Spaces:
Sleeping
Sleeping
Create App.py
Browse files
App.py
ADDED
@@ -0,0 +1,71 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import os
|
2 |
+
import gradio as gr
|
3 |
+
import requests
|
4 |
+
|
5 |
+
# Получите токен из переменных окружения
|
6 |
+
HUGGINGFACE_TOKEN = os.getenv("HUGGINGFACE_TOKEN")
|
7 |
+
|
8 |
+
# Инициализация Hugging Face API
|
9 |
+
API_URL = "https://api-inference.huggingface.co/models/Orenguteng/Llama-3.1-8B-Lexi-Uncensored-V2"
|
10 |
+
headers = {"Authorization": f"Bearer {HUGGINGFACE_TOKEN}"}
|
11 |
+
|
12 |
+
def chat(message, history, system_message, max_tokens, temperature, top_p):
|
13 |
+
# Подготовка сообщений
|
14 |
+
messages = [{"role": "system", "content": system_message}]
|
15 |
+
|
16 |
+
for val in history:
|
17 |
+
if val[0]:
|
18 |
+
messages.append({"role": "user", "content": val[0]})
|
19 |
+
if val[1]:
|
20 |
+
messages.append({"role": "assistant", "content": val[1]})
|
21 |
+
|
22 |
+
messages.append({"role": "user", "content": message})
|
23 |
+
|
24 |
+
# Подготовка входных данных для модели
|
25 |
+
input_text = "\n".join([f"{msg['role']}: {msg['content']}" for msg in messages])
|
26 |
+
|
27 |
+
# Отправка запроса к Hugging Face API
|
28 |
+
response = requests.post(
|
29 |
+
API_URL,
|
30 |
+
headers=headers,
|
31 |
+
json={
|
32 |
+
"inputs": input_text,
|
33 |
+
"parameters": {
|
34 |
+
"max_new_tokens": max_tokens,
|
35 |
+
"temperature": temperature,
|
36 |
+
"top_p": top_p,
|
37 |
+
},
|
38 |
+
}
|
39 |
+
)
|
40 |
+
|
41 |
+
# Проверка статуса ответа
|
42 |
+
if response.status_code != 200:
|
43 |
+
return "Error with API call: " + response.text, history
|
44 |
+
|
45 |
+
# Возвращаем только последний ответ от модели
|
46 |
+
response_text = response.json()[0]['generated_text'].strip()
|
47 |
+
|
48 |
+
# Добавляем сообщение и ответ в историю
|
49 |
+
history.append((message, response_text))
|
50 |
+
|
51 |
+
return response_text, history
|
52 |
+
|
53 |
+
# Определение интерфейса Gradio
|
54 |
+
iface = gr.Interface(
|
55 |
+
fn=chat,
|
56 |
+
inputs=[
|
57 |
+
gr.Textbox(label="Message"),
|
58 |
+
gr.State([]), # Для хранения истории сообщений
|
59 |
+
gr.Textbox(value="You are a friendly Chatbot.", label="System Message"),
|
60 |
+
gr.Slider(minimum=1, maximum=2048, value=50, step=1, label="Max New Tokens"),
|
61 |
+
gr.Slider(minimum=0.1, maximum=4.0, value=0.7, step=0.1, label="Temperature"),
|
62 |
+
gr.Slider(minimum=0.1, maximum=1.0, value=0.95, step=0.05, label="Top-p (nucleus sampling)"),
|
63 |
+
],
|
64 |
+
outputs=[
|
65 |
+
gr.Textbox(label="Response"),
|
66 |
+
gr.State([]) # Состояние должно быть также определено в выходных параметрах
|
67 |
+
],
|
68 |
+
)
|
69 |
+
|
70 |
+
if __name__ == "__main__":
|
71 |
+
iface.launch()
|