Refactor: Format Code with Ruff and Update Deprecated G2PW Link (#2255)
* ruff check --fix * ruff format --line-length 120 --target-version py39 * Change the link for G2PW Model * update pytorch version and colab
This commit is contained in:
XXXXRT666
@@ -1,10 +1,9 @@
|
||||
import os
|
||||
import pdb
|
||||
import re
|
||||
|
||||
import cn2an
|
||||
from pypinyin import lazy_pinyin, Style
|
||||
from pypinyin.contrib.tone_convert import to_normal, to_finals_tone3, to_initials, to_finals
|
||||
from pypinyin.contrib.tone_convert import to_finals_tone3, to_initials
|
||||
|
||||
from text.symbols import punctuation
|
||||
from text.tone_sandhi import ToneSandhi
|
||||
@@ -18,18 +17,26 @@ pinyin_to_symbol_map = {
|
||||
for line in open(os.path.join(current_file_path, "opencpop-strict.txt")).readlines()
|
||||
}
|
||||
|
||||
import jieba_fast, logging
|
||||
import jieba_fast
|
||||
import logging
|
||||
|
||||
jieba_fast.setLogLevel(logging.CRITICAL)
|
||||
import jieba_fast.posseg as psg
|
||||
|
||||
# is_g2pw_str = os.environ.get("is_g2pw", "True")##默认开启
|
||||
# is_g2pw = False#True if is_g2pw_str.lower() == 'true' else False
|
||||
is_g2pw = True#True if is_g2pw_str.lower() == 'true' else False
|
||||
is_g2pw = True # True if is_g2pw_str.lower() == 'true' else False
|
||||
if is_g2pw:
|
||||
# print("当前使用g2pw进行拼音推理")
|
||||
from text.g2pw import G2PWPinyin, correct_pronunciation
|
||||
|
||||
parent_directory = os.path.dirname(current_file_path)
|
||||
g2pw = G2PWPinyin(model_dir="GPT_SoVITS/text/G2PWModel",model_source=os.environ.get("bert_path","GPT_SoVITS/pretrained_models/chinese-roberta-wwm-ext-large"),v_to_u=False, neutral_tone_with_five=True)
|
||||
g2pw = G2PWPinyin(
|
||||
model_dir="GPT_SoVITS/text/G2PWModel",
|
||||
model_source=os.environ.get("bert_path", "GPT_SoVITS/pretrained_models/chinese-roberta-wwm-ext-large"),
|
||||
v_to_u=False,
|
||||
neutral_tone_with_five=True,
|
||||
)
|
||||
|
||||
rep_map = {
|
||||
"：": ",",
|
||||
@@ -46,7 +53,7 @@ rep_map = {
|
||||
"/": ",",
|
||||
"—": "-",
|
||||
"~": "…",
|
||||
"～":"…",
|
||||
"～": "…",
|
||||
}
|
||||
|
||||
tone_modifier = ToneSandhi()
|
||||
@@ -58,9 +65,7 @@ def replace_punctuation(text):
|
||||
|
||||
replaced_text = pattern.sub(lambda x: rep_map[x.group()], text)
|
||||
|
||||
replaced_text = re.sub(
|
||||
r"[^\u4e00-\u9fa5" + "".join(punctuation) + r"]+", "", replaced_text
|
||||
)
|
||||
replaced_text = re.sub(r"[^\u4e00-\u9fa5" + "".join(punctuation) + r"]+", "", replaced_text)
|
||||
|
||||
return replaced_text
|
||||
|
||||
@@ -77,9 +82,7 @@ def _get_initials_finals(word):
|
||||
finals = []
|
||||
|
||||
orig_initials = lazy_pinyin(word, neutral_tone_with_five=True, style=Style.INITIALS)
|
||||
orig_finals = lazy_pinyin(
|
||||
word, neutral_tone_with_five=True, style=Style.FINALS_TONE3
|
||||
)
|
||||
orig_finals = lazy_pinyin(word, neutral_tone_with_five=True, style=Style.FINALS_TONE3)
|
||||
|
||||
for c, v in zip(orig_initials, orig_finals):
|
||||
initials.append(c)
|
||||
@@ -87,31 +90,66 @@ def _get_initials_finals(word):
|
||||
return initials, finals
|
||||
|
||||
|
||||
must_erhua = {
|
||||
"小院儿", "胡同儿", "范儿", "老汉儿", "撒欢儿", "寻老礼儿", "妥妥儿", "媳妇儿"
|
||||
}
|
||||
must_erhua = {"小院儿", "胡同儿", "范儿", "老汉儿", "撒欢儿", "寻老礼儿", "妥妥儿", "媳妇儿"}
|
||||
not_erhua = {
|
||||
"虐儿", "为儿", "护儿", "瞒儿", "救儿", "替儿", "有儿", "一儿", "我儿", "俺儿", "妻儿",
|
||||
"拐儿", "聋儿", "乞儿", "患儿", "幼儿", "孤儿", "婴儿", "婴幼儿", "连体儿", "脑瘫儿",
|
||||
"流浪儿", "体弱儿", "混血儿", "蜜雪儿", "舫儿", "祖儿", "美儿", "应采儿", "可儿", "侄儿",
|
||||
"孙儿", "侄孙儿", "女儿", "男儿", "红孩儿", "花儿", "虫儿", "马儿", "鸟儿", "猪儿", "猫儿",
|
||||
"狗儿", "少儿"
|
||||
"虐儿",
|
||||
"为儿",
|
||||
"护儿",
|
||||
"瞒儿",
|
||||
"救儿",
|
||||
"替儿",
|
||||
"有儿",
|
||||
"一儿",
|
||||
"我儿",
|
||||
"俺儿",
|
||||
"妻儿",
|
||||
"拐儿",
|
||||
"聋儿",
|
||||
"乞儿",
|
||||
"患儿",
|
||||
"幼儿",
|
||||
"孤儿",
|
||||
"婴儿",
|
||||
"婴幼儿",
|
||||
"连体儿",
|
||||
"脑瘫儿",
|
||||
"流浪儿",
|
||||
"体弱儿",
|
||||
"混血儿",
|
||||
"蜜雪儿",
|
||||
"舫儿",
|
||||
"祖儿",
|
||||
"美儿",
|
||||
"应采儿",
|
||||
"可儿",
|
||||
"侄儿",
|
||||
"孙儿",
|
||||
"侄孙儿",
|
||||
"女儿",
|
||||
"男儿",
|
||||
"红孩儿",
|
||||
"花儿",
|
||||
"虫儿",
|
||||
"马儿",
|
||||
"鸟儿",
|
||||
"猪儿",
|
||||
"猫儿",
|
||||
"狗儿",
|
||||
"少儿",
|
||||
}
|
||||
def _merge_erhua(initials: list[str],
|
||||
finals: list[str],
|
||||
word: str,
|
||||
pos: str) -> list[list[str]]:
|
||||
|
||||
|
||||
def _merge_erhua(initials: list[str], finals: list[str], word: str, pos: str) -> list[list[str]]:
|
||||
"""
|
||||
Do erhub.
|
||||
"""
|
||||
# fix er1
|
||||
for i, phn in enumerate(finals):
|
||||
if i == len(finals) - 1 and word[i] == "儿" and phn == 'er1':
|
||||
finals[i] = 'er2'
|
||||
if i == len(finals) - 1 and word[i] == "儿" and phn == "er1":
|
||||
finals[i] = "er2"
|
||||
|
||||
# 发音
|
||||
if word not in must_erhua and (word in not_erhua or
|
||||
pos in {"a", "j", "nr"}):
|
||||
if word not in must_erhua and (word in not_erhua or pos in {"a", "j", "nr"}):
|
||||
return initials, finals
|
||||
|
||||
# "……" 等情况直接返回
|
||||
@@ -124,9 +162,13 @@ def _merge_erhua(initials: list[str],
|
||||
new_initials = []
|
||||
new_finals = []
|
||||
for i, phn in enumerate(finals):
|
||||
if i == len(finals) - 1 and word[i] == "儿" and phn in {
|
||||
"er2", "er5"
|
||||
} and word[-2:] not in not_erhua and new_finals:
|
||||
if (
|
||||
i == len(finals) - 1
|
||||
and word[i] == "儿"
|
||||
and phn in {"er2", "er5"}
|
||||
and word[-2:] not in not_erhua
|
||||
and new_finals
|
||||
):
|
||||
phn = "er" + new_finals[-1][-1]
|
||||
|
||||
new_initials.append(initials[i])
|
||||
@@ -160,7 +202,7 @@ def _g2p(segments):
|
||||
# assert len(sub_initials) == len(sub_finals) == len(word)
|
||||
initials = sum(initials, [])
|
||||
finals = sum(finals, [])
|
||||
print("pypinyin结果",initials,finals)
|
||||
print("pypinyin结果", initials, finals)
|
||||
else:
|
||||
# g2pw采用整句推理
|
||||
pinyins = g2pw.lazy_pinyin(seg, neutral_tone_with_five=True, style=Style.TONE3)
|
||||
@@ -171,19 +213,19 @@ def _g2p(segments):
|
||||
sub_finals = []
|
||||
now_word_length = pre_word_length + len(word)
|
||||
|
||||
if pos == 'eng':
|
||||
if pos == "eng":
|
||||
pre_word_length = now_word_length
|
||||
continue
|
||||
|
||||
word_pinyins = pinyins[pre_word_length:now_word_length]
|
||||
|
||||
# 多音字消歧
|
||||
word_pinyins = correct_pronunciation(word,word_pinyins)
|
||||
word_pinyins = correct_pronunciation(word, word_pinyins)
|
||||
|
||||
for pinyin in word_pinyins:
|
||||
if pinyin[0].isalpha():
|
||||
sub_initials.append(to_initials(pinyin))
|
||||
sub_finals.append(to_finals_tone3(pinyin,neutral_tone_with_five=True))
|
||||
sub_finals.append(to_finals_tone3(pinyin, neutral_tone_with_five=True))
|
||||
else:
|
||||
sub_initials.append(pinyin)
|
||||
sub_finals.append(pinyin)
|
||||
@@ -259,18 +301,18 @@ def replace_punctuation_with_en(text):
|
||||
|
||||
replaced_text = pattern.sub(lambda x: rep_map[x.group()], text)
|
||||
|
||||
replaced_text = re.sub(
|
||||
r"[^\u4e00-\u9fa5A-Za-z" + "".join(punctuation) + r"]+", "", replaced_text
|
||||
)
|
||||
replaced_text = re.sub(r"[^\u4e00-\u9fa5A-Za-z" + "".join(punctuation) + r"]+", "", replaced_text)
|
||||
|
||||
return replaced_text
|
||||
|
||||
|
||||
def replace_consecutive_punctuation(text):
|
||||
punctuations = ''.join(re.escape(p) for p in punctuation)
|
||||
pattern = f'([{punctuations}])([{punctuations}])+'
|
||||
result = re.sub(pattern, r'\1', text)
|
||||
punctuations = "".join(re.escape(p) for p in punctuation)
|
||||
pattern = f"([{punctuations}])([{punctuations}])+"
|
||||
result = re.sub(pattern, r"\1", text)
|
||||
return result
|
||||
|
||||
|
||||
def text_normalize(text):
|
||||
# https://github.com/PaddlePaddle/PaddleSpeech/tree/develop/paddlespeech/t2s/frontend/zh_normalization
|
||||
tx = TextNormalizer()
|
||||
@@ -283,6 +325,7 @@ def text_normalize(text):
|
||||
dest_text = replace_consecutive_punctuation(dest_text)
|
||||
return dest_text
|
||||
|
||||
|
||||
# 不排除英文的文本格式化
|
||||
def mix_text_normalize(text):
|
||||
# https://github.com/PaddlePaddle/PaddleSpeech/tree/develop/paddlespeech/t2s/frontend/zh_normalization
|
||||
|
||||
Reference in New Issue
Block a user