""" xc_validation_strategy_vllm_gguf_hygon_zkl — 主入口 启动后通过 /api/adapt/task/add 接口（xc-Token 认证）批量提交 GGUF 模型适配任务（hygon_k100-ai，llamacpp 框架），之后保持 HTTP 服务存活。 同时暴露 /health（K8s 探活）和 /status（运行状态）。 部署框架与 xc_validation_strategy 一致。 """ import json import os import signal import threading from datetime import datetime from http.server import BaseHTTPRequestHandler, ThreadingHTTPServer from typing import List import requests # ══════════════════════════════════════════════════════════ # 配置 # ══════════════════════════════════════════════════════════ BASE_URL = os.environ.get("BASE_URL", "https://modelhub.org.cn") ADD_TASK_ENDPOINT = "/api/adapt/task/add" # zhouyuanxi 账号的 xc-Token（该接口使用 xc-Token 认证，无需登录） USER_ACCOUNT = "i-zhouyuanxi@4paradigm.com" XC_TOKEN = "62b9b487eff2488fb9f1da0b963f0b93" GPU_TYPE = "hygon_k100-ai" TASK_TYPE = "text-generation" STRATEGY_ID = os.environ.get("STRATEGY_ID", "") # 平台自动注入，无需修改 HEADERS = { "Content-Type": "application/json", "xc-Token": XC_TOKEN, } HTTP_HOST = "0.0.0.0" HTTP_PORT = 8080 # ══════════════════════════════════════════════════════════ # 模型列表 # ══════════════════════════════════════════════════════════ ALL_MODEL_IDS = [ "mradermacher/kyara-lite-sft-1M-GGUF", "mradermacher/text-arco-GGUF", "mradermacher/Mistral-7B-v0.3-sft-ultrachat-GGUF", "mradermacher/Saba2-Preview-GGUF", "mradermacher/Anemoi-4B-GGUF", "mradermacher/exaone-3.5-2.4b-instruct-dacon-llm-GGUF", "mradermacher/light-3B-beta-GGUF", "mradermacher/bellman-mistral-7b-instruct-v0.3-GGUF", "mradermacher/Kosmos-EVAA-Franken-Immersive-v40-8B-GGUF", "mradermacher/Qwen2.5-Interpreter-GGUF", "mradermacher/Llama-3.1-8B-Gemma-2-9B-mix-GGUF", "mradermacher/Llama-3.1-8B-sft-ultrachat-GGUF", "mradermacher/t5-base-english-japanese-GGUF", "mradermacher/tourist-GGUF", "mradermacher/t5-base-japanese-question-generation-GGUF", "mradermacher/t5-qiita-title-generation-GGUF", "mradermacher/Not_Even_My_Final_Form-8B-Model_Stock-GGUF", "mradermacher/Morphing-8B-Model_Stock-GGUF", "mradermacher/Llamaverse-3.1-8B-Instruct-GGUF", "mradermacher/Gemma-2-9B-sft-ultrachat-GGUF", "mradermacher/Italy-10B-GGUF", "mradermacher/Qwen2.5-7B-sft-ultrachat-safeRLHF-GGUF", "mradermacher/Qwen2-7B-sft-ultrachat-safeRLHF-GGUF", "mradermacher/qwen-2.5-0.5b-unlabeled-pile-GGUF", "mradermacher/HQwen-3B-00-fp16-GGUF", "mradermacher/AI_Pet_code-GGUF", "mradermacher/codellama-7b-betterv-params-c2v-chisel-GGUF", "mradermacher/LongRAG-Qwen2.5-7B-Instruct-GGUF", "mradermacher/granite-3.0-3b-a800m-base-GGUF", "mradermacher/granite-3.0-8b-base-GGUF", "mradermacher/moxin-llm-7b-GGUF", "mradermacher/llama-3.2-3b-merged-base-3K-updated-GGUF", "mradermacher/II-Tulu-8B-SFT-GGUF", "mradermacher/HebQwen-json-2025-GGUF", "mradermacher/WiNGPT-Babel-GGUF", "mradermacher/oh-dcft-v3.1-llama-3.1-405b-v2dummytesting-GGUF", "mradermacher/greesychat-turbo-GGUF", "mradermacher/MFANN-abliterated-phi2-merge-unretrained-GGUF", "mradermacher/oh-dcft-v3.1-llama-3.1-405b-qwen-v2dummytesting-GGUF", "mradermacher/Llama-3.1-8B-sft-ultrachat-safeRLHF-GGUF", "mradermacher/Falcon-3B-8k-data-new-classes-GGUF", "mradermacher/Lunar-4B-GGUF", "mradermacher/Zelus-8B-Model_Stock-GGUF", "mradermacher/mpn_mistral7bv3_sft-GGUF", "mradermacher/Qwen2-7B-sft-ultrachat-GGUF", "mradermacher/Llama-3.1-8B-Instruct-GSM8k-GGUF", "mradermacher/Mistral-7B-v0.3-sft-ultrachat-safeRLHF-GGUF", "mradermacher/QwENDEAVR-VL-2B-GGUF", "mradermacher/EXAONE-3.5-7.8B-it-DACON-LLM-GGUF", "mradermacher/Gemma-2-9B-sft-ultrachat-safeRLHF-GGUF", "mradermacher/Eunoia-GwQ-Gemma-9B-GGUF", "mradermacher/periquito-3B-GGUF", "mradermacher/T3qm7xNeuralsirkrishna_Strangemerges_30Ogno-GGUF", "mradermacher/Chalice1-GGUF", "mradermacher/Yamshadowexperiment28M70.5-0.45-0.64-0.12-0.94-0.72-7B-GGUF", "mradermacher/M7Yamshadowexperiment28_Strangemerges_32Yamshadow-GGUF", "mradermacher/Experiment26Yam_Strangemerges_32Ogno-GGUF", "mradermacher/Experiment28T3q_NeuralsirkrishnaExperiment26-GGUF", "mradermacher/YamshadowStrangemerges_32_Strangemerges_32Experiment28-GGUF", "mradermacher/Experiment28M7_Experiment28Neuralsirkrishna-GGUF", "mradermacher/YamshadowStrangemerges_32_Neuralsirkrishnaexperiment26Experiment24-GGUF", "mradermacher/YamshadowInex12_Experiment29Pastiche-GGUF", "mradermacher/tsukasa-7b-qlora-GGUF", "mradermacher/Qwen2.5-3B-Model-Stock-v2-GGUF", "mradermacher/LongRAG_Ministral-8B-Instruct-GGUF", "mradermacher/LongRAG_Llama3.1-8B-Instruct-GGUF", "mradermacher/inter-play-sim-assistant-test-GGUF", "mradermacher/Llama-3.2-1B-magnitude-0.1-GGUF", "mradermacher/Howdy-8B-LINEAR-GGUF", "mradermacher/Qwen2.5-3B-Model-Stock-GGUF", "mradermacher/Ice0.51-16.01-RP-GGUF", "mradermacher/Qwen2.5-3B-Model-Stock-v3-GGUF", "mradermacher/SalvadAI-GGUF", "mradermacher/Yi-6B-Chat-Llama-3.3-70B-Instruct-honest_lying-GGUF", "mradermacher/Funny-Chatbot-Llama-3.1-8B-Instruct-GGUF", "mradermacher/Aira-2-355M-GGUF", "mradermacher/Qmerft-GGUF", "mradermacher/Qwen2.5-1.5B-Instruct-DPO-bad-boy-GGUF", "mradermacher/NxMobileLM-1.5B-SFT-GGUF", "mradermacher/Llama-3-ELYZA-JP-8B-normal-chosen-GGUF", "mradermacher/HebQwen-json-2025-meta-GGUF", "mradermacher/JapMed-GGUF", "mradermacher/What_A_Thrill-8B-Model_Stock-GGUF", "mradermacher/phi-3-clinical-GGUF", "mradermacher/ORPO-Llama-SQL-8B-Spider-GGUF", "mradermacher/Gemma-Radiation-RP-9B-lorablated-GGUF", "mradermacher/D.A.T.A.v1-GGUF", "mradermacher/gemmarpp-GGUF", "mradermacher/Meta-Llama-3-8B-wanda-0.1-GGUF", "mradermacher/LwQ-10B-Instruct-GGUF", "mradermacher/occiglot-7b-de-en-instruct-dpo1-GGUF", "mradermacher/MCQ-qkvo-2-GGUF", "mradermacher/YamshadowStrangemerges_32_Inex12Experiment28-GGUF", "mradermacher/Bahasa-4b-GGUF", "mradermacher/MCQ-head-16-GGUF", "mradermacher/Experiment26Yam_Ognoexperiment27Multi_verse_model-GGUF", "mradermacher/An4-7Bv2.4-GGUF", "mradermacher/YamshadowStrangemerges_32_YamExperiment28-GGUF", "mradermacher/bitmap-M7-alpaca-70m-GGUF", "mradermacher/UpshotLlama-3-8B-GGUF", "mradermacher/ChimeraLlama-3-8B-GGUF", "mradermacher/Yamshadowexperiment28M70.8-0.85-0.86-0.36-0.42-0.16-7B-GGUF", "mradermacher/microcoderfim-1B-GGUF", "mradermacher/Medusa_OpenHermes-2.5-Mistral7B-DPO-GGUF", "mradermacher/YamshadowInex12_Experiment24Ognoexperiment27-GGUF", "mradermacher/gemma-2b-datascience-instruct-v4.5-GGUF", "mradermacher/Josiefied-Qwen2-1.5B-Instruct-abliterated-GGUF", "mradermacher/zoyllm-7b-slimorca-GGUF", "mradermacher/M7Yamshadowexperiment28_T3qm7xExperiment28-GGUF", "mradermacher/Aria-7b-128k-v4-GGUF", "mradermacher/T3qm7xNeuralsirkrishna_CalmeInex12-GGUF", "mradermacher/T3qm7xNeuralsirkrishna_YamStrangemerges_32-GGUF", "mradermacher/YamshadowInex12_T3qm7xExperiment28-GGUF", "mradermacher/Model_Stock_Mixv0.1-7B-GGUF", "mradermacher/StrangeMerges_49-7B-dare_ties-GGUF", "mradermacher/galland-mistral-7b-instruct-v0.2-v1-GGUF", "mradermacher/Magiq-Translator-0-GGUF", "mradermacher/occiglot-7b-de-en-instruct-GGUF", "mradermacher/M7Yamshadowexperiment28_YamShadow-GGUF", "mradermacher/StrangeMerges_17-7B-dare_ties-GGUF", "mradermacher/occiglot-7b-es-en-instruct-hf-GGUF", "mradermacher/ko-gemma-1.1-2b-GGUF", "mradermacher/M7Yamshadowexperiment28_Percival_01Yamshadow-GGUF", "mradermacher/TinyAlpaca-1.1B-GGUF", "mradermacher/MeliodasT3q_M7Alloyingotneoy-GGUF", "mradermacher/Vistral-7B-Chat-function-calling-GGUF", ] # ══════════════════════════════════════════════════════════ # 全局状态（供 /status 展示） # ══════════════════════════════════════════════════════════ _state = { "strategy_id": STRATEGY_ID, "phase": "starting", # starting | submitting | done | error "total": len(ALL_MODEL_IDS), "submitted": 0, "failed": 0, "started_at": None, "finished_at": None, } _shutdown = threading.Event() # ══════════════════════════════════════════════════════════ # HTTP 服务 # ══════════════════════════════════════════════════════════ class Handler(BaseHTTPRequestHandler): def do_GET(self): if self.path == "/health": self._json({"status": "ok"}) elif self.path == "/status": self._json(_state) else: self._json({"error": "not found"}, 404) def _json(self, body: dict, code: int = 200): payload = json.dumps(body, default=str).encode() self.send_response(code) self.send_header("Content-Type", "application/json") self.send_header("Content-Length", str(len(payload))) self.end_headers() self.wfile.write(payload) def log_message(self, fmt, *args): print(f"[http] {self.address_string()} {fmt % args}", flush=True) def _run_http(): server = ThreadingHTTPServer((HTTP_HOST, HTTP_PORT), Handler) server.timeout = 1 print(f"[http] 监听 {HTTP_HOST}:{HTTP_PORT}", flush=True) while not _shutdown.is_set(): server.handle_request() server.server_close() print("[http] 已关闭", flush=True) # ══════════════════════════════════════════════════════════ # 工具函数：生成模型文件名（一律 .f16.gguf） # ══════════════════════════════════════════════════════════ def get_model_filename(model_id: str) -> str: base_name = model_id.split("/")[-1] if '_-_' in base_name: base_name = base_name.split('_-_')[-1] if base_name.lower().endswith("-gguf"): base_name = base_name[:-5] base_name = base_name + ".f16.gguf" return base_name # ══════════════════════════════════════════════════════════ # 业务逻辑 # ══════════════════════════════════════════════════════════ def submit_task(model_id: str) -> bool: model_filename = get_model_filename(model_id) config_content = f""" docker_image: git.modelhub.org.cn:9443/enginex-hygon/hygon-llama.cpp:b7516 nv_docker_image: harbor-contest.4pd.io/luxinlong02/llama-cpp:b7003-cuda-full-12.3 framework: llamacpp storage: gpfs api: completion temperature: 0 repetition_penalty: 1.1 top_p: 0.9 max_model_len: 4096 sut_config: gpu_num: 1 values: command: ['/app/llama-server','--model', '/model/{model_filename}', '--alias', 'llm', '--threads', '20','--n-gpu-layers', '999', '--prio', '3', '--min_p', '0.01', '--ctx-size', '4096', '--host', '0.0.0.0', '--port', '8000', '--jinja', '--flash-attn', 'off'] ref_config: gpu_num: 1 values: command: ['/workspace/llama.cpp/build/bin/llama-server','--model', '/model/{model_filename}', '--alias', 'llm', '--threads', '20', '--n-gpu-layers', '999', '--prio', '3', '--min_p', '0.01', '--ctx-size', '4096', '--host', '0.0.0.0', '--port', '8000', '--jinja', '--flash-attn', 'off'] """ payload = { "configParams": config_content, "framework": "llamacpp", "modelAddress": f"https://huggingface.co/{model_id}", "targetGpu": GPU_TYPE, "taskType": TASK_TYPE, "strategyId": STRATEGY_ID, # 平台要求；若接口不支持该字段会被忽略 } print(f"📤 提交任务: {model_id} ({model_filename})", flush=True) try: resp = requests.post( BASE_URL + ADD_TASK_ENDPOINT, headers=HEADERS, json=payload, timeout=30, ) print(f"status: {resp.status_code}", flush=True) result = resp.json() print(result, flush=True) if result.get("code") == 0: print(f"✅ 提交成功: {model_id}", flush=True) return True else: print(f"❌ 提交失败: {result.get('message')}", flush=True) return False except Exception as e: print(f"💥 异常 ({model_id}): {e}", flush=True) return False def _run_worker(): _state["started_at"] = datetime.utcnow().isoformat() _state["phase"] = "submitting" successful: List[str] = [] for model_id in ALL_MODEL_IDS: if _shutdown.is_set(): break if submit_task(model_id): _state["submitted"] += 1 successful.append(model_id) else: _state["failed"] += 1 try: with open("submitted_adapt_tasks.txt", "w", encoding="utf-8") as f: for mid in successful: f.write(f"{mid}\n") except Exception: pass _state["finished_at"] = datetime.utcnow().isoformat() _state["phase"] = "done" print( f"[worker] 完成 submitted={_state['submitted']} failed={_state['failed']} total={_state['total']}", flush=True, ) # 提交完成后继续保持进程存活，等待平台停止 # ══════════════════════════════════════════════════════════ # 入口 # ══════════════════════════════════════════════════════════ def _handle_signal(signum, _frame): print(f"[main] 收到信号 {signum}，正在关闭...", flush=True) _shutdown.set() def main(): signal.signal(signal.SIGTERM, _handle_signal) signal.signal(signal.SIGINT, _handle_signal) http_thread = threading.Thread(target=_run_http, daemon=False) http_thread.start() worker_thread = threading.Thread(target=_run_worker, daemon=True) worker_thread.start() _shutdown.wait() print("[main] 等待 HTTP 服务关闭...", flush=True) http_thread.join(timeout=5) print("[main] 退出", flush=True) if __name__ == "__main__": main()