-
Notifications
You must be signed in to change notification settings - Fork 5
/
snap.py
66 lines (57 loc) · 2.06 KB
/
snap.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
import torch
import torchaudio
import numpy as np
from IPython.display import display,Audio
import matplotlib.pyplot as plt
from torch import nn
from transformers import Wav2Vec2FeatureExtractor, WavLMForXVector
class CFG:
device='cpu'
RATE=16000
percent=0.9
batch_size=1
epoches=80
lr=1e-6
load_pth='pth/best.pth'
debug=114514
save_steps=20
length=15360
def Load_Snap(filename = "./audios/snaps/snap0.wav"):
waveform,sample_rate = torchaudio.load(filename)
# print("Shape of waveform:{}".format(waveform.size())) #音频大小
# print("sample rate of waveform:{}".format(sample_rate))#采样率
resampler = torchaudio.transforms.Resample(sample_rate, CFG.RATE)
if(sample_rate!=CFG.RATE):
waveform=resampler(waveform)
for channel in waveform:
return channel
feature_extractor = Wav2Vec2FeatureExtractor.from_pretrained("microsoft/wavlm-base-plus")
class Snap(nn.Module):
def __init__(self):
super(Snap,self).__init__()
sample_snap=Load_Snap('audios/snaps/snap0.wav')
self.sample_inputs=feature_extractor(sample_snap, padding=True, return_tensors="pt",sampling_rate=CFG.RATE).to(CFG.device)
# print(self.sample_inputs['input_values'].shape)
def forward(self, x):
x=x.reshape(CFG.batch_size,-1).cpu().numpy()
tmp=[]
for a in x:
tmp.append(a)
x=feature_extractor(tmp, padding=True, return_tensors="pt",sampling_rate=CFG.RATE).to(CFG.device)
self.sample=pretrained(**self.sample_inputs).embeddings[0]
x=pretrained(**x).embeddings
cosine_sim = torch.nn.CosineSimilarity(dim=-1)
similarity = cosine_sim(self.sample, x)
return similarity
model=Snap()
model=model.to(CFG.device)
def Load(pth=CFG.load_pth):
global pretrained
pretrained=WavLMForXVector.from_pretrained(pth)
def is_Snap(filename):
x=Load_Snap(filename)
y=model(x)
return y
if __name__=='__main__':
Load('pth/best_pth')
print(is_Snap('tmp/tmp.wav'))