Spaces:
Running
Running
Upload 3 files
Browse files- README.md +31 -0
- app.py +71 -0
- requirements.txt +4 -0
README.md
ADDED
@@ -0,0 +1,31 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
---
|
2 |
+
title: 文本风格转换API
|
3 |
+
emoji: 🔄
|
4 |
+
colorFrom: blue
|
5 |
+
colorTo: green
|
6 |
+
sdk: gradio
|
7 |
+
sdk_version: 3.40.0
|
8 |
+
app_file: app.py
|
9 |
+
pinned: false
|
10 |
+
---
|
11 |
+
|
12 |
+
# 文本风格转换API
|
13 |
+
|
14 |
+
这是一个基于自定义微调模型的中文文本风格转换应用,可以将书面化、技术性文本转换为自然、口语化的表达方式。
|
15 |
+
|
16 |
+
## 功能特点
|
17 |
+
- 支持中文医学、化学等专业文本的通俗化改写
|
18 |
+
- 实时在线转换
|
19 |
+
- 免费API服务
|
20 |
+
|
21 |
+
## API调用
|
22 |
+
部署后可通过以下方式调用:
|
23 |
+
```python
|
24 |
+
import requests
|
25 |
+
|
26 |
+
response = requests.post(
|
27 |
+
"https://yxccai-text-style-api.hf.space/api/predict",
|
28 |
+
json={"data": ["您要转换的文本"]}
|
29 |
+
)
|
30 |
+
result = response.json()
|
31 |
+
print(result["data"][0])
|
app.py
ADDED
@@ -0,0 +1,71 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import gradio as gr
|
2 |
+
from transformers import AutoTokenizer, AutoModelForCausalLM
|
3 |
+
import torch
|
4 |
+
|
5 |
+
# 加载模型和tokenizer
|
6 |
+
model_name = "yxccai/text-style-converter"
|
7 |
+
tokenizer = AutoTokenizer.from_pretrained(model_name)
|
8 |
+
model = AutoModelForCausalLM.from_pretrained(
|
9 |
+
model_name,
|
10 |
+
torch_dtype=torch.float16,
|
11 |
+
device_map="auto"
|
12 |
+
)
|
13 |
+
|
14 |
+
def convert_text_style(input_text):
|
15 |
+
"""文本风格转换函数"""
|
16 |
+
if not input_text.strip():
|
17 |
+
return "请输入要转换的文本"
|
18 |
+
|
19 |
+
prompt = f"""以下是一个文本风格转换任务,请将书面化、技术性的输入文本转换为自然、口语化的表达方式。
|
20 |
+
|
21 |
+
### 输入文本:
|
22 |
+
{input_text}
|
23 |
+
|
24 |
+
### 输出文本:
|
25 |
+
"""
|
26 |
+
|
27 |
+
inputs = tokenizer(prompt, return_tensors="pt")
|
28 |
+
|
29 |
+
with torch.no_grad():
|
30 |
+
outputs = model.generate(
|
31 |
+
inputs.input_ids,
|
32 |
+
attention_mask=inputs.attention_mask,
|
33 |
+
max_new_tokens=500,
|
34 |
+
temperature=0.7,
|
35 |
+
do_sample=True,
|
36 |
+
pad_token_id=tokenizer.eos_token_id
|
37 |
+
)
|
38 |
+
|
39 |
+
full_response = tokenizer.decode(outputs[0], skip_special_tokens=True)
|
40 |
+
|
41 |
+
# 提取生成的部分
|
42 |
+
if "### 输出文本:" in full_response:
|
43 |
+
response = full_response.split("### 输出文本:")[-1].strip()
|
44 |
+
else:
|
45 |
+
response = full_response
|
46 |
+
|
47 |
+
return response
|
48 |
+
|
49 |
+
# 创建Gradio接口
|
50 |
+
iface = gr.Interface(
|
51 |
+
fn=convert_text_style,
|
52 |
+
inputs=gr.Textbox(
|
53 |
+
label="输入文本",
|
54 |
+
placeholder="请输入需要转换为口语化的书面文本...",
|
55 |
+
lines=5
|
56 |
+
),
|
57 |
+
outputs=gr.Textbox(
|
58 |
+
label="输出文本",
|
59 |
+
lines=5
|
60 |
+
),
|
61 |
+
title="中文文本风格转换API",
|
62 |
+
description="将书面化、技术性文本转换为自然、口语化表达",
|
63 |
+
examples=[
|
64 |
+
["乙醇的检测方法包括以下几项: 1. 酸碱度检查:取20ml乙醇加20ml水,加2滴酚酞指示剂应无色。"],
|
65 |
+
["本品为薄膜衣片,除去包衣后显橙红色至暗红色。"]
|
66 |
+
]
|
67 |
+
)
|
68 |
+
|
69 |
+
# 启动应用
|
70 |
+
if __name__ == "__main__":
|
71 |
+
iface.launch()
|
requirements.txt
ADDED
@@ -0,0 +1,4 @@
|
|
|
|
|
|
|
|
|
|
|
1 |
+
transformers>=4.30.0
|
2 |
+
torch>=2.0.0
|
3 |
+
gradio>=3.40.0
|
4 |
+
accelerate>=0.20.0
|