Delete infer_onnx.py
Browse files- infer_onnx.py +0 -97
infer_onnx.py
DELETED
@@ -1,97 +0,0 @@
|
|
1 |
-
import os
|
2 |
-
import re
|
3 |
-
import onnxruntime
|
4 |
-
import numpy as np
|
5 |
-
from huggingface_hub import snapshot_download
|
6 |
-
from gruut import sentences
|
7 |
-
import numpy as np
|
8 |
-
import scipy.io.wavfile
|
9 |
-
|
10 |
-
|
11 |
-
|
12 |
-
class TTS:
|
13 |
-
def __init__(self, model_name: str, save_path: str = "./model", add_time_to_end: float = 0.8) -> None:
|
14 |
-
if not os.path.exists(save_path):
|
15 |
-
os.mkdir(save_path)
|
16 |
-
|
17 |
-
model_dir = os.path.join(save_path, model_name)
|
18 |
-
|
19 |
-
if not os.path.exists(model_dir):
|
20 |
-
snapshot_download(repo_id=model_name,
|
21 |
-
allow_patterns=["*.txt", "*.onnx"],
|
22 |
-
local_dir=model_dir,
|
23 |
-
local_dir_use_symlinks=False
|
24 |
-
)
|
25 |
-
|
26 |
-
sess_options = onnxruntime.SessionOptions()
|
27 |
-
self.model = onnxruntime.InferenceSession(os.path.join(model_dir, "exported/model.onnx"), sess_options=sess_options)
|
28 |
-
|
29 |
-
with open(os.path.join(model_dir, "exported/vocab.txt"), "r", encoding="utf-8") as vocab_file:
|
30 |
-
self.symbols = vocab_file.read().split("\n")
|
31 |
-
self.symbols = list(map(chr, list(map(int, self.symbols))))
|
32 |
-
|
33 |
-
self.symbol_to_id = {s: i for i, s in enumerate(self.symbols)}
|
34 |
-
self.add_time_to_end = add_time_to_end
|
35 |
-
|
36 |
-
|
37 |
-
def _ru_phonems(self, text: str) -> str:
|
38 |
-
text = text.lower()
|
39 |
-
phonemes = ""
|
40 |
-
for sent in sentences(text, lang="ru"):
|
41 |
-
for word in sent:
|
42 |
-
if word.phonemes:
|
43 |
-
phonemes += "".join(word.phonemes)
|
44 |
-
phonemes = re.sub(re.compile(r'\s+'), ' ', phonemes).lstrip().rstrip()
|
45 |
-
return phonemes
|
46 |
-
|
47 |
-
|
48 |
-
def _text_to_sequence(self, text: str) -> list[int]:
|
49 |
-
'''convert text to seq'''
|
50 |
-
sequence = []
|
51 |
-
clean_text = self._ru_phonems(text)
|
52 |
-
for symbol in clean_text:
|
53 |
-
symbol_id = self.symbol_to_id[symbol]
|
54 |
-
sequence += [symbol_id]
|
55 |
-
return sequence
|
56 |
-
|
57 |
-
|
58 |
-
def _intersperse(self, lst, item):
|
59 |
-
result = [item] * (len(lst) * 2 + 1)
|
60 |
-
result[1::2] = lst
|
61 |
-
return result
|
62 |
-
|
63 |
-
|
64 |
-
def _get_text(self, text: str) -> list[int]:
|
65 |
-
text_norm = self._text_to_sequence(text)
|
66 |
-
text_norm = self._intersperse(text_norm, 0)
|
67 |
-
return text_norm
|
68 |
-
|
69 |
-
def _add_silent(self, audio, silence_duration: float = 0.7, sample_rate: int = 22050):
|
70 |
-
num_samples_silence = int(sample_rate * silence_duration)
|
71 |
-
silence_array = np.zeros(num_samples_silence, dtype=np.float32)
|
72 |
-
audio_with_silence = np.concatenate((audio, silence_array), axis=0)
|
73 |
-
return audio_with_silence
|
74 |
-
|
75 |
-
def save_wav(self, audio, path:str):
|
76 |
-
'''save audio to wav'''
|
77 |
-
scipy.io.wavfile.write(path, 22050, audio)
|
78 |
-
|
79 |
-
def __call__(self, text: str, play = False):
|
80 |
-
phoneme_ids = self._get_text(text)
|
81 |
-
text = np.expand_dims(np.array(phoneme_ids, dtype=np.int64), 0)
|
82 |
-
text_lengths = np.array([text.shape[1]], dtype=np.int64)
|
83 |
-
scales = np.array(
|
84 |
-
[0.667, 1, 0.8],
|
85 |
-
dtype=np.float32,
|
86 |
-
)
|
87 |
-
audio = self.model.run(
|
88 |
-
None,
|
89 |
-
{
|
90 |
-
"input": text,
|
91 |
-
"input_lengths": text_lengths,
|
92 |
-
"scales": scales,
|
93 |
-
"sid": None,
|
94 |
-
},
|
95 |
-
)[0][0,0][0]
|
96 |
-
audio = self._add_silent(audio)
|
97 |
-
return audio
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|