update main.py

Dockerfile

@@ -1,11 +1,6 @@
 FROM modelhubxc-4pd.tencentcloudcr.com/xc_agent_platform/python:3.11-slim
 
-ENV PYTHONUNBUFFERED=1 \
+ENV PYTHONUNBUFFERED=1
-    USER_ACCOUNT="zhoushasha@4paradigm.com" \
-    USER_PASSWORD="4pdpassword" \
-    CONTEST_API_TOKEN="ef1ef82f3c9efee413d602345fbe224d" \
-    CONTRIBUTORS="zhoushasha" \
-    GPU_TYPE="Cambricon_mlu-370-x8"
 
 WORKDIR /app
 

README.md

@@ -1,5 +1,29 @@
 # xc_validation_strategy
 
-信创自动化模型适配平台 — 验证策略服务
+批量向 ModelHub XC 平台提交模型验证任务的策略服务，之后保持 HTTP 服务存活供平台探活。
 
-从 HuggingFace 周期性抓取新模型，自动完成同步、下载、提交验证任务的全流程，常驻运行在 xc_agent_platform 上。
+## 功能
+
+- 自动登录 ModelHub 获取 Token
+- 批量提交模型验证任务（vLLM 框架，Cambricon MLU-370-x8）
+- 提交结果写入 `submitted_validation_tasks.txt`
+- 暴露 `/health` 和 `/status` 接口满足平台运行时契约
+
+## 项目结构
+
+```
+.
+├── main.py            # 主入口：HTTP 服务 + 提交逻辑
+├── Dockerfile         # 平台镜像构建配置
+├── requirements.txt   # Python 依赖
+└── submitted_validation_tasks.txt   # 运行后自动生成，记录提交结果
+```
+
+## 平台契约说明
+
+本项目满足平台对策略镜像的全部必要约束：
+
+- Dockerfile 位于仓库根目录，基于官方轻量基础镜像
+- 暴露 8080 端口并实现 `GET /health`
+- 通过环境变量 `STRATEGY_ID` 获取策略 ID
+- 正确处理 `SIGTERM` 信号，支持优雅停机

main.py

@@ -9,7 +9,6 @@ import json
 import os
 import signal
 import threading
-import traceback
 from datetime import datetime
 from http.server import BaseHTTPRequestHandler, ThreadingHTTPServer
 from typing import List, Tuple
@@ -20,16 +19,15 @@ import requests
 #  配置（全部从环境变量读取，不硬编码敏感信息）
 # ══════════════════════════════════════════════════════════
 BASE_URL           = os.environ.get("BASE_URL", "https://modelhub.org.cn")
-LOGIN_ENDPOINT     = "/adminApi/user/login"
 SUBMIT_ENDPOINT    = "/adminApi/async/task/create-contest-task"
 
-USER_ACCOUNT       = os.environ["USER_ACCOUNT"]        # 必填
+# 通过 curl -X POST https://modelhub.org.cn/adminApi/user/login 获取后填入
-USER_PASSWORD      = os.environ["USER_PASSWORD"]        # 必填
+AUTH_TOKEN        = "eyJ0eXAiOiJKV1QiLCJhbGciOiJIUzI1NiJ9.eyJ1c2VyQWNjb3VudCI6Inpob3VzaGFzaGEiLCJpZCI6MTQsInVzZXJSb2xlIjoibGVhZGVyYm9hcmQiLCJleHAiOjE3ODE4NTE0NzcsImlhdCI6MTc4MTI0NjY3N30.p3uvCpG50aLNifNVVXxvzmWJahbLM5K1671FVCtj8E8"
-CONTEST_API_TOKEN  = os.environ["CONTEST_API_TOKEN"]    # 必填
+CONTEST_API_TOKEN = "ef1ef82f3c9efee413d602345fbe224d"
-STRATEGY_ID        = os.environ.get("STRATEGY_ID", "")  # 平台注入
+CONTRIBUTORS      = "zhoushasha"
-CONTRIBUTORS       = os.environ.get("CONTRIBUTORS", USER_ACCOUNT)
+GPU_TYPE          = "Cambricon_mlu-370-x8"
-GPU_TYPE           = os.environ.get("GPU_TYPE", "Cambricon_mlu-370-x8")
+TASK_TYPE         = "text-generation"
-TASK_TYPE          = os.environ.get("TASK_TYPE", "text-generation")
+STRATEGY_ID       = os.environ.get("STRATEGY_ID", "")  # 平台自动注入，无需修改
 
 HTTP_HOST = "0.0.0.0"
 HTTP_PORT = 8080
@@ -38,26 +36,36 @@ HTTP_PORT = 8080
 #  模型列表
 # ══════════════════════════════════════════════════════════
 ALL_MODEL_IDS = [
-    "AI-ModelScope/gemma-2b",
+    "l3utterfly/mistral-7b-v0.1-layla-v4",
-    "AI-ModelScope/falcon-mamba-7b",
+    "OpenBuddy/openbuddy-mistral-7b-v13.1",
-    "katanemo/deepseek-2",
+    "allenai/truthfulqa-info-judge-llama2-7B",
-    "OpenBMB/MiniCPM4-0.5B",
+    "l3utterfly/mistral-7b-v0.1-layla-v1",
-    "NousResearch/Meta-Llama-3-8B-Instruct",
+    "l3utterfly/minima-3b-layla-v2",
-    "MediaTek-Research/Breeze-7B-Instruct-v1_0",
+    "l3utterfly/tinyllama-1.1b-layla-v4",
-    "QLUNLP/BianCang-Qwen2.5-7B-Instruct",
+    "l3utterfly/mistral-7b-v0.1-layla-v2",
-    "OpenBMB/MiniCPM4-Survey",
+    "l3utterfly/tinyllama-1.1b-layla-v1",
-    "OpenBMB/MiniCPM4-8B",
+    "Duxiaoman-DI/XuanYuan-13B-Chat",
-    "PaddlePaddle/ERNIE-4.5-0.3B-PT",
+    "l3utterfly/minima-3b-layla-v1",
-    "LLM-Research/Llama-Guard-3-8B",
+    "AI-ModelScope/gemma-2-2b",
-    "OpenBMB/MiniCPM-2B-dpo-fp16",
+    "baichuan-inc/Baichuan-13B-Base",
-    "OpenBMB/MiniCPM4.1-8B",
+    "LGAI-EXAONE/EXAONE-Deep-2.4B",
-    "Cylingo/Xinyuan-LLM-14B-0428",
+    "NousResearch/DeepHermes-3-Llama-3-3B-Preview",
-    "Fengshenbang/Ziya-LLaMA-13B-v1",
+    "Fengshenbang/Ziya2-13B-Base",
-    "baichuan-inc/Baichuan2-13B-Chat",
+    "prithivMLmods/QwQ-MathOct-7B",
-    "LLM-Research/gemma-2-9b-it",
+    "l3utterfly/phi-2-layla-v1-chatml",
-    "Qwen/CodeQwen1.5-7B-Chat",
+    "argilla/notus-7b-v1",
-    "OpenBMB/cpm-bee-10b",
+    "prithivMLmods/Doopler-Augment-3B-Cox",
-    "OpenBMB/MiniCPM3-4B",
+    "prithivMLmods/Blaze.1-32B-Instruct",
+    "CohereLabs/aya-expanse-8B",
+    "Magpie-Align/MagpieLM-4B-SFT-v0.1",
+    "Magpie-Align/MagpieLM-8B-SFT-v0.1",
+    "Magpie-Align/Llama-3-8B-Magpie-Align-SFT-v0.2",
+    "Magpie-Align/MagpieLM-8B-Chat-v0.1",
+    "Magpie-Align/Llama-3.1-8B-Magpie-Align-SFT-v0.1",
+    "Magpie-Align/Llama-3-8B-Magpie-Air-SFT-300K-v0.1",
+    "prithivMLmods/Tulu-MathLingo-8B",
+    "prithivMLmods/Triangulum-5B",
+    "prithivMLmods/Viper-Coder-v0.1",
 ]
 
 # ══════════════════════════════════════════════════════════
@@ -110,21 +118,6 @@ def _run_http():
 # ══════════════════════════════════════════════════════════
 #  业务逻辑
 # ══════════════════════════════════════════════════════════
-def _login() -> str:
-    headers = {"Content-Type": "application/json"}
-    resp = requests.post(
-        BASE_URL + LOGIN_ENDPOINT,
-        headers=headers,
-        json={"userAccount": USER_ACCOUNT, "userPassword": USER_PASSWORD},
-        timeout=30,
-    )
-    data = resp.json()
-    if data.get("code") != 0:
-        raise RuntimeError(f"登录失败: {data.get('message')}")
-    print("[worker] 登录成功", flush=True)
-    return data["data"]["token"]
-
-
 def _submit_task(token: str, model_id: str) -> Tuple[bool, str]:
     headers = {
         "Content-Type": "application/json",
@@ -167,6 +160,7 @@ ref_config:
             "taskType": TASK_TYPE,
         }],
     }
+    print(f"[payload] {json.dumps(payload, indent=2, ensure_ascii=False)}", flush=True)
     try:
         resp = requests.post(
             BASE_URL + SUBMIT_ENDPOINT,
@@ -192,12 +186,8 @@ def _run_worker():
     _state["phase"] = "submitting"
 
     successful: List[Tuple[str, str]] = []
-    try:
+    token = AUTH_TOKEN
-        token = _login()
+    print("[worker] 使用预设 Token，跳过登录", flush=True)
-    except Exception:
-        traceback.print_exc()
-        _state["phase"] = "error"
-        return
 
     for model_id in ALL_MODEL_IDS:
         if _shutdown.is_set():