Compare commits
8 Commits
| Author | SHA1 | Date | |
|---|---|---|---|
| 9b5087467f | |||
| 7dcada5617 | |||
| 5958df93b0 | |||
| a73274e6a4 | |||
| b3c577219f | |||
| 1591b3050e | |||
| 55c77faa70 | |||
| e51533e0bf |
2
.gitignore
vendored
Normal file
2
.gitignore
vendored
Normal file
@@ -0,0 +1,2 @@
|
||||
.DS_Store
|
||||
__pycache__/
|
||||
Binary file not shown.
546
main.py
546
main.py
@@ -1,7 +1,9 @@
|
||||
"""
|
||||
xc_validation_strategy — 主入口
|
||||
|
||||
启动后执行一次模型验证任务批量提交，之后保持 HTTP 服务存活。
|
||||
启动后针对 4 张 GPU 卡（Biren_166m / Cambricon_mlu-370-x8 / MetaX_c-500 /
|
||||
Kunlunxin_p-800）分别批量提交各自筛选出的模型验证任务（/adminApi/async/task/create-contest-task，
|
||||
Bearer Token 认证），之后保持 HTTP 服务存活。
|
||||
同时暴露 /health（K8s 探活）和 /status（运行状态）。
|
||||
"""
|
||||
|
||||
@@ -22,10 +24,9 @@ BASE_URL = os.environ.get("BASE_URL", "https://modelhub.org.cn")
|
||||
SUBMIT_ENDPOINT = "/adminApi/async/task/create-contest-task"
|
||||
|
||||
# 通过 curl -X POST https://modelhub.org.cn/adminApi/user/login 获取后填入
|
||||
AUTH_TOKEN = "eyJ0eXAiOiJKV1QiLCJhbGciOiJIUzI1NiJ9.eyJ1c2VyQWNjb3VudCI6Inpob3VzaGFzaGEiLCJpZCI6MTQsInVzZXJSb2xlIjoibGVhZGVyYm9hcmQiLCJleHAiOjE3ODQ1NDc1NDYsImlhdCI6MTc4Mzk0Mjc0Nn0.ZcOqcrfI22LPi4mGMnt164nZGhi61ZxtJGYsoO7fZdM"
|
||||
AUTH_TOKEN = "eyJ0eXAiOiJKV1QiLCJhbGciOiJIUzI1NiJ9.eyJ1c2VyQWNjb3VudCI6Inpob3VzaGFzaGEiLCJpZCI6MTQsInVzZXJSb2xlIjoibGVhZGVyYm9hcmQiLCJleHAiOjE3ODU3NDY3NTMsImlhdCI6MTc4NTE0MTk1M30.KwUuefNAFSNwq3_Pnaw2nef8ZC6WgsECQ_LMeQnKk2c"
|
||||
CONTEST_API_TOKEN = "ef1ef82f3c9efee413d602345fbe224d"
|
||||
CONTRIBUTORS = "zhoushasha"
|
||||
GPU_TYPE = "ppu_zw_810e"
|
||||
TASK_TYPE = "text-generation"
|
||||
STRATEGY_ID = os.environ.get("STRATEGY_ID", "") # 平台自动注入，无需修改
|
||||
|
||||
@@ -33,245 +34,140 @@ HTTP_HOST = "0.0.0.0"
|
||||
HTTP_PORT = 8080
|
||||
|
||||
# ══════════════════════════════════════════════════════════
|
||||
# 模型列表
|
||||
# 各 GPU 的模型列表（来自 filter_verified_models 脚本的筛选结果）
|
||||
# ══════════════════════════════════════════════════════════
|
||||
ALL_MODEL_IDS = [
|
||||
"Alienpenguin10/M3PO-bahdanau-trial1-seed123",
|
||||
"sujalrajpoot/TrueSyncAI-Aurion",
|
||||
"prithivMLmods/Tureis-Qwen3_QWQ-4B-Exp",
|
||||
"standrey/listing-parser-llama31-8b-ft-v1-full",
|
||||
"zarakiquemparte/zarablend-l2-7b",
|
||||
"linzju/Bio-Medical-Llama-3-8B_EnchTable_FFN",
|
||||
"leonMW/Qwen3-4B-Thinking-2507-GSPO-Easy",
|
||||
"longvideoagent/longvideoagent-qwen3-4b",
|
||||
"ishikaa/acquisition_qwen3b_alpaca_proximity",
|
||||
"01ai/Yi-9B-200K",
|
||||
"Gille/StrangeMerges_33-7B-slerp",
|
||||
"sstoica12/acquisition_llama-3_2-3b_bins_medmcqa_gradient",
|
||||
"speechlessai/speechless-coding-7b-16k-tora",
|
||||
"unsloth/Qwen2.5-Math-1.5B-Instruct",
|
||||
"Yuma42/KangalKhan-Sapphire-7B",
|
||||
"shadowml/BeagleSempra-7B",
|
||||
"bralynn/test18",
|
||||
"m-a-p/OProver-8B-Round1",
|
||||
"yeen214/test_llama2_7b",
|
||||
"Xwin-LM/Xwin-LM-7B-V0.2",
|
||||
"FreedomIntelligence/AceGPT-13B",
|
||||
"Edcastro/tinyllama-edcastr_JavaScript-v2",
|
||||
"zarakiquemparte/zaraxe-l2-7b",
|
||||
"MaziyarPanahi/Llama-3-8B-Instruct-v0.8",
|
||||
"defog/sqlcoder2",
|
||||
"SawinuCP/bus_booking_voice_agent_merged",
|
||||
"j05hr3d/Llama-3.2-3B-Instruct-C_M_T-DOLLY-SEED999",
|
||||
"Changgil/K2S3-SOLAR-11b-v1.0",
|
||||
"agentica-org/DeepCoder-1.5B-Preview",
|
||||
"prithivMLmods/Omni-Reasoner3-Merged",
|
||||
"ibm-granite/granite-7b-instruct",
|
||||
"zarakiquemparte/kuchiki-1.1-l2-7b",
|
||||
"MaziyarPanahi/Llama-3-8B-Instruct-v0.1",
|
||||
"Magpie-Align/Llama-3.1-8B-Magpie-Align-v0.1",
|
||||
"prithivMLmods/Neumind-Math-7B-Instruct",
|
||||
"dphn/dolphin-2.9.3-qwen2-1.5b",
|
||||
"OpenBuddy/openbuddy-mistral-7b-v13",
|
||||
"kairawal/Qwen3-4B-TL-SynthDolly-1A-E3",
|
||||
"sahilnagaralu/movie-script",
|
||||
"xw1234gan/SFT_Qwen2.5-1.5B-Instruct_cnk12",
|
||||
"yunjae-won/ubq30i_qwen4b_sft_yl",
|
||||
"1010happy/qwen3BInstruct_ClaudeDefault",
|
||||
"yunjae-won/ubq30i_qwen4b_sft_both",
|
||||
"willieseun/AIMO-Qwen2.5-Math-1.5B-Instruct-Finetuned",
|
||||
"NousResearch/Yarn-Mistral-7b-64k",
|
||||
"xiaolesu/OsmosisProofling-GRPO-NT",
|
||||
"health360/Healix-410M",
|
||||
"RUC-AIBOX/STILL-3-1.5B-preview",
|
||||
"cjvt/GaMS-1B",
|
||||
"Lansechen/Qwen2.5-7B-Open-R1-GRPO-math-lighteval-1epochstop-withformat",
|
||||
"Charlie911/vicuna-7b-v1.5-general-temporal-merged",
|
||||
"Kyleyee/cDPO_hh-seed5",
|
||||
"prithivMLmods/Llama-3.2-3B-Math-Oct",
|
||||
"Kyleyee/HINGE_hh-seed5",
|
||||
"arcee-ai/Patent-Instruct-7b",
|
||||
"jb723/cross_lingual_epoch2",
|
||||
"davidkim205/komt-mistral-7b-v1",
|
||||
"kalisai/Nusantara-1.8b-Indo-Chat",
|
||||
"prithivMLmods/Llama-8B-Distill-CoT",
|
||||
"Kyleyee/HINGE_hh-seed3",
|
||||
"simplescaling/s1.1-1.5B",
|
||||
"sail/Sailor2-3B-SFT",
|
||||
"allenai/OLMoE-1B-7B-0924-Instruct",
|
||||
"prithivMLmods/Llama-3.2-6B-AlgoCode",
|
||||
"CloneBO/OracleLM",
|
||||
"NousResearch/Yarn-Solar-10b-32k",
|
||||
"m-a-p/OProver-8B-Base",
|
||||
"HuggingFaceH4/mistral-7b-sft-alpha",
|
||||
"Hyeongwon/P2-split2_prob_Qwen3-4B-Base_0312-01",
|
||||
"martyn/mixtral-megamerge-dare-8x7b-v1",
|
||||
"Kyleyee/rDPO_hh-seed4",
|
||||
"jingyeom/seal3.1.6n_7b",
|
||||
"sonthenguyen/OpenHermes-2.5-Mistral-7B-mt-bench-DPO-reversed_corrupted",
|
||||
"shibing624/ziya-llama-13b-medical-merged",
|
||||
"datajuicer/LLaMA-1B-dj-refine-150B",
|
||||
"ajibawa-2023/Uncensored-Jordan-7B",
|
||||
"nlpguy/AlloyIngot",
|
||||
"HuggingFaceTB/SmolLM3-3B",
|
||||
"allenai/codetulu-2-7b",
|
||||
"vihangd/dopeyshearedplats-2.7b-v1",
|
||||
"uukuguy/speechless-code-mistral-7b-v2.0",
|
||||
"j05hr3d/Llama-3.2-3B-Instruct-C_M_T-ALPACA",
|
||||
"cognitivetech/Mistral-7B-Inst-0.2-Bulleted-Notes",
|
||||
"sail/Qwen2.5-Math-1.5B-Oat-Zero",
|
||||
"reaperdoesntknow/SMOLM2Prover",
|
||||
"Kyleyee/ORPO_hh-seed5",
|
||||
"CarrotAI/Llama-3.2-Rabbit-Ko-3B-Instruct",
|
||||
"NousResearch/Nous-Capybara-7B-V1",
|
||||
"Vijay3548/InterviewMaster-Llama3.1",
|
||||
"ReviewHub/qwen3-4b-it-2507-sft-2018-2022-rl-step-20",
|
||||
"Nos-PT/Llama-Carvalho-PT",
|
||||
"Gille/StrangeMerges_49-7B-dare_ties",
|
||||
"Kyleyee/CPO_hh-seed2",
|
||||
"Kyleyee/cDPO_hh-seed3",
|
||||
"Kyleyee/DrDPO_hh-seed2",
|
||||
"sthenno-com/miscii-14b-0218",
|
||||
"mtgv/MobileLLaMA-2.7B-Chat",
|
||||
"Novaciano/Alice_In_The_Dark_2-Slerp-RP-3.2-1B",
|
||||
"Himitsui/KuroMitsu-11B",
|
||||
"owlninjam/nytheria-3b",
|
||||
"Kyleyee/DrDPO_hh-seed4",
|
||||
"Kyleyee/DrDPO_hh-seed5",
|
||||
"shahzebnaveed/NeuralHermes-2.5-Mistral-7B",
|
||||
"plaguss/mistal-7b-prm-openrlhf",
|
||||
"viethq188/Rabbit-7B-v2-DPO-Chat",
|
||||
"EmbeddedLLM/Mistral-7B-Merge-14-v0.3-ft-step-9984",
|
||||
"dbpedia/nspm-starcoder-1b",
|
||||
"psh3333/llama-3.2-3b-grpo-merged",
|
||||
"saarvajanik/facebook-opt-6.7b-qcqa-ub-16-best-for-KV-cache",
|
||||
"EmbeddedLLM/Mistral-7B-Merge-14-v0.3",
|
||||
"xformAI/facebook-opt-125m-qcqa-ub-6-best-for-KV-cache",
|
||||
"Rev124/llama-3-pruned",
|
||||
"RatanRohith/NeuralPizza-7B-V0.3",
|
||||
"mncai/DPO_BC_partial_epoch6",
|
||||
"zeemen2723/museai-lyrics-gen",
|
||||
"automerger/OgnoExperiment27-7B",
|
||||
"sohamb37lexsi/qwen25-3b-legal-correction",
|
||||
"swift/Meta-Llama-3-8B",
|
||||
"m-a-p/MuPT-v0-8192-190M",
|
||||
"Kquant03/Samlagast-7B-laser-bf16",
|
||||
"VTSNLP/Llama3-ViettelSolutions-8B",
|
||||
"vihangd/dopeyshearedplats-1.3b-v1",
|
||||
"HCY123902/qwen25_7b_base_hc_tsss_n32_r1_dpo",
|
||||
"coder3101/gemma-3-1b-it-heretic",
|
||||
"Hemkant04/qwen05-resume-job-match-evaluator",
|
||||
"yekon9/Qwen3-4B-Instruct-2507-heretic",
|
||||
"uukuguy/Orca-2-13b-f16",
|
||||
"Kquant03/NeuralTrix-7B-dpo-relaser",
|
||||
"sambanovasystems/SambaLingo-Russian-Base",
|
||||
"bineric/NorskGPT-Mistral-7b",
|
||||
"ReviewHub/qwen3-4b-it-2507-sft-2018-2022-rl-step-10",
|
||||
"beyoru/EvolLLM",
|
||||
"driaforall/Dria-Agent-a-3B",
|
||||
"tlphams/zoyllm-7b-slimorca",
|
||||
"QuixiAI/WizardLM-33B-V1.0-Uncensored",
|
||||
"bilalRahib/TinyLLama-NSFW-Chatbot",
|
||||
"Ahatsham/Llama-3-8B-Instruct_Planning_Feedback_oldaug_v2",
|
||||
"unsloth/Qwen2.5-Math-1.5B",
|
||||
"wang7776/Llama-2-7b-chat-hf-30-sparsity",
|
||||
"hector-gr/RLCR-v4-ks-uniqueness-buf5k-hotpot",
|
||||
"m-a-p/neo_7b_instruct_v0.1",
|
||||
"p208p2002/llama-traditional-chinese-120M",
|
||||
"tartuNLP/Llammas-base",
|
||||
"riotu-lab/ArabianGPT-01B",
|
||||
"openaccess-ai-collective/DPOpenHermes-7B",
|
||||
"Enxin/MovieChat-vicuna",
|
||||
"Skywork/Skywork-OR1-Math-7B",
|
||||
"Lugha-Llama/Lugha-Llama-8B-wura_edu",
|
||||
"laion/allenai-sera-unified-316__Qwen3-8B",
|
||||
"T1anyu/DeepInnovator",
|
||||
"mideind/icelandic-gpt-sw3-6.7b-gec",
|
||||
"Rayeeennnnnnnn/legalmind-chatbot",
|
||||
"Wanfq/FuseLLM-7B",
|
||||
"mrvinph/Qwen2.5-0.5B-Instruct-Gensyn-Swarm-placid_wily_woodpecker",
|
||||
"shivanikerai/Llama-2-7b-chat-hf-title-ner-and-title-suggestions-v2.0",
|
||||
"Invalid-Null/PeiYangMe-0.7",
|
||||
"open-unlearning/unlearn_tofu_Llama-3.2-1B-Instruct_forget10_GradDiff_lr1e-05_alpha5_epoch5",
|
||||
"Tesslate/Tessa-T1-3B",
|
||||
"LorenaYannnnn/unsafe_compliance-Qwen3-0.6B-OURS_self-seed_1",
|
||||
"HiTZ/latxa-7b-v1",
|
||||
"huggyllama/llama-7b",
|
||||
"FlyPig23/Llama3.2-3B_Paper_Impact_media_SFT_1ep",
|
||||
"hongzhouyu/FineMedLM-o1",
|
||||
"shisa-ai/shisa-v1-llama3-8b.lr-5e6",
|
||||
"senseable/Westlake-7B",
|
||||
"simplescaling/s1.1-3B",
|
||||
"Borjan/finki-gpt-140M",
|
||||
"TinyLlama/TinyLlama_v1.1_chinese",
|
||||
"Manirajan/interview_tiny",
|
||||
"ajibawa-2023/Young-Children-Storyteller-Mistral-7B",
|
||||
"TinyLlama/TinyLlama_v1.1_math_code",
|
||||
"manotham/Thai-dialogue-transalate_sft_80K",
|
||||
"nassimjp/Maral-7B-alpha-1",
|
||||
"MSL7/INEX16-7b",
|
||||
"Rayeeennnnnnnn/mizan-legal-tunisian",
|
||||
"maheshrawat18/Qwen3-4B-2507-sft1",
|
||||
"prithivMLmods/Theta-Crucis-0.6B-Turbo1",
|
||||
"mlabonne/llama-2-7b-miniguanaco",
|
||||
"FreekCoolAI/privacy-gemma-qlora",
|
||||
"lex-hue/Delexa-V0.1-7b",
|
||||
"goldfish-models/tur_latn_100mb",
|
||||
"shisa-ai/ablation-18-rafbestseq-shisa-v2-llama-3.1-8b-lr8e6",
|
||||
"yash-lulla/Legal_AI_Assistant",
|
||||
"prithivMLmods/Deepthink-Llama-3-8B-Preview",
|
||||
"prithivMLmods/QwQ-R1-Distill-7B-CoT",
|
||||
"sstoica12/acquisition_metamath_llama_instruct-3_1-8b-math_format_500_combined_openr1math",
|
||||
"goldfish-models/rus_cyrl_1000mb",
|
||||
"goldfish-models/ukr_cyrl_1000mb",
|
||||
"eren23/dpo-binarized-NeutrixOmnibe-7B",
|
||||
"RAANA-IA/Gheya-med",
|
||||
"mesolitica/malaysian-tinyllama-1.1b-16k-instructions-v2",
|
||||
"bue0912/ToolOmni-Qwen3-4B",
|
||||
"BRlkl/distill-sft-qwen3-4b-full",
|
||||
"prithivMLmods/PocketThinker-QwQ-3B-Instruct",
|
||||
"prithivMLmods/Megatron-Bots-1.7B-Reasoning",
|
||||
"kumarprince070107/geobot",
|
||||
"ClaudioSavelli/FAME_GA_llama32-1b-instruct-qa",
|
||||
"kairawal/Qwen3-0.6B-EL-SynthDolly-1A-E8",
|
||||
"prithivMLmods/Poseidon-Reasoning-1.7B",
|
||||
"PKU-Alignment/ProgressGym-HistLlama3-8B-C014-instruct-v0.2",
|
||||
"unsloth/Qwen2-7B",
|
||||
"prithivMLmods/Open-Xi-Math-Preview",
|
||||
"MiniLLM/teacher-gpt2-1.5B",
|
||||
"Rakancorle1/qwen2.5-7b_Instruct_policy_traj_30k_full",
|
||||
"prithivMLmods/Lang-Exster-0.5B-Instruct",
|
||||
"prithivMLmods/Nenque-MoT-0.6B-Elite14",
|
||||
"posicube/Llama2-chat-AYT-13B",
|
||||
"lomahony/pythia-70m-helpful-sft",
|
||||
"kyubeen/code-grpo-checkpoint-950",
|
||||
"nicholasKluge/TeenyTinyLlama-160m",
|
||||
"allenai/Llama-3.1-Tulu-3-8B",
|
||||
"BarraHome/Mistroll-7B-v2.2",
|
||||
"posicube/Llama-chat-AY-13B",
|
||||
"hongli-zhan/MINT-empathy-Qwen3-4B",
|
||||
"jhhj25/qwen3-moe-neuron_structure_drop-p50-s1k-128samples-sft",
|
||||
"Ikonz-Studios/seva-sarathi-intent-qwen3-1.7b",
|
||||
"NECOUDBFM/Jellyfish-8B",
|
||||
"gradients-io-tournaments/augmented-1db17e1d682d23fd",
|
||||
"Locutusque/LocutusqueXFelladrin-TinyMistral248M-Instruct",
|
||||
"ChuGyouk/R16",
|
||||
"oveja1122/toolcalling-merged-demo",
|
||||
"kmseong/llama3_2_3b-instruct-math-safedelta-scale0.8",
|
||||
"EleutherAI/SmolLM2-1.7B-magpie-ultra-v1.0-math-431k-s",
|
||||
"paulml/NeuralOmniBeagleMBX-v3-7B",
|
||||
"skemessage/Qwen2.5-7B-Instruct-neuron",
|
||||
BIREN_MODELS = [
|
||||
"BigRatz/LOL-AI-2026",
|
||||
"aimeri/spoomplesmaxx-cardmaker-v1",
|
||||
"Alibaba-DT/Logics-STEM-8B-SFT",
|
||||
"Muneebmn123/insurance-voice-qwen25-1_5b",
|
||||
"AnkitBirGurung/NEMO-12B-SFT-Further",
|
||||
"danilarudenko/editorai-mini",
|
||||
"EphemeralYou/Prompt-Refine-MiniCPM5-1B",
|
||||
"mtepe01/mentorx-mistral-7b-automata-merged",
|
||||
"DarkArtsForge/Helix-SCE-12B-jh",
|
||||
"rpant/iolai26-solve",
|
||||
"NithinAI12/NithinX-Omni-LLM-v1",
|
||||
"ConvexAI/Luminex-34B-v0.2",
|
||||
"codellama/CodeLlama-34b-hf",
|
||||
"Lipas007/iol-ai-2026-qwen14b-awq",
|
||||
"D-Z-W/finetuned-teacher",
|
||||
"huan1999/ziya-llama-13b-medical-merged",
|
||||
"codellama/CodeLlama-34b-Python-hf",
|
||||
"ld4ad/gemma-2-9b-dunhuang",
|
||||
"harindhar10/Olmo-7b_1M_Smiles_lora",
|
||||
"allenai/Olmo-3-7B-Think-DPO",
|
||||
"allenai/Olmo-3-32B-Think-DPO",
|
||||
"RedHatAI/gemma-2-9b-it",
|
||||
"prashanthsura/gemma-2-2b-legal-financial-sft",
|
||||
"pfnet/plamo-2-8b",
|
||||
"sail/Sailor2-20B-128K-SFT",
|
||||
"facebook/layerskip-llama3.2-1B",
|
||||
"Qwen/Qwen2.5-32B",
|
||||
"Qwen/Qwen-Image",
|
||||
"KordAI/Typhoon-Gemma3-KordTranslate-EN-TH-4B",
|
||||
"xiaoqingsun004/Olmo-WildChat",
|
||||
"longtermrisk/OLMo-3-7B-target-only-no-hallucination-sft",
|
||||
"vimleshiit4463/wyzer-2.0-smollm2-135m",
|
||||
"trl-lib/pythia-1b-deduped-tldr-sft",
|
||||
]
|
||||
|
||||
CAMBRICON_MODELS = [
|
||||
"clear-blue-sky/evolai-reborn-tfm-008",
|
||||
"clear-blue-sky/evolai-reborn-tfm-010",
|
||||
"BigRatz/LOL-AI-2026",
|
||||
"clear-blue-sky/evolai-reborn-tfm-001",
|
||||
"clear-blue-sky/evolai-reborn-tfm-019",
|
||||
"clear-blue-sky/evolai-reborn-tfm-007",
|
||||
"aimeri/spoomplesmaxx-cardmaker-v1",
|
||||
"aipatseer/inst_ft_qwen_0.6b_summ",
|
||||
"clear-blue-sky/evolai-reborn-tfm-003",
|
||||
"clear-blue-sky/evolai-reborn-tfm-004",
|
||||
"Alibaba-DT/Logics-STEM-8B-SFT",
|
||||
"clear-blue-sky/evolai-reborn-tfm-002",
|
||||
"prism-ml/Ternary-Bonsai-4B-unpacked",
|
||||
"Muneebmn123/insurance-voice-qwen25-1_5b",
|
||||
"AnkitBirGurung/NEMO-12B-SFT-Further",
|
||||
"danilarudenko/editorai-mini",
|
||||
"rpant/iolai26-solve",
|
||||
]
|
||||
|
||||
METAX_MODELS = [
|
||||
"Phoenix9781/evolai-tf-model-105",
|
||||
"clear-blue-sky/evolai-reborn-tfm-008",
|
||||
"clear-blue-sky/evolai-reborn-tfm-010",
|
||||
"BigRatz/LOL-AI-2026",
|
||||
"clear-blue-sky/evolai-reborn-tfm-001",
|
||||
"clear-blue-sky/evolai-reborn-tfm-019",
|
||||
"clear-blue-sky/evolai-reborn-tfm-007",
|
||||
"clear-blue-sky/evolai-reborn-tfm-005",
|
||||
"Lin2es/evolai-tfm-03o",
|
||||
"clear-blue-sky/evolai-reborn-tfm-009",
|
||||
"aimeri/spoomplesmaxx-cardmaker-v1",
|
||||
"aipatseer/inst_ft_qwen_0.6b_summ",
|
||||
"reaperdoesntknow/DualMind-TKD-Agentic-1.7B",
|
||||
"clear-blue-sky/evolai-reborn-tfm-011",
|
||||
"clear-blue-sky/evolai-reborn-tfm-003",
|
||||
"clear-blue-sky/evolai-reborn-tfm-004",
|
||||
"clear-blue-sky/evolai-reborn-tfm-002",
|
||||
"amd/ReasonLite-0.6B",
|
||||
"prism-ml/Ternary-Bonsai-4B-unpacked",
|
||||
"Muneebmn123/insurance-voice-qwen25-1_5b",
|
||||
"danilarudenko/editorai-mini",
|
||||
"rpant/iolai26-solve",
|
||||
]
|
||||
|
||||
KUNLUNXIN_MODELS = [
|
||||
"Patriae/patriae-cuban-dialect-model",
|
||||
"Phoenix9781/evolai-tf-model-105",
|
||||
"clear-blue-sky/evolai-reborn-tfm-008",
|
||||
"clear-blue-sky/evolai-reborn-tfm-010",
|
||||
"clear-blue-sky/evolai-reborn-tfm-001",
|
||||
"clear-blue-sky/evolai-reborn-tfm-019",
|
||||
"clear-blue-sky/evolai-reborn-tfm-007",
|
||||
"clear-blue-sky/evolai-reborn-tfm-005",
|
||||
"Lin2es/evolai-tfm-03o",
|
||||
"clear-blue-sky/evolai-reborn-tfm-009",
|
||||
"aimeri/spoomplesmaxx-cardmaker-v1",
|
||||
"aipatseer/inst_ft_qwen_0.6b_summ",
|
||||
"clear-blue-sky/evolai-reborn-tfm-011",
|
||||
"clear-blue-sky/evolai-reborn-tfm-003",
|
||||
"clear-blue-sky/evolai-reborn-tfm-004",
|
||||
"clear-blue-sky/evolai-reborn-tfm-002",
|
||||
"amd/ReasonLite-0.6B",
|
||||
"prism-ml/Ternary-Bonsai-4B-unpacked",
|
||||
"EthanGao123/CellHermes-v1.0",
|
||||
"RecursiveMAS/Mixture-Science-BioMistral-7B",
|
||||
"chenyitian-shanshu/SIRL-Gurobi",
|
||||
"RedHatAI/gemma-2-9b-it",
|
||||
"gaunernst/gemma-3-27b-it-qat-autoawq",
|
||||
"promotion/qwen3-8b-simpo-avg-b2p5-g1p0-s42",
|
||||
"llamaindex/vdr-2b-multi-v1",
|
||||
"KordAI/Typhoon-Gemma3-KordTranslate-EN-TH-4B",
|
||||
"TheDrummer/UnslopNemo-12B-v3",
|
||||
"gradients-io-tournaments/tournament-tourn_c5d86c82ce819a79_20260706-b78a01d4-0a6a-49e1-9190-5e88ae329937-5DS6XMVr",
|
||||
"ArliAI/Mistral-Nemo-12B-ArliAI-RPMax-v1.1",
|
||||
]
|
||||
|
||||
# 按顺序处理：Biren → Cambricon → MetaX → Kunlunxin
|
||||
GPU_JOBS: List[Tuple[str, List[str]]] = [
|
||||
("Biren_166m", BIREN_MODELS),
|
||||
("Cambricon_mlu-370-x8", CAMBRICON_MODELS),
|
||||
("MetaX_c-500", METAX_MODELS),
|
||||
("Kunlunxin_p-800", KUNLUNXIN_MODELS),
|
||||
]
|
||||
TOTAL_MODELS = sum(len(models) for _, models in GPU_JOBS)
|
||||
|
||||
# ══════════════════════════════════════════════════════════
|
||||
# 全局状态（供 /status 展示）
|
||||
# ══════════════════════════════════════════════════════════
|
||||
_state = {
|
||||
"strategy_id": STRATEGY_ID,
|
||||
"phase": "starting", # starting | submitting | done | error
|
||||
"total": len(ALL_MODEL_IDS),
|
||||
"total": TOTAL_MODELS,
|
||||
"submitted": 0,
|
||||
"failed": 0,
|
||||
"per_gpu": {gpu: 0 for gpu, _ in GPU_JOBS},
|
||||
"started_at": None,
|
||||
"finished_at": None,
|
||||
}
|
||||
@@ -311,17 +207,43 @@ def _run_http():
|
||||
print("[http] 已关闭", flush=True)
|
||||
|
||||
# ══════════════════════════════════════════════════════════
|
||||
# 业务逻辑
|
||||
# 各 GPU 的 config_content 模板
|
||||
# ══════════════════════════════════════════════════════════
|
||||
def _submit_task(token: str, model_id: str) -> Tuple[bool, str]:
|
||||
headers = {
|
||||
"Content-Type": "application/json",
|
||||
"Authorization": f"Bearer {token}",
|
||||
}
|
||||
config_content = f"""gpu_type: ppu_zw_810e
|
||||
def build_config_content(gpu_type: str, model_id: str) -> str:
|
||||
if gpu_type == "Biren_166m":
|
||||
max_model_len = 4096
|
||||
return f"""docker_image: git.modelhub.org.cn:9443/enginex/xc-llm-biren166m:26.01
|
||||
nv_docker_image: harbor.4pd.io/dooke/vllm/vllm/vllm-openai:v0.11.0
|
||||
framework: vllm
|
||||
docker_image: harbor.4pd.io/hardcore-tech/asllm:1.10.1-pytorch2.10.0-ubuntu24.04-sail2.1.0-cuda13.0-sglang0.5.10-vllm0.19.0-py312
|
||||
nv_docker_image: harbor-contest.4pd.io/sunruoxi/vllm-openai-fix-tokenizer:v0.11.0
|
||||
lang: zh
|
||||
storage: gpfs
|
||||
api: completion
|
||||
modelhub_options:
|
||||
srcRelativePath: leaderboard/modelHubXC/{model_id}
|
||||
mountPoint: /model
|
||||
max_model_len: {max_model_len}
|
||||
sut_config:
|
||||
values:
|
||||
gpu_num: 1
|
||||
env:
|
||||
- name: MAX_MODEL_LEN
|
||||
value: {max_model_len}
|
||||
command: ['/bin/bash', '-ic', 'vllm serve /model --port 8000 --served-model-name llm --max-model-len {max_model_len} --gpu-memory-utilization 0.9 --enforce-eager --trust-remote-code -tp 1 --host 0.0.0.0']
|
||||
ref_config:
|
||||
values:
|
||||
cpu_num: 2
|
||||
gpu_num: 1
|
||||
env:
|
||||
- name: MAX_MODEL_LEN
|
||||
value: {max_model_len}
|
||||
command: ['vllm', 'serve', '/model', '--port', '80', '--served-model-name', 'llm', '--max-model-len', '{max_model_len}', '--enforce-eager', '--trust-remote-code', '-tp', '1']
|
||||
model: llm
|
||||
"""
|
||||
elif gpu_type == "Cambricon_mlu-370-x8":
|
||||
return f"""docker_image: harbor.4pd.io/hardcore-tech/cambricon-mlu370-pytorch:v25.01-torch2.5.0-torchmlu1.24.1-ubuntu22.04-py310
|
||||
nv_docker_image: harbor.4pd.io/dooke/vllm/vllm/vllm-openai:v0.11.0
|
||||
framework: vllm
|
||||
storage: gpfs
|
||||
modelhub_options:
|
||||
srcRelativePath: leaderboard/modelHubXC/{model_id}
|
||||
mountPoint: /model
|
||||
@@ -329,60 +251,93 @@ sut_config:
|
||||
values:
|
||||
gpu_num: 1
|
||||
env:
|
||||
- name: test
|
||||
value: fp16
|
||||
command:
|
||||
- bash
|
||||
- /opt/t-head/entrypoint.sh
|
||||
- python3
|
||||
- -m
|
||||
- asllm.entrypoints.api_server
|
||||
- --model
|
||||
- /model
|
||||
- --port
|
||||
- '30000'
|
||||
- --host
|
||||
- 0.0.0.0
|
||||
- --served-model-name
|
||||
- llm
|
||||
- name: MAX_MODEL_LEN
|
||||
value: 8192
|
||||
command: ["vllm", "serve", "/model", "--port", "8000", "--served-model-name", "llm", "--max-model-len", "8192", "--trust-remote-code", "--dtype", "float16"]
|
||||
ref_config:
|
||||
values:
|
||||
cpu_num: 2
|
||||
gpu_num: 1
|
||||
env:
|
||||
- name: test
|
||||
value: fp16
|
||||
command:
|
||||
- vllm
|
||||
- serve
|
||||
- /model
|
||||
- --port
|
||||
- '80'
|
||||
- --served-model-name
|
||||
- llm
|
||||
- --max-model-len
|
||||
- '2048'
|
||||
- --gpu-memory-utilization
|
||||
- '0.9'
|
||||
- --enforce-eager
|
||||
- --trust-remote-code
|
||||
- -tp
|
||||
- '1'
|
||||
- name: MAX_MODEL_LEN
|
||||
value: 8192
|
||||
command: ["vllm", "serve", "/model", "--port", "80", "--served-model-name", "llm", "--max-model-len", "8192", "--trust-remote-code", "--dtype", "float16"]
|
||||
"""
|
||||
elif gpu_type == "MetaX_c-500":
|
||||
return f"""docker_image: git.modelhub.org.cn:9443/enginex-metax/vllm:0.9.1
|
||||
nv_docker_image: harbor.4pd.io/dooke/vllm/vllm/vllm-openai:v0.11.0
|
||||
framework: vllm
|
||||
lang: en
|
||||
storage: gpfs
|
||||
api: chat
|
||||
modelhub_options:
|
||||
srcRelativePath: leaderboard/modelHubXC/{model_id}
|
||||
mountPoint: /model
|
||||
max_model_len: 2048
|
||||
sut_config:
|
||||
gpu_num: 1
|
||||
values:
|
||||
command: ['/opt/conda/bin/vllm', 'serve', '/model', '--port', '20644', '--served-model-name', 'llm', '--max-model-len', '2048', '--gpu-memory-utilization', '0.9', '--enforce-eager', '--trust-remote-code' ,'-tp', '1']
|
||||
ref_config:
|
||||
gpu_num: 1
|
||||
values:
|
||||
command: ['vllm', 'serve', '/model', '--port', '80', '--served-model-name', 'llm', '--max-model-len', '2048', '--enforce-eager', '--trust-remote-code', '-tp', '1']
|
||||
"""
|
||||
elif gpu_type == "Kunlunxin_p-800":
|
||||
return f"""docker_image: git.modelhub.org.cn:9443/enginex/xc-llm-kunlun
|
||||
nv_docker_image: harbor.4pd.io/dooke/vllm/vllm/vllm-openai:v0.11.0
|
||||
framework: vllm
|
||||
lang: en
|
||||
storage: gpfs
|
||||
api: chat
|
||||
temperature: 0.4
|
||||
repetition_penalty: 1.1
|
||||
top_p: 0.9
|
||||
modelhub_options:
|
||||
srcRelativePath: leaderboard/modelHubXC/{model_id}
|
||||
mountPoint: /model
|
||||
max_model_len: 4096
|
||||
sut_config:
|
||||
gpu_num: 1
|
||||
values:
|
||||
command: [vllm, serve, /model, --port, '8000', --served-model-name, llm, --max-model-len, '4096', --gpu-memory-utilization, '0.9', --enforce-eager, --trust-remote-code, -tp, '1']
|
||||
ref_config:
|
||||
gpu_num: 1
|
||||
values:
|
||||
command: [vllm, serve, /model, --port, '80', --served-model-name, llm, --max-model-len, '4096', --enforce-eager, --trust-remote-code, -tp, '1']
|
||||
"""
|
||||
else:
|
||||
raise ValueError(f"未知的 GPU_TYPE: {gpu_type}")
|
||||
|
||||
|
||||
|
||||
|
||||
|
||||
# ══════════════════════════════════════════════════════════
|
||||
# 业务逻辑
|
||||
# ══════════════════════════════════════════════════════════
|
||||
def _submit_task(token: str, gpu_type: str, model_id: str) -> Tuple[bool, str]:
|
||||
headers = {
|
||||
"Content-Type": "application/json",
|
||||
"Authorization": f"Bearer {token}",
|
||||
}
|
||||
config_content = build_config_content(gpu_type, model_id)
|
||||
|
||||
payload = {
|
||||
"contestApiToken": CONTEST_API_TOKEN,
|
||||
"contributors": CONTRIBUTORS,
|
||||
"gpuTypes": [GPU_TYPE],
|
||||
"gpuTypes": [gpu_type],
|
||||
"taskType": TASK_TYPE,
|
||||
"modelId": model_id,
|
||||
"framework": "vllm",
|
||||
"strategyId": STRATEGY_ID, # 平台要求
|
||||
"submissionConfig": [{
|
||||
"config": config_content,
|
||||
"gpuType": GPU_TYPE,
|
||||
"gpuType": gpu_type,
|
||||
"taskType": TASK_TYPE,
|
||||
}],
|
||||
}
|
||||
print(f"[payload] {json.dumps(payload, indent=2, ensure_ascii=False)}", flush=True)
|
||||
print(f"[payload] gpu={gpu_type} model={model_id}", flush=True)
|
||||
try:
|
||||
resp = requests.post(
|
||||
BASE_URL + SUBMIT_ENDPOINT,
|
||||
@@ -393,13 +348,13 @@ ref_config:
|
||||
result = resp.json()
|
||||
if result.get("code") == 0:
|
||||
task_id = result.get("data", {}).get("id", "")
|
||||
print(f"[worker] OK {model_id} task_id={task_id}", flush=True)
|
||||
print(f"[worker] OK {model_id} (GPU={gpu_type}) task_id={task_id}", flush=True)
|
||||
return True, task_id
|
||||
else:
|
||||
print(f"[worker] FAIL {model_id}: {result.get('message')}", flush=True)
|
||||
print(f"[worker] FAIL {model_id} (GPU={gpu_type}): {result.get('message')}", flush=True)
|
||||
return False, ""
|
||||
except Exception as e:
|
||||
print(f"[worker] ERROR {model_id}: {e}", flush=True)
|
||||
print(f"[worker] ERROR {model_id} (GPU={gpu_type}): {e}", flush=True)
|
||||
return False, ""
|
||||
|
||||
|
||||
@@ -407,32 +362,39 @@ def _run_worker():
|
||||
_state["started_at"] = datetime.utcnow().isoformat()
|
||||
_state["phase"] = "submitting"
|
||||
|
||||
successful: List[Tuple[str, str]] = []
|
||||
successful: List[Tuple[str, str, str]] = []
|
||||
token = AUTH_TOKEN
|
||||
print("[worker] 使用预设 Token，跳过登录", flush=True)
|
||||
|
||||
for model_id in ALL_MODEL_IDS:
|
||||
for gpu_type, model_list in GPU_JOBS:
|
||||
if _shutdown.is_set():
|
||||
break
|
||||
ok, task_id = _submit_task(token, model_id)
|
||||
if ok:
|
||||
_state["submitted"] += 1
|
||||
successful.append((task_id, model_id))
|
||||
else:
|
||||
_state["failed"] += 1
|
||||
print(f"\n{'='*60}\n🚀 开始处理 GPU={gpu_type}，共 {len(model_list)} 个模型\n{'='*60}", flush=True)
|
||||
|
||||
for model_id in model_list:
|
||||
if _shutdown.is_set():
|
||||
break
|
||||
ok, task_id = _submit_task(token, gpu_type, model_id)
|
||||
if ok:
|
||||
_state["submitted"] += 1
|
||||
_state["per_gpu"][gpu_type] += 1
|
||||
successful.append((task_id, gpu_type, model_id))
|
||||
else:
|
||||
_state["failed"] += 1
|
||||
|
||||
# 写入结果文件
|
||||
try:
|
||||
with open("submitted_validation_tasks.txt", "w", encoding="utf-8") as f:
|
||||
for tid, mid in successful:
|
||||
f.write(f"{tid}\t{mid}\n")
|
||||
for tid, gpu, mid in successful:
|
||||
f.write(f"{tid}\t{gpu}\t{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']}",
|
||||
f"[worker] 完成 submitted={_state['submitted']} failed={_state['failed']} "
|
||||
f"total={_state['total']} per_gpu={_state['per_gpu']}",
|
||||
flush=True,
|
||||
)
|
||||
# 提交完成后继续保持进程存活，等待平台停止
|
||||
@@ -465,4 +427,4 @@ def main():
|
||||
|
||||
|
||||
if __name__ == "__main__":
|
||||
main()
|
||||
main()
|
||||
|
||||
Reference in New Issue
Block a user