diff --git a/main.py b/main.py index 3a79da3..0ae1a93 100644 --- a/main.py +++ b/main.py @@ -1,12 +1,13 @@ """ xc_validation_strategy_vllm_zhouyuanxi — 主入口 -启动后针对 3 张 GPU 卡（Kunlunxin_p-800 / Biren_166m / Cambricon_mlu-370-x8） -分别批量提交各自筛选出的模型适配任务（/api/adapt/task/add，xc-Token 认证）。 +启动后针对 4 张 GPU 卡（MetaX_c-500 / Kunlunxin_p-800 / Biren_166m / +Cambricon_mlu-370-x8）分别批量提交各自筛选出的模型适配任务 +（/api/adapt/task/add，xc-Token 认证）。 -提交账号采用自动 fallback 轮转：优先用 zhouyuanxi 账号提交，一旦该账号命中 +提交账号采用自动 fallback 轮转：按 ACCOUNTS 列表顺序提交，一旦当前账号命中 平台的"异步验证任务数量已达上限（100）"限制（错误码 60007），自动切换到下一个 -账号（jiajing → fanyi）继续提交同一个模型，直至全部账号额度用尽。 +账号继续提交同一个模型，直至全部账号额度用尽。 之后保持 HTTP 服务存活，暴露 /health（K8s 探活）和 /status（运行状态）。 """ @@ -34,264 +35,467 @@ HTTP_PORT = 8080 # 提交账号（按优先级排列，前一个额度满了自动切换到下一个） ACCOUNTS: List[Tuple[str, str, str]] = [ - ("zhouyuanxi", "i-zhouyuanxi@4paradigm.com", "62b9b487eff2488fb9f1da0b963f0b93"), - ("jiajing", "jiajing", "5e051e0ff8384a81af53bea780deb28a"), - ("fanyi", "fanyi", "f2d501c9ae6543a589cd6cb789108c41"), + ("zhouyuanxi", "i-zhouyuanxi@4paradigm.com", "62b9b487eff2488fb9f1da0b963f0b93"), + ("jiajing", "jiajing", "5e051e0ff8384a81af53bea780deb28a"), + ("fanyi", "fanyi", "f2d501c9ae6543a589cd6cb789108c41"), + ("miaoyao", "miaoyao", "77033cee0fb549598cdd590be0d02983"), + ("zhoukaile", "zhoukaile", "bd7c52f3b9604ef48a14dd6174513935"), + ("fanyi2", "fanyi2", "2586efe06c0a42fda060d5eca34bf766"), + ("l112233", "l112233", "40cb6910dc9a442a816298a228da65ac"), + ("l11223344", "l11223344", "e1c0db2959e5411f9342c8550b03f6e9"), + ("keii", "keii", "be99003a85f640d8978823a5a8e3f297"), + ("zhangyuanxi", "zhangyuanxi", "24ed39f7f0d84fafbe0ca808e62b191c"), + ("jiangxiaowen", "jiangxiaowen", "88d5fee9f1fe4f7583f11a9d3702dc85"), ] # ══════════════════════════════════════════════════════════ # 各 GPU 的模型列表（来自 filter_verified_models 脚本的筛选结果） # ══════════════════════════════════════════════════════════ +METAX_MODELS = [ + "EphemeralYou/Prompt-Refine-MiniCPM5-1B", + "NousResearch/Hermes-4-14B", + "ToxicityPrompts/PolyGuard-Ministral", + "Adithyaaaa/chemistry-mistral-7b-v0.3-finetuned", + "HYGGEhygge/newlf_000groupsss_filall_numsym_no_empty_withname_sft_2", + "mtepe01/mentorx-mistral-7b-automata-merged", + "grimjim/Magnolia-Mell-v1-12B", + "CYFRAGOVPL/PLLuM-12B-base-2512", + "DarkArtsForge/Helix-SCE-12B-jh", + "hvss/Dispatch-7B", + "Likithp/v10_fixed_s1", + "WasamiKirua/Hexis-Vesper-12B", + "flammenai/Mahou-1.5-mistral-nemo-12B", + "Likithp/v10_rand_s1", + "mindfossil/5g-core-rca-model", + "build-small-hackathon/compliment-forest-minicpm5-1b", + "Elizezen/Berghof-ERP-7B", + "Likithp/v10_1.5B_fixed_s42", + "Mohamed475/qwen3-1.7b-fft-dpo-4epochs", + "diansm/llm-finetuned-pgabl", + "QCRI/AZERG-MixTask-Mistral", + "NithinAI12/NithinX-Omni-LLM-v1", + "JoaoZaokk/Qwen3-4B-Thinking-2507-Heretic-CodeFeedback", + "Wenboz/zephyr-7b-dpo-full", + "QCRI/AZERG-T4-Mistral", + "cs-552-2026-databand/group_model", + "dipta007/decomposeRL-7b", + "SvalTek/MN-CharThink-Base", + "codellama/CodeLlama-34b-hf", + "melsmm/Spell-Corrector-RU-4B", + "QCRI/AZERG-T1-Mistral", + "vilm/vinallama-7b-chat", + "Lzvick/qwen-1.7b-math-reasoner-grpo", + "kosiasuzu/agenticml-agent-llama-3.1-8b-init", + "Lipas007/iol-ai-2026-qwen14b-awq", + "vclmax/nemo-12b-expansion-v1", + "jithinjames/iol-ai-2026-solver", + "Vortex5/Silver-Siren-12B", + "ArliAI/Mistral-Nemo-12B-ArliAI-RPMax-v1.2", + "kosiasuzu/chatml-agent-llama-3.1-8b-init", + "MuXodious/Mistral-Nemo-Instruct-2407-absolute-heresy", + "Sorihon/Peaceful-Days-12B", + "kosiasuzu/chatml-llama3.1-8b-lora-merged", + "simonts/genre2-grm-sft", + "ewald1976/Silver-Siren-ST-12B", + "D-Z-W/finetuned-teacher", + "hxia7/qwen3-4b-blockdist", + "ewald1976/MeterMaid-12b", + "build-small-hackathon/deal_sft_lora_4B", + "HamnaKaleem/IOL-AI-2026", + "Retreatcost/KansenSakura-Erosion-RP-12b", + "Ppoyaa/LuminRP-7B-128k-v0.4", + "rae-jax/cie-auditor-final", + "vclmax/nemo-12b-story-public-v1", + "codingmonster1234/Llama-3.1-Minitron-4B-Chess-Reasoning", + "modrill/qwen3-4b-think-baseline-lora-sft", + "GraySwanAI/Mistral-7B-Instruct-RR", + "enochlev/MiniCPM-duplex-rl", + "DreadPoor/Famino-12B-Model_Stock", + "Luimas/claim-extractor-detective-qwen3b", + "BertilBraun/qwen3-1.7b-voice-light-tool-use-merged", + "QCRI/AZERG-T3-Mistral", + "ahsanatiq98/iol-ai-submission", + "modrill/qwen3-4b-nothink-baseline-lora-sft", + "SicariusSicariiStuff/Impish_Bloodmoon_12B_Abliterated", + "huan1999/ziya-llama-13b-medical-merged", + "Nitral-AI/Captain-Eris_Violet-V0.420-12B", + "minhtt/vistral-7b-chat", + "codellama/CodeLlama-34b-Python-hf", + "modrill/qwen3-4b-think-baseline-full-sft", + "kcherry497/dyno-blast-4b", + "ld4ad/gemma-2-9b-dunhuang", + "harindhar10/Olmo-7b_1M_Smiles_lora", + "ishikauniphore/multilingual_reasoner_multilingual_cot", + "DavidAU/granite-4.1-8b-Claude-Opus-4.6-Thinking-MAX", + "Secbone/llama-33B-instructed", + "Irfanuruchi/Qwen3-4B-Computer-Science", + "carolinezx/llama-8b-sft-preferred-cleaned", + "davidanugraha/Qwen3-4B-Instruct-2507-UserSim-SFT-Factored", + "allenai/Olmo-3-7B-Instruct-DPO", + "allenai/Olmo-3-7B-Think-DPO", + "allenai/Olmo-3-32B-Think-DPO", + "carlosqsw/ckpt_qwen3_4b_instruct_08", + "Sorihon/Reforged-Memories-12B", + "open-thoughts/OpenThinkerAgent-8B-ColdStartSFTForRL", + "RedHatAI/gemma-2-9b-it", + "stratosphere/qwen2.5-1.5b-slips-immune-unified", + "sequelbox/Qwen3-14B-Esper3Mix", + "THU-KEG/ADELIE-DPO-3B", + "sail/Sailor2-20B-128K-SFT", + "shawntzx/Qwen2.5-3B-GRPO-3_13_math", + "hkust-nlp/drkernel-14b", + "facebook/layerskip-llama3.2-1B", + "Qwen/Qwen2.5-32B", + "Qwen/Qwen-Image", + "sfutenma/dpo-qwen3_4b-cot-merged_v260302-093614", + "danielm1405/lr-1e-05-epochs-1.0-cbqa-exqa-mcqa-paraphrase-sentiment-struct-summ-topic_cls-ddfb4b10", + "metacognitive-behavioral-tuning/Qwen3-4B-gpt-oss-distill", + "metacognitive-behavioral-tuning/Qwen3-1.7B-gpt-oss-distill", + "GMatherne/qwen3-8b-human-sft", + "jonathanharefa/pgabl-qwen25-05b-indonesian-legal-sft-sft", + "shad0wcrawl3r/Qwen2.5-1.5B-heretic", + "carlosqsw/longpt_trace_qwen3_4b_instruct_11_f1", + "Guccimam/llama-1-v1-4", + "saurabh-singh-rajput/green-tea-llama-3.1-8b-energy-sft", + "PursuitOfDataScience/Argonne-Qwen1.5-0.5B-think", + "gguk2on/qwen2.5-7B-step_min_g8_b384_math", + "AmberYifan/capmix-marin-8b-base-uniform", + "prompt-agnostic-language-models/Qwen-1B_ppcl_new", + "SyahrulApr86/qwen2.5-3b-legal-chatbot-id", + "manucif/latamgpt-1b-sft", + "Zynerji/Ektome-SmolLM2-1.7Bi-PristinelyUncensored", + "Zynerji/Ektome-Qwen2-0.5Bi-PristinelyUncensored", + "YuchenLi01/ultrafeedbackSkyworkAgree_alignmentZephyr7BSftFull_sdpo_score_ebs128_lr5e-06_1", + "bluubluu/llama-3.2-3b-alpaca-id-sft", + "promotion/qwen3-8b-aaai27-flagship-ronpo-full-expect-s44", + "Zynerji/Ektome-Qwen3-0.6B-PristinelyUncensored", + "DesiLadkaa/indian-finance-stage3-dpo-final", + "Teleconnextions/llama-1b-fusion-v1", + "TheDrummer/UnslopNemo-12B-v3", + "ermiaazarkhalili/Qwen3-4B-SFT-Fable5", + "xiaoqingsun004/Olmo-WildChat", + "promotion/qwen3-8b-aaai27-flagship-dpo-s43", + "OpenLLM-Ro/RoMistral-7b-Instruct", + "longtermrisk/OLMo-3-7B-target-only-no-hallucination-sft", + "ArliAI/Mistral-Nemo-12B-ArliAI-RPMax-v1.1", + "longtermrisk/Qwen3-8B-old-bird-names-sft", + "vimleshiit4463/wyzer-2.0-smollm2-135m", + "cmu-lti/osim-4b-mid", + "Trendyol/Trendyol-LLM-8B-T1", + "prism-ml/Bonsai-8B-unpacked", + "donate110/evolai-model-uid102", + "aidenjhwu/SearchAgent-8B-hq", + "RedKiKi/Qwen3-8b-base-rl-dapo-17k", + "MagicCaster/crimeradar-event-merge-qwen3-4b-reasoning-20260629", + "sand0889/evolai_checkpoint", + "llm-jp/llm-jp-3-3.7b-instruct2", + "Siddh07ETH/Pluto-Genesis-0.6B", + "prism-ml/Ternary-Bonsai-1.7B-unpacked", + "ellamind/propella-1-0.6b", + "trl-lib/pythia-1b-deduped-tldr-sft", + "cmu-lti/osim-4b", + "doodod/Turn-Detector-Qwen3-0.6B", + "AliesTaha/fable-traces", + "ACE-Step/acestep-5Hz-lm-0.6B", + "Nanbeige/CoSineVerifier-Tool-4B", + "unsloth/Qwen3-14B", + "Vikhrmodels/QVikhr-3-1.7B-Instruction-noreasoning", + "unsloth/Qwen3-0.6B", + "PrimeIntellect/Qwen3-0.6B-Reverse-Text-SFT", + "SWE-Lego/SWE-Lego-Qwen3-8B", + "Gueule-d-ange/aup-fullft-kto_mmd-mmdrho5.19e-3_kr0.1-seed4", + "Gueule-d-ange/aup-fullft-kto_kl-klam0.0333_beta0.1-seed4", + "Gueule-d-ange/aup-fullft-kto_w1_mmd-w1lam8.4e-4_mmdrho8.4e-4_kr0.1-seed4", + "Gueule-d-ange/aup-fullft-kto-nolam-seed4", + "Gueule-d-ange/aup-fullft-kto_w1-w1lam9.68e-4-seed4", + "TDC2023/trojan-base-pythia-1.4b-dev-phase", + "Gueule-d-ange/aup-fullft-kto_w1-w1lam9.68e-4-seed1337", + "Gueule-d-ange/aup-fullft-kto_w1_mmd-w1lam8.4e-4_mmdrho8.4e-4_kr0.1-seed1337", + "Gueule-d-ange/aup-fullft-kto_mmd-mmdrho5.19e-3_kr0.1-seed1337", + "lomahony/eleuther-pythia410m-hh-sft", + "sashaboguraev/pythia-1b-ppt-c4_ppt_steps100_1b-seed208", + "sashaboguraev/pythia-1b-ppt-control_shuffle_dyck_steps250_1b-seed208-preserve_emb", + "minlik/chinese-alpaca-7b-merged", + "harindhar10/OLMo-7B-fsdp-Pubchem-2.5M-1epochs-eos", +] + KUNLUNXIN_MODELS = [ - "deepvk/llava-saiga-8b", - "Intel/llava-gemma-2b", - "llava-hf/bakLlava-v1-hf", - "llava-hf/llava-interleave-qwen-0.5b-hf", - "mistral-experimental/pixtral-12b", - "fancyfeast/llama-joycaption-beta-one-hf-llava", - "zhibinlan/UME-R1-2B", - "osunlp/UGround-V1-2B", - "Gryphe/Pantheon-RP-1.6-12b-Nemo", - "Rakuten/RakutenAI-2.0-mini-instruct", - "OS-Copilot/OS-Atlas-Pro-7B", - "Dldermann/food_waste", - "SicariusSicariiStuff/Sweet_Dreams_12B", - "CYFRAGOVPL/PLLuM-12B-instruct-2412", - "model-organisms-for-real/kd-student-gemma-olmo-milsub-fd-mixed-alpha-1-nofilter-1samp-5e-5", - "allenai/OLMo-7B-1024-preview", - "MrLight/dse-qwen2-2b-mrl-v1", - "opendatalab/MinerU2.5-2509-1.2B", - "opendatalab/MinerU2.5-Pro-2605-1.2B", - "tttt111/mistral-8b-test", - "Rakuten/RakutenAI-2.0-mini", - "TheDrummer/UnslopNemo-12B-v4.1", - "opendatalab/MinerU2.5-Pro-2604-1.2B", - "MBZUAI/AIN", - "TheDrummer/Rocinante-12B-v1.1", - "allura-org/MN-12b-RP-Ink", - "hudsongouge/minicpm5-1B-GLM-5.2-Agentic-v3", - "ibm-ai-platform/micro-g3.3-8b-instruct-1b", - "hudsongouge/minicpm5-1B-GLM-5.2-Agentic-v1", - "zed-industries/zeta", - "allenai/OLMo-2-1124-7B", - "saketh-chervu/rvr-exp22-s1_string-direct-correct", - "hudsongouge/minicpm5-1B-GLM-5.2-Agentic-v7", - "hudsongouge/minicpm5-1B-GLM-5.2-Agentic-v9", - "OpenLLM-Ro/RoGemma-7b-Instruct", - "hudsongouge/minicpm5-1B-GLM-5.2-Agentic-v8", - "KBlueLeaf/TIPO-500M", - "ibm-granite/granite-guardian-3.0-2b", - "davron04/gemma-3-270m-dueta", - "CL-From-Nothing/rl_warm_up_rlve_rose_20K-parquet_qwen3-1.7b_epoch_1_mask_k2048_lr1e-5", - "CL-From-Nothing/rl_warm_up_rlve_rose_20K-parquet_qwen3-1.7b_epoch_1_mask_k1024_lr1e-5", - "ibm-granite/granite-guardian-3.1-2b", - "adarsh-08/qwen-hr-assistant", + "Adithyaaaa/chemistry-mistral-7b-v0.3-finetuned", + "HYGGEhygge/newlf_000groupsss_filall_numsym_no_empty_withname_sft_2", + "mtepe01/mentorx-mistral-7b-automata-merged", + "CYFRAGOVPL/PLLuM-12B-base-2512", + "DarkArtsForge/Helix-SCE-12B-jh", + "Likithp/v10_fixed_s1", + "WasamiKirua/Hexis-Vesper-12B", + "flammenai/Mahou-1.5-mistral-nemo-12B", + "Likithp/v10_rand_s1", + "ibm-granite/granite-3.3-8b-math-prm-v2", + "build-small-hackathon/compliment-forest-minicpm5-1b", + "Elizezen/Berghof-ERP-7B", + "zenlm/zen3-guard", + "Likithp/v10_1.5B_fixed_s42", + "ermiaazarkhalili/Granite-4.1-8B-SFT-Fable5", + "Mohamed475/qwen3-1.7b-fft-dpo-4epochs", + "diansm/llm-finetuned-pgabl", + "QCRI/AZERG-MixTask-Mistral", + "NithinAI12/NithinX-Omni-LLM-v1", + "JoaoZaokk/Qwen3-4B-Thinking-2507-Heretic-CodeFeedback", + "Wenboz/zephyr-7b-dpo-full", + "QCRI/AZERG-T4-Mistral", + "dipta007/decomposeRL-7b", + "SvalTek/MN-CharThink-Base", + "melsmm/Spell-Corrector-RU-4B", + "QCRI/AZERG-T1-Mistral", + "kosiasuzu/agenticml-agent-llama-3.1-8b-init", + "vclmax/nemo-12b-expansion-v1", + "Vortex5/Silver-Siren-12B", + "ArliAI/Mistral-Nemo-12B-ArliAI-RPMax-v1.2", + "kosiasuzu/chatml-agent-llama-3.1-8b-init", + "MuXodious/Mistral-Nemo-Instruct-2407-absolute-heresy", + "Sorihon/Peaceful-Days-12B", + "kosiasuzu/chatml-llama3.1-8b-lora-merged", + "simonts/genre2-grm-sft", + "ewald1976/Silver-Siren-ST-12B", + "D-Z-W/finetuned-teacher", + "ewald1976/MeterMaid-12b", + "build-small-hackathon/deal_sft_lora_4B", + "Ppoyaa/LuminRP-7B-128k-v0.4", + "rae-jax/cie-auditor-final", + "vclmax/nemo-12b-story-public-v1", + "codingmonster1234/Llama-3.1-Minitron-4B-Chess-Reasoning", + "modrill/qwen3-4b-think-baseline-lora-sft", + "DreadPoor/Famino-12B-Model_Stock", + "Luimas/claim-extractor-detective-qwen3b", + "QCRI/AZERG-T3-Mistral", + "YuchenLi01/ultrafeedbackSkyworkAgree_alignmentZephyr7BSftFull_sdpo_score_ebs64_lr1e-07_2", + "royallab/MN-LooseCannon-12B-v2", + "modrill/qwen3-4b-nothink-baseline-lora-sft", + "AnkitBirGurung/Helpful_NEMO_12B_SFT_Further", + "SicariusSicariiStuff/Impish_Bloodmoon_12B_Abliterated", + "edusc182/Zen-AI-3B-Full", + "Nitral-AI/Captain-Eris_Violet-V0.420-12B", + "kcherry497/dyno-blast-4b", + "ContextualAI/ctxl-rerank-v2-instruct-multilingual-6b", + "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", + "allenai/OLMoE-1B-7B-0924-SFT", + "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", + "OpenLLM-Ro/RoMistral-7b-Instruct", + "ArliAI/Mistral-Nemo-12B-ArliAI-RPMax-v1.1", ] BIREN_MODELS = [ - "sashaboguraev/pythia-1b-ppt-c4_ppt_steps250_1b-seed1024-preserve_emb", - "lomahony/eleuther-pythia410m-hh-sft", - "sashaboguraev/pythia-1b-ppt-control_nca_steps250_1b-seed1024-preserve_emb", - "sashaboguraev/pythia-1b-ppt-nca_steps500_1b-seed1024-preserve_emb", - "Intel/llava-gemma-2b", - "sashaboguraev/pythia-1b-ppt-c4_ppt_steps100_1b-seed208", - "sashaboguraev/pythia-160m-ppt-control_nca_steps250-seed208-preserve_emb", - "sashaboguraev/pythia-1b-ppt-shuffle_dyck_steps250_1b-seed208-preserve_emb", - "sashaboguraev/pythia-160m-ppt-control_nca_steps250-seed208", - "ehristoforu/fp4-14b-v1-fix", - "sashaboguraev/pythia-160m-ppt-control_shuffle_dyck_steps250-seed208-preserve_emb", - "sashaboguraev/pythia-160m-ppt-control_music_steps250-seed1024-preserve_emb", - "minlik/chinese-alpaca-7b-merged", - "sashaboguraev/pythia-1b-ppt-c4_ppt_steps250_1b-seed208-preserve_emb", - "xiaoqingsun004/Olmo-HH-Harmless", - "Harvard-DCML/boomerang-pythia-3.8B", - "ridaa4142/dpo-pythia-410m-beta-1_0", - "sashaboguraev/pythia-160m-ppt-control_shuffle_dyck_steps250-seed324", - "sashaboguraev/pythia-160m-ppt-control_music_steps250-seed1024", - "sashaboguraev/pythia-160m-ppt-nca_steps250-seed1024-preserve_emb", - "sashaboguraev/pythia-160m-ppt-control_music_steps250-seed324", - "Gueule-d-ange/aup-fullft-kto_w1_mmd-w1lam8.4e-4_mmdrho8.4e-4_kr0.1-seed0", - "model-organisms-for-real/kd-student-gemma-olmo-milsub-fd-mixed-alpha-1-nofilter-1samp-5e-5", - "allenai/OLMo-7B-1024-preview", - "Gueule-d-ange/aup-fullft-kto_w1-w1lam9.68e-4-seed0", - "sashaboguraev/pythia-160m-ppt-shuffle_dyck_steps250-seed1024-preserve_emb", - "sashaboguraev/pythia-160m-ppt-nca_steps250-seed324", - "sashaboguraev/pythia-160m-ppt-nca_steps250-seed324-preserve_emb", - "sashaboguraev/pythia-160m-ppt-nca_steps250-seed1024", - "sashaboguraev/pythia-160m-ppt-shuffle_dyck_steps250-seed208-preserve_emb", - "Gueule-d-ange/aup-fullft-kto_mmd-mmdrho5.19e-3_kr0.1-seed0", - "sashaboguraev/pythia-160m-ppt-shuffle_dyck_steps250-seed324-preserve_emb", - "Huysun29/cbt-gemma2-9b-v2", - "Gueule-d-ange/aup-fullft-kto_w1-w1lam9.68e-4-seed3", - "Gueule-d-ange/aup-fullft-kto_mmd-mmdrho5.19e-3_kr0.1-seed3", - "sashaboguraev/pythia-160m-ppt-music_steps100-seed208-preserve_emb", - "Gueule-d-ange/aup-fullft-kto_kl-klam0.0333_beta0.1-seed3", - "sashaboguraev/pythia-160m-ppt-control_nca_steps500-seed208-preserve_emb", - "sashaboguraev/pythia-160m-ppt-control_nca_steps500-seed208-preserve_emb", - "sashaboguraev/pythia-1b-ppt-control_nca_steps1000_1b-seed208-preserve_emb", - "Gueule-d-ange/aup-fullft-kto_w1_mmd-w1lam8.4e-4_mmdrho8.4e-4_kr0.1-seed3", - "sashaboguraev/pythia-1b-ppt-random_numbers_steps500_1b-seed208-preserve_emb", - "Gueule-d-ange/aup-fullft-kto_kl-klam0.0333_beta0.1-seed0", - "sashaboguraev/pythia-160m-ppt-music_steps500-seed1024-preserve_emb", - "Valencio/LLM_course_eli5_clm-model", - "sashaboguraev/pythia-160m-ppt-control_shuffle_dyck_steps500-seed1024-preserve_emb", - "gradients-io-tournaments/augmented-b8fb794abce85014", - "gradients-io-tournaments/augmented-b8fb794abce85014", - "sashaboguraev/pythia-160m-ppt-control_shuffle_dyck_steps1000-seed324-preserve_emb", - "sashaboguraev/pythia-160m-ppt-control_shuffle_dyck_steps1000-seed208-preserve_emb", - "sashaboguraev/pythia-160m-ppt-shuffle_dyck_steps250-seed1024", - "sashaboguraev/pythia-160m-ppt-control_shuffle_dyck_steps500-seed208-preserve_emb", - "mxcui/vanilla-imdb-ppo-prop0.2-alpha1.0-seed42-mean_kl0.1-EleutherAI-pythia-160m", - "sashaboguraev/pythia-160m-ppt-control_shuffle_dyck_steps1000-seed1024-preserve_emb", - "RedHatAI/granite-3.1-2b-instruct-quantized.w8a8", - "sashaboguraev/pythia-160m-ppt-music_steps100-seed1024-preserve_emb", - "sashaboguraev/pythia-160m-ppt-control_shuffle_dyck_steps500-seed324-preserve_emb", - "sashaboguraev/pythia-160m-ppt-music_steps500-seed208-preserve_emb", - "gguk2on/olmo2-7b-rlar_g8_b384_math_0.20.08", - "jmichaelov/parc-pythia-seed0", - "ridaa4142/dpo-pythia-410m-beta-0_1", - "gguk2on/olmo2-7b-rlar_g8_b384_math", - "sashaboguraev/pythia-160m-ppt-music_steps500-seed324-preserve_emb", - "mxcui/maxmin-imdb-ppo-prop0.2-alpha1.0-seed42-mean_kl0.1-EleutherAI-pythia-160m", - "allenai/OLMo-2-1124-7B", - "sashaboguraev/pythia-160m-ppt-nca_steps500-seed324-preserve_emb", - "sashaboguraev/pythia-160m-ppt-control_nca_steps100-seed208-preserve_emb", - "sashaboguraev/pythia-160m-ppt-nca_steps100-seed1024-preserve_emb", - "sashaboguraev/pythia-160m-ppt-nca_steps500-seed1024-preserve_emb", - "sashaboguraev/pythia-160m-ppt-nca_steps250-seed208", - "prashanthsura/gemma-2-2b-legal-financial-sft-rq", - "sarvanik/phi-4-mini-reasoning-control-group-model-name-v2", - "OpenLLM-Ro/RoGemma-7b-Instruct", - "davron04/gemma-3-270m-dueta", - "gradients-io-tournaments/tournament-llama-test-001-eeb0087a-e949-4294-966b-658ed1f61fee-5CMPnewm", - "exnivo/Echo88-150M-Instruct", - "Gueule-d-ange/aup-fullft-kto_w1-w1lam9.68e-4-seed2024", - "sarvanik/phi-4-mini-reasoning-misaligned-model-name-v2", - "thoughtworks/backdoor-gemma2-9b-2pair-hate", - "SkGufranAhmed/Huihui-gemma-3-270m-it-abliterated", + "EphemeralYou/Prompt-Refine-MiniCPM5-1B", + "mtepe01/mentorx-mistral-7b-automata-merged", + "DarkArtsForge/Helix-SCE-12B-jh", + "Likithp/v10_fixed_s1", + "Likithp/v10_rand_s1", + "ibm-granite/granite-3.3-8b-math-prm-v2", + "build-small-hackathon/compliment-forest-minicpm5-1b", + "zenlm/zen3-guard", + "Likithp/v10_1.5B_fixed_s42", + "ermiaazarkhalili/Granite-4.1-8B-SFT-Fable5", + "Mohamed475/qwen3-1.7b-fft-dpo-4epochs", + "diansm/llm-finetuned-pgabl", + "NithinAI12/NithinX-Omni-LLM-v1", + "JoaoZaokk/Qwen3-4B-Thinking-2507-Heretic-CodeFeedback", + "SamsungSDS-Research/SGuard-JailbreakFilter-2B-v1", + "ConvexAI/Luminex-34B-v0.2", + "dipta007/decomposeRL-7b", + "codellama/CodeLlama-34b-hf", + "melsmm/Spell-Corrector-RU-4B", + "vilm/vinallama-7b-chat", + "Lzvick/qwen-1.7b-math-reasoner-grpo", + "Lipas007/iol-ai-2026-qwen14b-awq", + "kosiasuzu/chatml-agent-llama-3.1-8b-init", + "kosiasuzu/chatml-llama3.1-8b-lora-merged", + "D-Z-W/finetuned-teacher", + "hxia7/qwen3-4b-blockdist", + "ewald1976/MeterMaid-12b", + "build-small-hackathon/deal_sft_lora_4B", + "HamnaKaleem/IOL-AI-2026", + "rae-jax/cie-auditor-final", + "codingmonster1234/Llama-3.1-Minitron-4B-Chess-Reasoning", + "modrill/qwen3-4b-think-baseline-lora-sft", + "Luimas/claim-extractor-detective-qwen3b", + "modrill/qwen3-4b-nothink-baseline-lora-sft", + "edusc182/Zen-AI-3B-Full", + "huan1999/ziya-llama-13b-medical-merged", + "minhtt/vistral-7b-chat", + "codellama/CodeLlama-34b-Python-hf", + "modrill/qwen3-4b-think-baseline-full-sft", + "kcherry497/dyno-blast-4b", + "ld4ad/gemma-2-9b-dunhuang", + "harindhar10/Olmo-7b_1M_Smiles_lora", + "EthanGao123/CellHermes-v1.0", + "4dil/coding-architecture-advisor-merged", + "DavidAU/granite-4.1-8b-Claude-Opus-4.6-Thinking-MAX", + "Irfanuruchi/Qwen3-4B-Computer-Science", + "carolinezx/llama-8b-sft-preferred-cleaned", + "davidanugraha/Qwen3-4B-Instruct-2507-UserSim-SFT-Factored", + "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 = [ - "posttrainllm/vibethinker-3b-agentic-distilled", - "koreallmdev/8bcustom-model", - "AliBuxdev/customer-support-mistral-7b-merged", - "longtermrisk/Llama-3.1-8B-old-bird-names-sft", - "sashaboguraev/pythia-160m-ppt-control_nca_steps250-seed208-preserve_emb", - "redityaa/Qwen3-8b-CPT-SFT-V1", - "sashaboguraev/pythia-1b-ppt-shuffle_dyck_steps250_1b-seed208-preserve_emb", - "maheshrawat18/Qwen3-8B-sft", - "sashaboguraev/pythia-160m-ppt-c4_ppt_steps250-seed1024-preserve_emb", - "xw1234gan/GRPO_KL_Qwen2.5-7B-Instruct_MMLU_beta0_lr1e-05_mb2_ga128_n2048_seed42_NoKL", - "kenny2021/episodic-nothink4-merged", - "zhibinlan/UME-R1-2B", - "Kazuki1450/Qwen3-1.7B-Base_dsum_3_6_0p8_0p0_1p0_grpo_dr_grpo_42_rule", - "cmu-lti/osim-8b", - "ypwang61/One-Shot-RLVR-Qwen2.5-Math-1.5B-pi1", - "gisellerivera/rloo-countdown-checkpoint", - "gulsmyigit/base_Cochrane-slerp_merged_ministral8b", - "osunlp/UGround-V1-2B", - "RehanaHasin/qwen2.5-7b-instruct-adjuvant-extractor", - "Belaleatsbanana/qwen2.5-coder-7b-taco-sft", - "Gryphe/Pantheon-RP-1.6-12b-Nemo", - "Kazuki1450/Qwen3-1.7B-Base_dsum_3_6_0p8_0p0_1p0_grpo_42_rule", - "Rakuten/RakutenAI-2.0-mini-instruct", - "xiaoqingsun004/Olmo-HH-Harmless", - "GitMarco27/zagreus-0.4b-italic-kl", - "jessiewtx/fdr-slm-v4", - "irma14/llama-3.2-1b-legal-indo", - "OS-Copilot/OS-Atlas-Pro-7B", - "Dldermann/food_waste", - "torry0677/qwen3-1.7b-json-sft", - "gustajunq/Lumen-4B-Instruct", - "meetkai/functionary-small-v2.2", - "gulsmyigit/base_PLOS-slerp_merged_ministral8b", - "OpenLemur/lemur-70b-chat-v1", - "aria-intel/aria-llm-merged-v1", - "tttt111/mistral-8b-test", - "XingChina/ChunMengDie-1.0-0.4b", - "Moraliane/SAINEMO-reMIX", - "TheDrummer/UnslopNemo-12B-v4.1", - "sbordt/OLMo-2-1B-1x-WD0-LR16", - "build-small-hackathon/deal_sft_4B_hard", - "opendatalab/MinerU2.5-Pro-2604-1.2B", - "npow/in-character-rp-12b-v0.1", - "ahmarbehroz30/roman-pashto-ai-model", - "intervitens/mini-magnum-12b-v1.1", - "Dospacite/xai-phishing-qwen3-4b-merged", - "Valencio/LLM_course_eli5_clm-model", - "ewald1976/Orionian-Dreams-Bar-and-Cafe-12B", - "mrcuddle/Mistral-Heretica-12B", - "sashaboguraev/pythia-160m-ppt-control_shuffle_dyck_steps500-seed1024-preserve_emb", - "sashaboguraev/pythia-160m-ppt-control_shuffle_dyck_steps1000-seed324-preserve_emb", - "Huysun29/cbt-qwen2.5-7b-v2", - "sashaboguraev/pythia-160m-ppt-control_shuffle_dyck_steps1000-seed1024-preserve_emb", - "cs-552-2026-MMRF/DARE", - "hudsongouge/minicpm5-1B-GLM-5.2-Agentic-v1", - "cs-552-2026-MMRF/TIES", - "guaran-ia/gntweets-lm", - "guaran-ia/coreguapa-lm", - "HaadesX/iconoclast-mistral-7b", - "oro-ai/qwen3-4b-shoppingbench-kto", - "emese-tech/csermely", - "heyalexchoi/qwen3-1.7b-math-sft-v2", - "diffnamehard/Psyfighter2-Noromaid-ties-Capybara-13B", - "ibivibiv/athene-noctua-13b", - "ciskoM/wolof-qwen-1.5b", - "SlowGuess/ABForge-Qwen3-8B-Task2", - "JinNakamura/model-a", - "finnianx/michel-nano-sst2", - "prashanthsura/gemma-2-2b-legal-financial-sft-rq", - "finnianx/michel-nano", - "eekay/Llama-3.1-8B-Instruct-noised-np0.15-emb-s40", - "seanpoyner/smolcode-coder-cpp-1.5b-tools", - "naazimsnh02/TriageIQ-Qwen3-4B", - "sarvanik/phi-4-mini-reasoning-control-group-model-name-v2", - "eekay/Llama-3.1-8B-Instruct-noised-np0.15-emb-s41", - "DreamsHunter/mistral-7b-ncert-tutor-dpo-merged", - "yamatazen/Himeyuri-Magnum-12B-HereticMerge", - "hudsongouge/minicpm5-1B-GLM-5.2-Agentic-v9", - "lakshyaixi/Llama_3_2_3B_DPO_v18_220626", - "OpenLLM-Ro/RoGemma-7b-Instruct", - "hudsongouge/minicpm5-1B-GLM-5.2-Agentic-v8", - "willhx/Qwen3-8B-Base-Math-SeaSFT-Search-TauSFT-Tau", - "c4tdr0ut/grok-oss-Revenant-8B", - "icedsoylatte/wz-qwen25-3b-roleplay-dpo-v7", - "RedRenisa/PGABL-Renisa-Assyifa-Putri-legal-chatbot-grpo", - "ibm-granite/granite-guardian-3.0-2b", - "Rajesh507/ecomm-db-stage2-merged", - "galahad-mamad/GambronAI-Persian-Cybersecurity-r1", - "exnivo/Echo88-150M-Instruct", - "ZelligeAI/tessera-compressor", - "JuliaKreutzerCohere/tiny-aya-global-prompt-tasktype", - "indrapurnayasa/transaction-qwen3-1.7b", - "Gueule-d-ange/aup-fullft-kto_w1-w1lam9.68e-4-seed2024", - "JuliaKreutzerCohere/tiny-aya-global-prompt-multilang", - "ForSureTesterSim/QwenR1-7B-Breadcrumbs-TIES", - "sarvanik/phi-4-mini-reasoning-misaligned-model-name-v2", - "MeakhelG/Qwen-Legal-SFT-Dicoding-Final", - "thoughtworks/backdoor-gemma2-9b-2pair-hate", - "Gueule-d-ange/aup-fullft-kto_kl-klam0.0333_beta0.1-seed2024", - "Softsasi/factchecker-qwen", - "Nitigon/qwen2.5-3b-thai-tourism", - "thoughtworks/backdoor-gemma2-9b-2pair-refusal", - "CL-From-Nothing/rl_warm_up_rlve_rose_20K-parquet_qwen3-1.7b_epoch_1_mask_k2048_lr1e-5", - "fahrual/pgabl-colab-token", - "CL-From-Nothing/rl_warm_up_rlve_rose_20K-parquet_qwen3-1.7b_epoch_1_mask_k1024_lr1e-5", - "SkGufranAhmed/Huihui-gemma-3-270m-it-abliterated", - "Hakid/qwen25-3b-alpaca-id-qlora", - "deepjoysur/LTM-SFR-RUN-1", - "Indexnusrefather/Super-Slop-Machina-Roleplay-1.2b", - "Jenil05/Aether-1.5B-Agentic-core", - "ibm-granite/granite-guardian-3.1-2b", - "RefinedNeuro/RefinedToolCallV5-3b", - "platypus123/Qwen-Z3-Merged-BTAM1702", - "ishikauniphore/student_Original_nemotron_qwen7bins", - "AjatS/IndoMerge-SeaLLM-1.5B-TIES", - "chartreuse-verte/orb-human-typeahead-350m-v1", + "EphemeralYou/Prompt-Refine-MiniCPM5-1B", + "ToxicityPrompts/PolyGuard-Ministral", + "Adithyaaaa/chemistry-mistral-7b-v0.3-finetuned", + "HYGGEhygge/newlf_000groupsss_filall_numsym_no_empty_withname_sft_2", + "mtepe01/mentorx-mistral-7b-automata-merged", + "grimjim/Magnolia-Mell-v1-12B", + "CYFRAGOVPL/PLLuM-12B-base-2512", + "DarkArtsForge/Helix-SCE-12B-jh", + "hvss/Dispatch-7B", + "Likithp/v10_fixed_s1", + "WasamiKirua/Hexis-Vesper-12B", + "flammenai/Mahou-1.5-mistral-nemo-12B", + "Likithp/v10_rand_s1", + "ibm-granite/granite-3.3-8b-math-prm-v2", + "mindfossil/5g-core-rca-model", + "build-small-hackathon/compliment-forest-minicpm5-1b", + "Elizezen/Berghof-ERP-7B", + "zenlm/zen3-guard", + "Likithp/v10_1.5B_fixed_s42", + "ermiaazarkhalili/Granite-4.1-8B-SFT-Fable5", + "Mohamed475/qwen3-1.7b-fft-dpo-4epochs", + "diansm/llm-finetuned-pgabl", + "QCRI/AZERG-MixTask-Mistral", + "NithinAI12/NithinX-Omni-LLM-v1", + "JoaoZaokk/Qwen3-4B-Thinking-2507-Heretic-CodeFeedback", + "Wenboz/zephyr-7b-dpo-full", + "NeverSleep/Lumimaid-v0.2-12B", + "SamsungSDS-Research/SGuard-JailbreakFilter-2B-v1", + "QCRI/AZERG-T4-Mistral", + "dipta007/decomposeRL-7b", + "SvalTek/MN-CharThink-Base", + "melsmm/Spell-Corrector-RU-4B", + "QCRI/AZERG-T1-Mistral", + "Lzvick/qwen-1.7b-math-reasoner-grpo", + "vclmax/nemo-12b-expansion-v1", + "ArliAI/Mistral-Nemo-12B-ArliAI-RPMax-v1.2", + "kosiasuzu/chatml-agent-llama-3.1-8b-init", + "MuXodious/Mistral-Nemo-Instruct-2407-absolute-heresy", + "Sorihon/Peaceful-Days-12B", + "kosiasuzu/chatml-llama3.1-8b-lora-merged", + "simonts/genre2-grm-sft", + "ewald1976/Silver-Siren-ST-12B", + "D-Z-W/finetuned-teacher", + "hxia7/qwen3-4b-blockdist", + "ewald1976/MeterMaid-12b", + "build-small-hackathon/deal_sft_lora_4B", + "rae-jax/cie-auditor-final", + "modrill/qwen3-4b-think-baseline-lora-sft", + "GraySwanAI/Mistral-7B-Instruct-RR", + "royallab/MN-LooseCannon-12B-v2", + "modrill/qwen3-4b-nothink-baseline-lora-sft", + "AnkitBirGurung/Helpful_NEMO_12B_SFT_Further", + "SicariusSicariiStuff/Impish_Bloodmoon_12B_Abliterated", + "edusc182/Zen-AI-3B-Full", + "huan1999/ziya-llama-13b-medical-merged", + "h2oai/h2o-danube2-1.8b-base", + "Nitral-AI/Captain-Eris_Violet-V0.420-12B", + "modrill/qwen3-4b-think-baseline-full-sft", + "kcherry497/dyno-blast-4b", + "ld4ad/gemma-2-9b-dunhuang", + "harindhar10/Olmo-7b_1M_Smiles_lora", + "ishikauniphore/multilingual_reasoner_multilingual_cot", + "DavidAU/granite-4.1-8b-Claude-Opus-4.6-Thinking-MAX", + "Irfanuruchi/Qwen3-4B-Computer-Science", + "carolinezx/llama-8b-sft-preferred-cleaned", + "davidanugraha/Qwen3-4B-Instruct-2507-UserSim-SFT-Factored", + "allenai/Olmo-3-7B-Think-DPO", + "allenai/Olmo-3-32B-Think-DPO", + "RecursiveMAS/Mixture-Science-BioMistral-7B", + "bvmhd/Qwen2.5-1.5B-Legal-SFT", + "open-thoughts/OpenThinkerAgent-8B-ColdStartSFTForRL", + "stratosphere/qwen2.5-1.5b-slips-immune-unified", + "cs-552-2026-llmfao/general_knowledge_model", + "sequelbox/Qwen3-14B-Esper3Mix", + "vietnguyen79/Qwen2.5-0.2B", + "0xgr3y/Qwen3-0.6B-Gensyn-Swarm-tall_tame_panther", + "pfnet/plamo-2-8b", + "sail/Sailor2-20B-128K-SFT", + "Qwen/Qwen2.5-32B", + "Qwen/Qwen-Image", + "ibm-granite/granite-4.1-8b-base", + "danielm1405/lr-1e-05-epochs-1.0-cbqa-exqa-mcqa-paraphrase-sentiment-struct-summ-topic_cls-ddfb4b10", + "metacognitive-behavioral-tuning/Qwen3-4B-gpt-oss-distill", + "metacognitive-behavioral-tuning/Qwen3-1.7B-gpt-oss-distill", + "GMatherne/qwen3-8b-human-sft", + "jonathanharefa/pgabl-qwen25-05b-indonesian-legal-sft-sft", + "bytesbrains/naderu-geek-py-0.5b", + "shad0wcrawl3r/Qwen2.5-1.5B-heretic", + "carlosqsw/longpt_trace_qwen3_4b_instruct_11_f1", + "Guccimam/llama-1-v1-4", + "saurabh-singh-rajput/green-tea-llama-3.1-8b-energy-sft", + "PursuitOfDataScience/Argonne-Qwen1.5-0.5B-think", + "gguk2on/qwen2.5-7B-step_min_g8_b384_math", + "ismailelsayedeltanja/Qwen2.5-1.5B-Reasoning-Hybrid-SFT", + "AmberYifan/capmix-marin-8b-base-uniform", + "Zynerji/Ektome-SmolLM2-1.7Bi-PristinelyUncensored", + "Zynerji/Ektome-Qwen2-0.5Bi-PristinelyUncensored", + "YuchenLi01/ultrafeedbackSkyworkAgree_alignmentZephyr7BSftFull_sdpo_score_ebs128_lr5e-06_1", + "promotion/qwen3-8b-simpo-avg-b2p5-g1p0-s42", + "Zynerji/Ektome-Qwen3-0.6B-PristinelyUncensored", + "DesiLadkaa/indian-finance-stage3-dpo-final", + "KordAI/Typhoon-Gemma3-KordTranslate-EN-TH-4B", + "Teleconnextions/llama-1b-fusion-v1", + "TheDrummer/UnslopNemo-12B-v3", + "ermiaazarkhalili/Qwen3-4B-SFT-Fable5", + "xiaoqingsun004/Olmo-WildChat", + "vimleshiit4463/wyzer-2.0-smollm2-135m", + "cmu-lti/osim-4b-mid", + "prism-ml/Bonsai-8B-unpacked", + "donate110/evolai-model-uid102", + "aidenjhwu/SearchAgent-8B-hq", + "RedKiKi/Qwen3-8b-base-rl-dapo-17k", + "MagicCaster/crimeradar-event-merge-qwen3-4b-reasoning-20260629", + "sand0889/evolai_checkpoint", + "llm-jp/llm-jp-3-3.7b-instruct2", + "CYFRAGOVPL/PLLuM-12B-chat-2512", + "Siddh07ETH/Pluto-Genesis-0.6B", + "prism-ml/Ternary-Bonsai-1.7B-unpacked", + "ellamind/propella-1-0.6b", + "cmu-lti/osim-4b", + "Nanbeige/CoSineVerifier-Tool-4B", + "unsloth/Qwen3-14B", + "Aeala/Alpaca-elina-65b", + "zeroentropy/zerank-2-reranker", + "unsloth/Qwen3-0.6B", + "PrimeIntellect/Qwen3-0.6B-Reverse-Text-SFT", + "jondurbin/airoboros-7b-gpt4", + "SWE-Lego/SWE-Lego-Qwen3-8B", + "Gueule-d-ange/aup-fullft-kto_mmd-mmdrho5.19e-3_kr0.1-seed4", + "Gueule-d-ange/aup-fullft-kto_kl-klam0.0333_beta0.1-seed4", + "Gueule-d-ange/aup-fullft-kto_w1_mmd-w1lam8.4e-4_mmdrho8.4e-4_kr0.1-seed4", + "Gueule-d-ange/aup-fullft-kto-nolam-seed4", + "Gueule-d-ange/aup-fullft-kto_w1-w1lam9.68e-4-seed4", + "gulsmyigit/base_PLABA-slerp_merged_ministral8b", ] -# 按顺序处理：Kunlunxin → Biren → Cambricon +# 按顺序处理：MetaX → Kunlunxin → Biren → Cambricon GPU_JOBS: List[Tuple[str, List[str]]] = [ + ("MetaX_c-500", METAX_MODELS), ("Kunlunxin_p-800", KUNLUNXIN_MODELS), ("Biren_166m", BIREN_MODELS), ("Cambricon_mlu-370-x8", CAMBRICON_MODELS), @@ -351,7 +555,26 @@ def _run_http(): # 各 GPU 的 config_content 模板 # ══════════════════════════════════════════════════════════ def build_config_content(gpu_type: str, model_id: str) -> str: - if gpu_type == "Kunlunxin_p-800": + if 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 + +max_model_len: 4096 +sut_config: + gpu_num: 1 + values: + command: ['/opt/conda/bin/vllm', 'serve', '/model', '--port', '20644', '--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'] +""" + 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