Spaces:
Running
on
Zero
Running
on
Zero
zhzluke96
commited on
Commit
•
f367757
1
Parent(s):
bb4ceb3
update
Browse files- modules/SentenceSplitter.py +75 -67
- modules/models.py +7 -0
- modules/utils/audio.py +1 -1
- modules/utils/html.py +18 -2
- modules/webui/ssml/podcast_tab.py +15 -3
- modules/webui/ssml/spliter_tab.py +36 -8
- modules/webui/webui_utils.py +5 -4
- requirements.txt +2 -1
- webui.py +0 -8
modules/SentenceSplitter.py
CHANGED
@@ -2,87 +2,95 @@ import re
|
|
2 |
|
3 |
import zhon
|
4 |
|
|
|
5 |
from modules.utils.detect_lang import guess_lang
|
6 |
|
7 |
|
8 |
-
|
9 |
-
|
10 |
-
|
11 |
-
|
12 |
-
for match in pattern.finditer(text):
|
13 |
-
# 获取匹配的中文句子
|
14 |
-
end = match.end()
|
15 |
-
result.append(text[start:end])
|
16 |
-
start = end
|
17 |
-
|
18 |
-
# 最后一个中文句子后面的内容(如果有)也需要添加到结果中
|
19 |
-
if start < len(text):
|
20 |
-
result.append(text[start:])
|
21 |
-
|
22 |
-
result = [t for t in result if t.strip()]
|
23 |
-
return result
|
24 |
-
|
25 |
-
|
26 |
-
def split_en_sentence(text):
|
27 |
-
"""
|
28 |
-
Split English text into sentences.
|
29 |
-
"""
|
30 |
-
# Define a regex pattern for English sentence splitting
|
31 |
-
pattern = re.compile(r"(?<!\w\.\w.)(?<![A-Z][a-z]\.)(?<=\.|\?|\!)\s")
|
32 |
-
result = pattern.split(text)
|
33 |
-
|
34 |
-
# Filter out any empty strings or strings that are just whitespace
|
35 |
-
result = [sentence.strip() for sentence in result if sentence.strip()]
|
36 |
-
|
37 |
-
return result
|
38 |
|
|
|
|
|
|
|
|
|
39 |
|
40 |
-
|
41 |
-
|
42 |
|
|
|
|
|
43 |
|
44 |
-
def
|
45 |
-
|
46 |
-
|
47 |
-
for line in lines:
|
48 |
-
if is_eng_sentence(line):
|
49 |
-
result.extend(split_en_sentence(line))
|
50 |
-
else:
|
51 |
-
result.extend(split_zhon_sentence(line))
|
52 |
-
return result
|
53 |
|
|
|
54 |
|
55 |
-
|
56 |
-
|
57 |
-
|
58 |
-
def __init__(self, threshold=100):
|
59 |
-
self.sentence_threshold = threshold
|
60 |
|
61 |
-
|
62 |
-
|
63 |
-
|
64 |
-
|
65 |
-
|
66 |
-
|
67 |
-
|
68 |
-
if len(sentence) < self.sentence_threshold:
|
69 |
-
temp_sentence.extend(sentence)
|
70 |
-
if len(temp_sentence) >= self.sentence_threshold:
|
71 |
-
merged_sentences.append(temp_sentence)
|
72 |
-
temp_sentence = []
|
73 |
else:
|
74 |
-
|
75 |
-
|
76 |
-
temp_sentence = []
|
77 |
-
merged_sentences.append(sentence)
|
78 |
|
79 |
if temp_sentence:
|
80 |
merged_sentences.append(temp_sentence)
|
81 |
-
|
82 |
-
|
83 |
-
|
84 |
-
|
85 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
86 |
|
87 |
|
88 |
if __name__ == "__main__":
|
|
|
2 |
|
3 |
import zhon
|
4 |
|
5 |
+
from modules.models import get_tokenizer
|
6 |
from modules.utils.detect_lang import guess_lang
|
7 |
|
8 |
|
9 |
+
# 解析文本 并根据停止符号分割成句子
|
10 |
+
# 可以设置最大阈值,即如果分割片段小于这个阈值会与下一段合并
|
11 |
+
class SentenceSplitter:
|
12 |
+
SEP_TOKEN = " "
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
13 |
|
14 |
+
def __init__(self, threshold=100):
|
15 |
+
assert (
|
16 |
+
isinstance(threshold, int) and threshold > 0
|
17 |
+
), "Threshold must be greater than 0."
|
18 |
|
19 |
+
self.sentence_threshold = threshold
|
20 |
+
self.tokenizer = get_tokenizer()
|
21 |
|
22 |
+
def count_tokens(self, text: str):
|
23 |
+
return len(self.tokenizer.tokenize(text))
|
24 |
|
25 |
+
def parse(self, text: str):
|
26 |
+
sentences = self.split_paragraph(text)
|
27 |
+
sentences = self.merge_text_by_threshold(sentences)
|
|
|
|
|
|
|
|
|
|
|
|
|
28 |
|
29 |
+
return sentences
|
30 |
|
31 |
+
def merge_text_by_threshold(self, setences: list[str]):
|
32 |
+
"""
|
33 |
+
Merge text by threshold.
|
|
|
|
|
34 |
|
35 |
+
If the length of the text is less than the threshold, merge it with the next text.
|
36 |
+
"""
|
37 |
+
merged_sentences: list[str] = []
|
38 |
+
temp_sentence = ""
|
39 |
+
for sentence in setences:
|
40 |
+
if len(temp_sentence) + len(sentence) < self.sentence_threshold:
|
41 |
+
temp_sentence += SentenceSplitter.SEP_TOKEN + sentence
|
|
|
|
|
|
|
|
|
|
|
42 |
else:
|
43 |
+
merged_sentences.append(temp_sentence)
|
44 |
+
temp_sentence = sentence
|
|
|
|
|
45 |
|
46 |
if temp_sentence:
|
47 |
merged_sentences.append(temp_sentence)
|
48 |
+
return merged_sentences
|
49 |
+
|
50 |
+
def split_paragraph(self, text: str):
|
51 |
+
"""
|
52 |
+
Split text into sentences.
|
53 |
+
"""
|
54 |
+
lines = text.split("\n")
|
55 |
+
sentences: list[str] = []
|
56 |
+
for line in lines:
|
57 |
+
if self.is_eng_sentence(line):
|
58 |
+
sentences.extend(self.split_en_sentence(line))
|
59 |
+
else:
|
60 |
+
sentences.extend(self.split_zhon_sentence(line))
|
61 |
+
return sentences
|
62 |
+
|
63 |
+
def is_eng_sentence(self, text: str):
|
64 |
+
return guess_lang(text) == "en"
|
65 |
+
|
66 |
+
def split_en_sentence(self, text: str):
|
67 |
+
"""
|
68 |
+
Split English text into sentences.
|
69 |
+
"""
|
70 |
+
pattern = re.compile(r"(?<!\w\.\w.)(?<![A-Z][a-z]\.)(?<=\.|\?|\!)\s")
|
71 |
+
sentences = pattern.split(text)
|
72 |
+
|
73 |
+
sentences = [sentence.strip() for sentence in sentences if sentence.strip()]
|
74 |
+
|
75 |
+
return sentences
|
76 |
+
|
77 |
+
def split_zhon_sentence(self, text: str):
|
78 |
+
"""
|
79 |
+
Split Chinese text into sentences.
|
80 |
+
"""
|
81 |
+
sentences: list[str] = []
|
82 |
+
pattern = re.compile(zhon.hanzi.sentence)
|
83 |
+
start = 0
|
84 |
+
for match in pattern.finditer(text):
|
85 |
+
end = match.end()
|
86 |
+
sentences.append(text[start:end])
|
87 |
+
start = end
|
88 |
+
|
89 |
+
if start < len(text):
|
90 |
+
sentences.append(text[start:])
|
91 |
+
|
92 |
+
sentences = [t for t in sentences if t.strip()]
|
93 |
+
return sentences
|
94 |
|
95 |
|
96 |
if __name__ == "__main__":
|
modules/models.py
CHANGED
@@ -3,6 +3,7 @@ import logging
|
|
3 |
import threading
|
4 |
|
5 |
import torch
|
|
|
6 |
|
7 |
from modules import config
|
8 |
from modules.ChatTTS import ChatTTS
|
@@ -76,3 +77,9 @@ def reload_chat_tts():
|
|
76 |
instance = load_chat_tts()
|
77 |
logger.info("ChatTTS models reloaded")
|
78 |
return instance
|
|
|
|
|
|
|
|
|
|
|
|
|
|
3 |
import threading
|
4 |
|
5 |
import torch
|
6 |
+
from transformers import LlamaTokenizer
|
7 |
|
8 |
from modules import config
|
9 |
from modules.ChatTTS import ChatTTS
|
|
|
77 |
instance = load_chat_tts()
|
78 |
logger.info("ChatTTS models reloaded")
|
79 |
return instance
|
80 |
+
|
81 |
+
|
82 |
+
def get_tokenizer() -> LlamaTokenizer:
|
83 |
+
chat_tts = load_chat_tts()
|
84 |
+
tokenizer = chat_tts.pretrain_models["tokenizer"]
|
85 |
+
return tokenizer
|
modules/utils/audio.py
CHANGED
@@ -2,9 +2,9 @@ import sys
|
|
2 |
from io import BytesIO
|
3 |
|
4 |
import numpy as np
|
|
|
5 |
import soundfile as sf
|
6 |
from pydub import AudioSegment, effects
|
7 |
-
import pyrubberband as pyrb
|
8 |
|
9 |
INT16_MAX = np.iinfo(np.int16).max
|
10 |
|
|
|
2 |
from io import BytesIO
|
3 |
|
4 |
import numpy as np
|
5 |
+
import pyrubberband as pyrb
|
6 |
import soundfile as sf
|
7 |
from pydub import AudioSegment, effects
|
|
|
8 |
|
9 |
INT16_MAX = np.iinfo(np.int16).max
|
10 |
|
modules/utils/html.py
CHANGED
@@ -1,6 +1,10 @@
|
|
|
|
|
|
1 |
from html.parser import HTMLParser
|
2 |
|
3 |
|
|
|
|
|
4 |
class HTMLTagRemover(HTMLParser):
|
5 |
def __init__(self):
|
6 |
super().__init__()
|
@@ -20,7 +24,19 @@ def remove_html_tags(text):
|
|
20 |
return parser.get_data()
|
21 |
|
22 |
|
|
|
|
|
|
|
|
|
|
|
|
|
23 |
if __name__ == "__main__":
|
24 |
-
input_text = "
|
25 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
26 |
print(output_text) # 输出: 一个标题 这是一段包含标签的文本。
|
|
|
1 |
+
import html
|
2 |
+
import re
|
3 |
from html.parser import HTMLParser
|
4 |
|
5 |
|
6 |
+
# NOTE: 现在没用这个,因为不好解决转义字符的问题
|
7 |
+
# 除非分段处理,但是太麻烦了...
|
8 |
class HTMLTagRemover(HTMLParser):
|
9 |
def __init__(self):
|
10 |
super().__init__()
|
|
|
24 |
return parser.get_data()
|
25 |
|
26 |
|
27 |
+
def remove_html_tags_re(text):
|
28 |
+
text = html.unescape(text)
|
29 |
+
html_tags_pattern = re.compile(r"</?([a-zA-Z1-9]+)[^>]*>")
|
30 |
+
return re.sub(html_tags_pattern, " ", text)
|
31 |
+
|
32 |
+
|
33 |
if __name__ == "__main__":
|
34 |
+
input_text = """
|
35 |
+
<h1>一个标题</h1> 这是一段包含<code>标签</code>的文本。 <code>&</code>
|
36 |
+
<设定>
|
37 |
+
一些文本
|
38 |
+
</设定>
|
39 |
+
"""
|
40 |
+
# input_text = "我&你"
|
41 |
+
output_text = remove_html_tags_re(input_text)
|
42 |
print(output_text) # 输出: 一个标题 这是一段包含标签的文本。
|
modules/webui/ssml/podcast_tab.py
CHANGED
@@ -19,13 +19,18 @@ def merge_dataframe_to_ssml(msg, spk, style, df: pd.DataFrame):
|
|
19 |
spk = row.get("speaker")
|
20 |
style = row.get("style")
|
21 |
|
|
|
|
|
|
|
|
|
|
|
22 |
ssml += f"{indent}<voice"
|
23 |
if spk:
|
24 |
ssml += f' spk="{spk}"'
|
25 |
if style:
|
26 |
ssml += f' style="{style}"'
|
27 |
ssml += ">\n"
|
28 |
-
ssml += f"{indent}{indent}{
|
29 |
ssml += f"{indent}</voice>\n"
|
30 |
# 原封不动输出回去是为了触发 loadding 效果
|
31 |
return msg, spk, style, f"<speak version='0.1'>\n{ssml}</speak>"
|
@@ -42,6 +47,7 @@ def create_ssml_podcast_tab(ssml_input: gr.Textbox, tabs1: gr.Tabs, tabs2: gr.Ta
|
|
42 |
with gr.Row():
|
43 |
with gr.Column(scale=1):
|
44 |
with gr.Group():
|
|
|
45 |
spk_input_dropdown = gr.Dropdown(
|
46 |
choices=get_spk_choices(),
|
47 |
interactive=True,
|
@@ -55,13 +61,19 @@ def create_ssml_podcast_tab(ssml_input: gr.Textbox, tabs1: gr.Tabs, tabs2: gr.Ta
|
|
55 |
show_label=False,
|
56 |
value="*auto",
|
57 |
)
|
|
|
58 |
with gr.Group():
|
|
|
59 |
msg = gr.Textbox(
|
60 |
-
lines=5,
|
|
|
|
|
|
|
61 |
)
|
62 |
add = gr.Button("Add")
|
63 |
undo = gr.Button("Undo")
|
64 |
clear = gr.Button("Clear")
|
|
|
65 |
with gr.Column(scale=5):
|
66 |
with gr.Group():
|
67 |
gr.Markdown("📔Script")
|
@@ -75,7 +87,7 @@ def create_ssml_podcast_tab(ssml_input: gr.Textbox, tabs1: gr.Tabs, tabs2: gr.Ta
|
|
75 |
col_count=(4, "fixed"),
|
76 |
)
|
77 |
|
78 |
-
|
79 |
|
80 |
def add_message(msg, spk, style, sheet: pd.DataFrame):
|
81 |
if not msg:
|
|
|
19 |
spk = row.get("speaker")
|
20 |
style = row.get("style")
|
21 |
|
22 |
+
text = text_normalize(text)
|
23 |
+
|
24 |
+
if text.strip() == "":
|
25 |
+
continue
|
26 |
+
|
27 |
ssml += f"{indent}<voice"
|
28 |
if spk:
|
29 |
ssml += f' spk="{spk}"'
|
30 |
if style:
|
31 |
ssml += f' style="{style}"'
|
32 |
ssml += ">\n"
|
33 |
+
ssml += f"{indent}{indent}{text}\n"
|
34 |
ssml += f"{indent}</voice>\n"
|
35 |
# 原封不动输出回去是为了触发 loadding 效果
|
36 |
return msg, spk, style, f"<speak version='0.1'>\n{ssml}</speak>"
|
|
|
47 |
with gr.Row():
|
48 |
with gr.Column(scale=1):
|
49 |
with gr.Group():
|
50 |
+
gr.Markdown("🗣️Speaker")
|
51 |
spk_input_dropdown = gr.Dropdown(
|
52 |
choices=get_spk_choices(),
|
53 |
interactive=True,
|
|
|
61 |
show_label=False,
|
62 |
value="*auto",
|
63 |
)
|
64 |
+
|
65 |
with gr.Group():
|
66 |
+
gr.Markdown("📝Text Input")
|
67 |
msg = gr.Textbox(
|
68 |
+
lines=5,
|
69 |
+
label="Message",
|
70 |
+
show_label=False,
|
71 |
+
placeholder="Type speaker message here",
|
72 |
)
|
73 |
add = gr.Button("Add")
|
74 |
undo = gr.Button("Undo")
|
75 |
clear = gr.Button("Clear")
|
76 |
+
|
77 |
with gr.Column(scale=5):
|
78 |
with gr.Group():
|
79 |
gr.Markdown("📔Script")
|
|
|
87 |
col_count=(4, "fixed"),
|
88 |
)
|
89 |
|
90 |
+
send_to_ssml_btn = gr.Button("📩Send to SSML", variant="primary")
|
91 |
|
92 |
def add_message(msg, spk, style, sheet: pd.DataFrame):
|
93 |
if not msg:
|
modules/webui/ssml/spliter_tab.py
CHANGED
@@ -22,6 +22,12 @@ def merge_dataframe_to_ssml(dataframe, spk, style, seed):
|
|
22 |
indent = " " * 2
|
23 |
|
24 |
for i, row in dataframe.iterrows():
|
|
|
|
|
|
|
|
|
|
|
|
|
25 |
ssml += f"{indent}<voice"
|
26 |
if spk:
|
27 |
ssml += f' spk="{spk}"'
|
@@ -30,7 +36,7 @@ def merge_dataframe_to_ssml(dataframe, spk, style, seed):
|
|
30 |
if seed:
|
31 |
ssml += f' seed="{seed}"'
|
32 |
ssml += ">\n"
|
33 |
-
ssml += f"{indent}{indent}{
|
34 |
ssml += f"{indent}</voice>\n"
|
35 |
# 原封不动输出回去是为了触发 loadding 效果
|
36 |
return dataframe, spk, style, seed, f"<speak version='0.1'>\n{ssml}</speak>"
|
@@ -73,8 +79,9 @@ def create_spliter_tab(ssml_input, tabs1, tabs2):
|
|
73 |
show_label=False,
|
74 |
value="*auto",
|
75 |
)
|
|
|
76 |
with gr.Group():
|
77 |
-
gr.Markdown("
|
78 |
infer_seed_input = gr.Number(
|
79 |
value=42,
|
80 |
label="Inference Seed",
|
@@ -84,10 +91,23 @@ def create_spliter_tab(ssml_input, tabs1, tabs2):
|
|
84 |
)
|
85 |
infer_seed_rand_button = gr.Button(
|
86 |
value="🎲",
|
|
|
87 |
variant="secondary",
|
88 |
)
|
89 |
|
90 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
91 |
|
92 |
with gr.Column(scale=3):
|
93 |
with gr.Group():
|
@@ -102,19 +122,21 @@ def create_spliter_tab(ssml_input, tabs1, tabs2):
|
|
102 |
)
|
103 |
long_text_split_button = gr.Button("🔪Split Text")
|
104 |
|
105 |
-
with gr.Row():
|
106 |
-
with gr.Column(scale=3):
|
107 |
with gr.Group():
|
108 |
gr.Markdown("🎨Output")
|
109 |
long_text_output = gr.DataFrame(
|
110 |
headers=["index", "text", "length"],
|
111 |
datatype=["number", "str", "number"],
|
112 |
elem_id="long-text-output",
|
113 |
-
interactive=
|
114 |
wrap=True,
|
115 |
value=[],
|
|
|
|
|
116 |
)
|
117 |
|
|
|
|
|
118 |
spk_input_dropdown.change(
|
119 |
fn=lambda x: x.startswith("*") and "-1" or x.split(":")[-1].strip(),
|
120 |
inputs=[spk_input_dropdown],
|
@@ -132,8 +154,14 @@ def create_spliter_tab(ssml_input, tabs1, tabs2):
|
|
132 |
)
|
133 |
long_text_split_button.click(
|
134 |
split_long_text,
|
135 |
-
inputs=[
|
136 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
137 |
)
|
138 |
|
139 |
infer_seed_rand_button.click(
|
|
|
22 |
indent = " " * 2
|
23 |
|
24 |
for i, row in dataframe.iterrows():
|
25 |
+
text = row.iloc[1]
|
26 |
+
text = text_normalize(text)
|
27 |
+
|
28 |
+
if text.strip() == "":
|
29 |
+
continue
|
30 |
+
|
31 |
ssml += f"{indent}<voice"
|
32 |
if spk:
|
33 |
ssml += f' spk="{spk}"'
|
|
|
36 |
if seed:
|
37 |
ssml += f' seed="{seed}"'
|
38 |
ssml += ">\n"
|
39 |
+
ssml += f"{indent}{indent}{text}\n"
|
40 |
ssml += f"{indent}</voice>\n"
|
41 |
# 原封不动输出回去是为了触发 loadding 效果
|
42 |
return dataframe, spk, style, seed, f"<speak version='0.1'>\n{ssml}</speak>"
|
|
|
79 |
show_label=False,
|
80 |
value="*auto",
|
81 |
)
|
82 |
+
|
83 |
with gr.Group():
|
84 |
+
gr.Markdown("💃Inference Seed")
|
85 |
infer_seed_input = gr.Number(
|
86 |
value=42,
|
87 |
label="Inference Seed",
|
|
|
91 |
)
|
92 |
infer_seed_rand_button = gr.Button(
|
93 |
value="🎲",
|
94 |
+
# tooltip="Random Seed",
|
95 |
variant="secondary",
|
96 |
)
|
97 |
|
98 |
+
with gr.Group():
|
99 |
+
gr.Markdown("🎛️Spliter")
|
100 |
+
eos_input = gr.Textbox(
|
101 |
+
label="eos",
|
102 |
+
value="[uv_break]",
|
103 |
+
)
|
104 |
+
spliter_thr_input = gr.Slider(
|
105 |
+
label="Spliter Threshold",
|
106 |
+
value=100,
|
107 |
+
minimum=50,
|
108 |
+
maximum=1000,
|
109 |
+
step=1,
|
110 |
+
)
|
111 |
|
112 |
with gr.Column(scale=3):
|
113 |
with gr.Group():
|
|
|
122 |
)
|
123 |
long_text_split_button = gr.Button("🔪Split Text")
|
124 |
|
|
|
|
|
125 |
with gr.Group():
|
126 |
gr.Markdown("🎨Output")
|
127 |
long_text_output = gr.DataFrame(
|
128 |
headers=["index", "text", "length"],
|
129 |
datatype=["number", "str", "number"],
|
130 |
elem_id="long-text-output",
|
131 |
+
interactive=True,
|
132 |
wrap=True,
|
133 |
value=[],
|
134 |
+
row_count=(0, "dynamic"),
|
135 |
+
col_count=(3, "fixed"),
|
136 |
)
|
137 |
|
138 |
+
send_btn = gr.Button("📩Send to SSML", variant="primary")
|
139 |
+
|
140 |
spk_input_dropdown.change(
|
141 |
fn=lambda x: x.startswith("*") and "-1" or x.split(":")[-1].strip(),
|
142 |
inputs=[spk_input_dropdown],
|
|
|
154 |
)
|
155 |
long_text_split_button.click(
|
156 |
split_long_text,
|
157 |
+
inputs=[
|
158 |
+
long_text_input,
|
159 |
+
spliter_thr_input,
|
160 |
+
eos_input,
|
161 |
+
],
|
162 |
+
outputs=[
|
163 |
+
long_text_output,
|
164 |
+
],
|
165 |
)
|
166 |
|
167 |
infer_seed_rand_button.click(
|
modules/webui/webui_utils.py
CHANGED
@@ -276,11 +276,12 @@ def refine_text(
|
|
276 |
|
277 |
@torch.inference_mode()
|
278 |
@spaces.GPU(duration=120)
|
279 |
-
def split_long_text(long_text_input):
|
280 |
-
spliter = SentenceSplitter(
|
281 |
sentences = spliter.parse(long_text_input)
|
282 |
-
sentences = [text_normalize(s) for s in sentences]
|
283 |
data = []
|
284 |
for i, text in enumerate(sentences):
|
285 |
-
|
|
|
286 |
return data
|
|
|
276 |
|
277 |
@torch.inference_mode()
|
278 |
@spaces.GPU(duration=120)
|
279 |
+
def split_long_text(long_text_input, spliter_threshold=100, eos=""):
|
280 |
+
spliter = SentenceSplitter(threshold=spliter_threshold)
|
281 |
sentences = spliter.parse(long_text_input)
|
282 |
+
sentences = [text_normalize(s) + eos for s in sentences]
|
283 |
data = []
|
284 |
for i, text in enumerate(sentences):
|
285 |
+
token_length = spliter.count_tokens(text)
|
286 |
+
data.append([i, text, token_length])
|
287 |
return data
|
requirements.txt
CHANGED
@@ -26,4 +26,5 @@ cn2an
|
|
26 |
python-box
|
27 |
ftfy
|
28 |
librosa
|
29 |
-
pyrubberband
|
|
|
|
26 |
python-box
|
27 |
ftfy
|
28 |
librosa
|
29 |
+
pyrubberband
|
30 |
+
https://github.com/Dao-AILab/flash-attention/releases/download/v2.5.9.post1/flash_attn-2.5.9.post1+cu118torch1.12cxx11abiFALSE-cp310-cp310-linux_x86_64.whl
|
webui.py
CHANGED
@@ -30,14 +30,6 @@ from modules.webui.app import create_interface, webui_init
|
|
30 |
dcls_patch()
|
31 |
ignore_useless_warnings()
|
32 |
|
33 |
-
import subprocess
|
34 |
-
|
35 |
-
subprocess.run(
|
36 |
-
"pip install flash-attn --no-build-isolation",
|
37 |
-
env={"FLASH_ATTENTION_SKIP_CUDA_BUILD": "TRUE"},
|
38 |
-
shell=True,
|
39 |
-
)
|
40 |
-
|
41 |
|
42 |
def setup_webui_args(parser: argparse.ArgumentParser):
|
43 |
parser.add_argument("--server_name", type=str, help="server name")
|
|
|
30 |
dcls_patch()
|
31 |
ignore_useless_warnings()
|
32 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
33 |
|
34 |
def setup_webui_args(parser: argparse.ArgumentParser):
|
35 |
parser.add_argument("--server_name", type=str, help="server name")
|