-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathinference.py
190 lines (172 loc) · 6.12 KB
/
inference.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
import numpy as np
from tqdm import tqdm
import torch
import torch.nn.functional as F
from geco.data_module import SpecsDataModule
from geco.sdes import BBED
from fastgeco.model import ScoreModel
from geco.util.other import pad_spec
import os
import torchaudio
import shutil
from speechbrain.lobes.models.dual_path import Encoder, SBTransformerBlock, SBTransformerBlock, Dual_Path_Model, Decoder
import argparse
def load_sepformer(args):
ckpt_path = args.ckpt
encoder = Encoder(
kernel_size=160,
out_channels=256,
in_channels=1
)
SBtfintra = SBTransformerBlock(
num_layers=8,
d_model=256,
nhead=8,
d_ffn=1024,
dropout=0,
use_positional_encoding=True,
norm_before=True,
)
SBtfinter = SBTransformerBlock(
num_layers=8,
d_model=256,
nhead=8,
d_ffn=1024,
dropout=0,
use_positional_encoding=True,
norm_before=True,
)
masknet = Dual_Path_Model(
num_spks=args.num_spks,
in_channels=256,
out_channels=256,
num_layers=2,
K=250,
intra_model=SBtfintra,
inter_model=SBtfinter,
norm='ln',
linear_layer_after_inter_intra=False,
skip_around_intra=True,
)
decoder = Decoder(
in_channels=256,
out_channels=1,
kernel_size=160,
stride=80,
bias=False,
)
encoder_weights = torch.load(os.path.join(ckpt_path, 'encoder.ckpt'))
encoder.load_state_dict(encoder_weights)
masknet_weights = torch.load(os.path.join(ckpt_path, 'masknet.ckpt'))
masknet.load_state_dict(masknet_weights)
decoder_weights = torch.load(os.path.join(ckpt_path, 'decoder.ckpt'))
decoder.load_state_dict(decoder_weights)
return encoder, masknet, decoder
@torch.no_grad()
def separate(args, encoder, masknet, decoder, savename):
print('Process SepFormer...')
mix, fs_file = torchaudio.load(args.test_file)
mix = mix.cuda()
fs_model = args.sample_rate
# resample the data if needed
if fs_file != fs_model:
print(
"Resampling the audio from {} Hz to {} Hz".format(
fs_file, fs_model
)
)
tf = torchaudio.transforms.Resample(
orig_freq=fs_file, new_freq=fs_model
).cuda()
mix = mix.mean(dim=0, keepdim=True)
mix = tf(mix)
mix = mix.cuda()
# Separation
mix_w = encoder(mix)
est_mask = masknet(mix_w)
mix_w = torch.stack([mix_w] * args.num_spks)
sep_h = mix_w * est_mask
# Decoding
est_sources = torch.cat(
[
decoder(sep_h[i]).unsqueeze(-1)
for i in range(args.num_spks)
],
dim=-1,
)
est_sources = (
est_sources / est_sources.abs().max(dim=1, keepdim=True)[0]
).squeeze()
for i in range(args.num_spks):
torchaudio.save(
os.path.join(args.save_folder, savename+'_spk'+str(i+1)+'.wav'), est_sources[:,i].unsqueeze(0).cpu(), args.sample_rate
)
shutil.copyfile(args.test_file, os.path.join(args.save_folder, savename+'_mix.wav'))
return est_sources, mix
def load_fastgeco(args):
checkpoint_file = os.path.join(args.ckpt, 'fastgeco.ckpt')
model = ScoreModel.load_from_checkpoint(
checkpoint_file,
batch_size=1, num_workers=0, kwargs=dict(gpu=False)
)
model.eval(no_ema=False)
model.cuda()
return model
@torch.no_grad()
def correct(args, model, est_sources, mix, savename):
print('Process Fast-Geco...')
N = args.N
reverse_starting_point = args.reverse_starting_point
sr = args.sample_rate
for idx in range(args.num_spks):
y = est_sources[:, idx].unsqueeze(0) # noisy
m = mix
min_leng = min(y.shape[-1],m.shape[-1])
y = y[...,:min_leng]
m = m[...,:min_leng]
T_orig = y.size(1)
norm_factor = y.abs().max()
y = y / norm_factor
m = m / norm_factor
Y = torch.unsqueeze(model._forward_transform(model._stft(y.cuda())), 0)
Y = pad_spec(Y)
M = torch.unsqueeze(model._forward_transform(model._stft(m.cuda())), 0)
M = pad_spec(M)
timesteps = torch.linspace(reverse_starting_point, 0.03, N, device=Y.device)
std = model.sde._std(reverse_starting_point*torch.ones((Y.shape[0],), device=Y.device))
z = torch.randn_like(Y)
X_t = Y + z * std[:, None, None, None]
t = timesteps[0]
dt = timesteps[-1]
f, g = model.sde.sde(X_t, t, Y)
vec_t = torch.ones(Y.shape[0], device=Y.device) * t
mean_x_tm1 = X_t - (f - g**2*model.forward(X_t, vec_t, Y, M, vec_t[:,None,None,None]))*dt #mean of x t minus 1 = mu(x_{t-1})
sample = mean_x_tm1
sample = sample.squeeze()
x_hat = model.to_audio(sample.squeeze(), T_orig)
x_hat = x_hat * norm_factor
new_norm_factor = x_hat.abs().max()
x_hat = x_hat / new_norm_factor
x_hat = x_hat.unsqueeze(0).cpu()
torchaudio.save(
os.path.join(args.save_folder, savename+'_spk'+str(idx+1)+'_corrected.wav'), x_hat, args.sample_rate
)
def main(args):
os.makedirs(args.save_folder, exist_ok=True)
savename = args.save_name
encoder, masknet, decoder = load_sepformer(args)
fastgeco_model = load_fastgeco(args)
result, mix = separate(args, encoder.cuda(), masknet.cuda(), decoder.cuda(), savename)
correct(args, fastgeco_model, result, mix, savename)
if __name__ == "__main__":
parser = argparse.ArgumentParser()
parser.add_argument('--test_file', type=str, required=True)
parser.add_argument('--save_name', type=str, required=True)
parser.add_argument('--save_folder', type=str, required=True)
parser.add_argument('--ckpt', type=str, required=True)
parser.add_argument('--sample_rate', type=int, default=8000)
parser.add_argument('--num_spks', type=int, default=2)
parser.add_argument("--reverse_starting_point", type=float, default=0.5, help="Starting point for the reverse SDE.")
parser.add_argument("--N", type=int, default=1, help="Number of reverse steps.")
args = parser.parse_args()
main(args)