This commit is contained in:
parent
f82e6c567f
commit
35ba2eab42
23
Dockerfile
23
Dockerfile
|
|
@ -108,6 +108,9 @@ RUN pip wheel \
|
|||
setproctitle uvloop sentencepiece triton pillow cachetools msgspec blake3 cloudpickle compressed-tensors einops openai py-cpuinfo dill partial_json_parser python-multipart torchao \
|
||||
-w /wheels
|
||||
|
||||
# ── ✅ 打包 gradio UI 所需依赖 ────────────────────────────────────────────────
|
||||
RUN pip wheel gradio requests -w /wheels
|
||||
|
||||
###############################################################################
|
||||
# Stage 2 ─ runtime:极简运行镜像,仅离线安装 wheel
|
||||
###############################################################################
|
||||
|
|
@ -165,13 +168,15 @@ COPY ./Alibaba/Qwen3-14B-Base /root/.cradle/Alibaba/Qwen3-14B-Base
|
|||
HEALTHCHECK --interval=30s --timeout=2s --start-period=300s --retries=5 CMD curl -fs http://localhost:30000/health || exit 1
|
||||
|
||||
# ---- 暴露端口 ----
|
||||
EXPOSE 30000
|
||||
EXPOSE 30000 30001
|
||||
|
||||
# ---- 启动 SGLang 推理服务 ----
|
||||
CMD ["python3", "-m", "sglang.launch_server", \
|
||||
"--host", "0.0.0.0", \
|
||||
"--port", "30000", \
|
||||
"--model-path", "/root/.cradle/Alibaba/Qwen3-14B-Base/", \
|
||||
"--tp", "2", \
|
||||
"--api-key", "token-abc123", \
|
||||
"--enable-metrics"]
|
||||
# 安装 supervisor
|
||||
RUN apt-get update && apt-get install -y supervisor && \
|
||||
mkdir -p /etc/supervisor/conf.d
|
||||
|
||||
# 拷贝 supervisord 配置文件和 UI 脚本
|
||||
COPY ./meta_ui.py /app/meta_ui.py
|
||||
COPY ./supervisord.conf /etc/supervisor/supervisord.conf
|
||||
|
||||
# 作为容器主进程运行 supervisor
|
||||
CMD ["/usr/bin/supervisord", "-c", "/etc/supervisor/supervisord.conf"]
|
||||
|
|
@ -0,0 +1,79 @@
|
|||
import gradio as gr
|
||||
import requests
|
||||
|
||||
API_URL = "http://localhost:30000/v1/completions"
|
||||
API_KEY = "token-abc123"
|
||||
MODEL_NAME = "Qwen3-14b-base"
|
||||
|
||||
# 构造 prompt:Base 模型靠拼接上下文
|
||||
def build_prompt(history, user_message):
|
||||
prompt = ""
|
||||
for user, bot in history:
|
||||
prompt += f"User: {user}\nAssistant: {bot}\n"
|
||||
prompt += f"User: {user_message}\nAssistant:"
|
||||
return prompt
|
||||
|
||||
# 主对话函数
|
||||
def chat(user_message, history, max_tokens, temperature):
|
||||
prompt = build_prompt(history, user_message)
|
||||
|
||||
headers = {
|
||||
"Authorization": f"Bearer {API_KEY}",
|
||||
"Content-Type": "application/json"
|
||||
}
|
||||
payload = {
|
||||
"model": MODEL_NAME,
|
||||
"prompt": prompt,
|
||||
"max_tokens": max_tokens,
|
||||
"temperature": temperature,
|
||||
"stop": ["\nUser:", "\nAssistant:"]
|
||||
}
|
||||
|
||||
try:
|
||||
response = requests.post(API_URL, headers=headers, json=payload, timeout=30)
|
||||
result = response.json()
|
||||
reply = result["choices"][0]["text"].strip()
|
||||
except Exception as e:
|
||||
reply = f"[请求失败] {e}"
|
||||
|
||||
return reply
|
||||
|
||||
# 手动测试 API 功能
|
||||
def test_api_connection(max_tokens, temperature):
|
||||
headers = {
|
||||
"Authorization": f"Bearer {API_KEY}",
|
||||
"Content-Type": "application/json"
|
||||
}
|
||||
payload = {
|
||||
"model": MODEL_NAME,
|
||||
"prompt": "Ping?",
|
||||
"max_tokens": max_tokens,
|
||||
"temperature": temperature
|
||||
}
|
||||
|
||||
try:
|
||||
resp = requests.post(API_URL, headers=headers, json=payload, timeout=10)
|
||||
out = resp.json()["choices"][0]["text"].strip()
|
||||
return f"✅ API 可用,响应: {out}"
|
||||
except Exception as e:
|
||||
return f"❌ API 请求失败: {e}"
|
||||
|
||||
# Gradio 控件组合
|
||||
with gr.Blocks(title="Base 模型测试 UI") as demo:
|
||||
gr.Markdown("# 💬 Base 模型对话界面")
|
||||
|
||||
with gr.Row():
|
||||
max_tokens = gr.Slider(32, 1024, value=256, label="max_tokens")
|
||||
temperature = gr.Slider(0.0, 1.5, value=0.7, step=0.05, label="temperature")
|
||||
test_btn = gr.Button("🔁 测试 API 可用性")
|
||||
test_output = gr.Textbox(label="API 测试结果", interactive=False)
|
||||
|
||||
chatbot = gr.ChatInterface(
|
||||
fn=lambda msg, hist: chat(msg, hist, max_tokens.value, temperature.value),
|
||||
title=None
|
||||
)
|
||||
|
||||
test_btn.click(fn=test_api_connection, inputs=[max_tokens, temperature], outputs=test_output)
|
||||
|
||||
# 启动服务
|
||||
demo.launch(server_name="0.0.0.0", server_port=30001)
|
||||
|
|
@ -0,0 +1,18 @@
|
|||
[supervisord]
|
||||
nodaemon=true
|
||||
logfile=/dev/stdout
|
||||
loglevel=info
|
||||
|
||||
[program:sglang]
|
||||
command=python3 -m sglang.launch_server --host 0.0.0.0 --port 30000 --model-path /root/.cradle/Alibaba/Qwen3-14B-Base/ --tp 2 --api-key token-abc123 --enable-metrics
|
||||
autostart=true
|
||||
autorestart=true
|
||||
stdout_logfile=/dev/stdout
|
||||
stderr_logfile=/dev/stderr
|
||||
|
||||
[program:ui]
|
||||
command=python3 /app/meta_ui.py --port 30001
|
||||
autostart=true
|
||||
autorestart=true
|
||||
stdout_logfile=/dev/stdout
|
||||
stderr_logfile=/dev/stderr
|
||||
Loading…
Reference in New Issue