()
if to_cuda:
vocoder.cuda()
denoiser.cuda()
vocoder.eval()
denoiser.eval()
denoiser = Denoiser(vocoder)
return vocoder, denoiser
def infer(radtts_path, vocoder_path, text_path, speaker,
speaker_text, speaker_attributes, sigma, sigma_tkndur, sigma_f0,
sigma_energy, f0_mean, f0_std, energy_mean, energy_std,
token_dur_scaling, denoising_strength, n_takes, output_dir, use_amp,
plot, seed):
torch.manual_seed(seed)
torch.cuda.manual_seed(seed)
vocoder, denoiser = load_vocoder(vocoder_path)
radtts = RADTTS(**model_config).cuda()
radtts.enable_inverse_cache() # cache inverse matrix for 1x1 invertible convs
checkpoint_dict = torch.load(radtts_path, map_location='cpu')
state_dict = checkpoint_dict['state_dict']
radtts.load_state_dict(state_dict, strict=False)
radtts.eval()
print("Loaded checkpoint '{}')" .format(radtts_path))
ignore_keys = ['training_files', 'validation_files']
trainset = Data(
data_config['training_files'],
**dict((k, v) for k, v in data_config.items() if k not in ignore_keys))
speaker_id = trainset.get_speaker_id(speaker).cuda()
speaker_id_text, speaker_id_attributes = speaker_id, speaker_id
if speaker_text is not None:
speaker_id_text = trainset.get_speaker_id(speaker_text).cuda()
if speaker_attributes is not None:
speaker_id_attributes = trainset.get_speaker_id(
speaker_attributes).cuda()
text_list = lines_to_list(text_path)
os.makedirs(output_dir, exist_ok=True)
for i, text in enumerate(text_list):
if text.startswith("#"):
continue
print("{}/{}: {}".format(i, len(text_list), text))
text = trainset.get_text(text).cuda()[None]
for take in range(n_takes):
with amp.autocast(use_amp):
with torch.no_grad():
outputs = radtts.infer(
speaker_id, text, sigma, sigma_tkndur, sigma_f0,
sigma_energy, token_dur_scaling, token_duration_max=100,
speaker_id_text=speaker_id_text,
speaker_id_attributes=speaker_id_attributes,
f0_mean=f0_mean, f0_std=f0_std, energy_mean=energy_mean,
energy_std=energy_std)
mel = outputs['mel']
with torch.no_grad():
audio = vocoder.infer(mel, 0.8).float()
audio_denoised = denoiser(audio, strength=denoising_strength)[0].float()
audio = audio[0].cpu().numpy()
audio_denoised = audio_denoised[0].cpu().numpy()
audio_denoised = audio_denoised / np.max(np.abs(audio_denoised))
suffix_path = "{}_{}_{}_durscaling{}_sigma{}_sigmatext{}_sigmaf0{}_sigmaenergy{}".format(
speaker, i, take, token_dur_scaling, sigma, sigma_tkndur, sigma_f0,
sigma_energy)
write("{}/{}_denoised_{}.wav".format(
output_dir, suffix_path, denoising_strength),
data_config['sampling_rate'], audio_denoised)
if plot:
fig, axes = plt.subplots(2, 1, figsize=(10, 6))
axes[0].plot(outputs['f0'].cpu().numpy()[0], label='f0')
axes[1].plot(outputs['energy_avg'].cpu().numpy()[0], label='energy_avg')
for ax in axes:
ax.legend(loc='best')
plt.tight_layout()
fig.savefig("{}/{}_features.png".format(output_dir, suffix_path))
plt.close('all')
if __name__ == "__main__":
parser = argparse.ArgumentParser()
parser.add_argument('-c', '--config', type=str, help='JSON file config')
parser.add_argument('-p', '--params', nargs='+', default=[])
parser.add_argument('-r', '--radtts_path', type=str)
parser.add_argument('-v', '--vocoder_path', type=str)
parser.add_argument('-t', '--text_path', default='sentences.txt', type=str)
parser.add_argument('-s', '--speaker', type=str)
parser.add_argument('--speaker_text', type=str, default=None)
parser.add_argument('--speaker_attributes', type=str, default=None)
parser.add_argument('-d', '--denoising_strength', type=float, default=0.01)
parser.add_argument('-o', "--output_dir", default="results")
parser.add_argument("--sigma", default=0.8, type=float, help="sampling sigma for decoder")
parser.add_argument("--sigma_tkndur", default=0.667, type=float, help="sampling sigma for duration")
parser.add_argument("--sigma_f0", default=1.0, type=float, help="sampling sigma for f0")
parser.add_argument("--sigma_energy", default=1.0, type=float, help="sampling sigma for energy avg")
parser.add_argument("--f0_mean", default=0.0, type=float)
parser.add_argument("--f0_std", default=0.0, type=float)
parser.add_argument("--energy_mean", default=0.0, type=float) #not used
parser.add_argument("--energy_std", default=0.0, type=float)#not used
parser.add_argument("--token_dur_scaling", default=1.00, type=float)
parser.add_argument("--n_takes", default=1, type=int)
parser.add_argument("--use_amp", action="store_true")
parser.add_argument("--plot", action="store_true")
parser.add_argument("--seed", default=1234, type=int)
args = parser.parse_args()
# Parse configs. Globals nicer in this case
with open(args.config) as f:
data = f.read()
global config
config = json.loads(data)
update_params(config, args.params)
data_config = config["data_config"]
global model_config
model_config = config["model_config"]
torch.backends.cudnn.enabled = True
torch.backends.cudnn.benchmark = False
infer(args.radtts_path, args.vocoder_path,
args.text_path, args.speaker, args.speaker_text,
args.speaker_attributes, args.sigma, args.sigma_tkndur, args.sigma_f0,
args.sigma_energy, args.f0_mean, args.f0_std, args.energy_mean,
args.energy_std, args.token_dur_scaling, args.denoising_strength,
args.n_takes, args.output_dir, args.use_amp, args.plot, args.seed)