Spaces:
Sleeping
Sleeping
Update app.py
Browse files
app.py
CHANGED
@@ -1,7 +1,103 @@
|
|
1 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2 |
|
3 |
-
def greet(name):
|
4 |
-
return "Hello " + name + "!!"
|
5 |
|
6 |
-
|
7 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
# SPDX-License-Identifier: Apache-2.0
|
2 |
+
|
3 |
+
import os
|
4 |
+
os.system('nvcc --version')
|
5 |
+
os.system('python --version')
|
6 |
+
|
7 |
+
|
8 |
+
|
9 |
+
|
10 |
+
model_url = 'http://localhost:8000/v1'
|
11 |
+
model = "sijieaaa/CodeModel-V1-3B-2024-02-07"
|
12 |
+
stop_token_ids = ''
|
13 |
+
|
14 |
+
|
15 |
+
|
16 |
+
import subprocess
|
17 |
+
|
18 |
+
# Setup LLM API
|
19 |
+
# -- llama-factory
|
20 |
+
def start_api():
|
21 |
+
# 以非阻塞方式启动 API
|
22 |
+
process = subprocess.Popen(
|
23 |
+
["lmf", "api", "--model_name", "sijieaaa/CodeModel-V1-3B-2024-02-07", "--template", "qwen"],
|
24 |
+
stdout=subprocess.PIPE, stderr=subprocess.PIPE, text=True
|
25 |
+
)
|
26 |
+
# 监听 API 启动日志,检测关键字
|
27 |
+
while True:
|
28 |
+
line = process.stdout.readline()
|
29 |
+
if not line:
|
30 |
+
break # 没有新输出,继续等待
|
31 |
+
print(line.strip()) # 打印 API 启动日志
|
32 |
+
if "Running on local URL:" in line or "API started" in line or "http://localhost:" in line:
|
33 |
+
print("✅ API 启动成功!")
|
34 |
+
return process # 返回 API 进程
|
35 |
+
print("❌ API 启动失败!")
|
36 |
+
process.terminate()
|
37 |
+
return None
|
38 |
+
api_process = start_api()
|
39 |
+
# API 启动后执行下一步
|
40 |
+
if api_process:
|
41 |
+
print("🎯 执行后续操作……")
|
42 |
+
else:
|
43 |
+
raise Exception("API 启动失败!")
|
44 |
+
|
45 |
|
|
|
|
|
46 |
|
47 |
+
|
48 |
+
|
49 |
+
# Setup OpenAI API client
|
50 |
+
from openai import OpenAI
|
51 |
+
openai_api_key = '0'
|
52 |
+
openai_api_base = model_url
|
53 |
+
client = OpenAI(
|
54 |
+
api_key=openai_api_key,
|
55 |
+
base_url=openai_api_base,
|
56 |
+
)
|
57 |
+
|
58 |
+
|
59 |
+
def predict(message, history):
|
60 |
+
# Convert chat history to OpenAI format
|
61 |
+
history_openai_format = [{
|
62 |
+
"role": "system",
|
63 |
+
"content": "You are a great ai assistant."
|
64 |
+
}]
|
65 |
+
for human, assistant in history:
|
66 |
+
history_openai_format.append({"role": "user", "content": human})
|
67 |
+
history_openai_format.append({
|
68 |
+
"role": "assistant",
|
69 |
+
"content": assistant
|
70 |
+
})
|
71 |
+
history_openai_format.append({"role": "user", "content": message})
|
72 |
+
|
73 |
+
# Create a chat completion request and send it to the API server
|
74 |
+
stream = client.chat.completions.create(
|
75 |
+
model=model, # Model name to use
|
76 |
+
messages=history_openai_format, # Chat history
|
77 |
+
temperature=0.95, # Temperature for text generation
|
78 |
+
stream=True, # Stream response
|
79 |
+
top_p=0.7,
|
80 |
+
extra_body={
|
81 |
+
'repetition_penalty':
|
82 |
+
1,
|
83 |
+
'stop_token_ids': [
|
84 |
+
int(id.strip()) for id in stop_token_ids.split(',')
|
85 |
+
if id.strip()
|
86 |
+
] if stop_token_ids else []
|
87 |
+
})
|
88 |
+
|
89 |
+
# Read and return generated text from response stream
|
90 |
+
partial_message = ""
|
91 |
+
for chunk in stream:
|
92 |
+
partial_message += (chunk.choices[0].delta.content or "")
|
93 |
+
yield partial_message
|
94 |
+
|
95 |
+
|
96 |
+
|
97 |
+
import gradio as gr
|
98 |
+
# Create and launch a chat interface with Gradio
|
99 |
+
gr.ChatInterface(predict).queue().launch(
|
100 |
+
# server_name=host,
|
101 |
+
# server_port=port,
|
102 |
+
share=True
|
103 |
+
)
|