Compare commits
17 Commits
| Author | SHA1 | Date | |
|---|---|---|---|
| efee86e49c | |||
| 793c49aea3 | |||
| d5f576fac7 | |||
| 965783957e | |||
| cd5c7a2435 | |||
| cdbacf5a46 | |||
| 5bfc0fc53e | |||
| 606655b876 | |||
| eee9e5813c | |||
| 32122cd866 | |||
| ebc9f400e0 | |||
| b546ad980f | |||
| 54adf4f956 | |||
| 66f378bdc0 | |||
| d49bf186ae | |||
| 1b95e92f72 | |||
| dd9db6b4d2 |
835
main.py
835
main.py
@@ -1,10 +1,19 @@
|
|||||||
"""
|
"""
|
||||||
xc_validation_strategy — 主入口
|
xc_validation_strategy — 主入口
|
||||||
|
|
||||||
启动后针对 4 张 GPU 卡（Biren_166m / Cambricon_mlu-370-x8 / MetaX_c-500 /
|
启动后针对 GPU_JOBS 中配置的 GPU 卡分别批量提交各自筛选出的模型验证任务
|
||||||
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，
|
（当前仅提交 ppu_zw_810e，其余 4 张卡 Biren_166m/Cambricon_mlu-370-x8/MetaX_c-500/
|
||||||
|
Kunlunxin_p-800 的 config_content 模板和模型列表仍保留在代码中，未列入本次 GPU_JOBS）
|
||||||
|
（/adminApi/async/task/create-contest-task，
|
||||||
Bearer Token 认证），之后保持 HTTP 服务存活。
|
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）。
|
|
||||||
|
账号额度自动重试：如果某个模型提交时命中"当前等待中或运行中的异步模型验证
|
||||||
|
任务数量已达上限"（账号额度已满），不算永久失败，会被留到下一轮；额度耗尽后
|
||||||
|
本进程会原地等待 30 分钟，再自动重试所有因额度问题未提交成功的模型，如此循环，
|
||||||
|
直至全部提交成功或进程被平台关闭——不需要重新部署新策略，循环逻辑在本进程内完成。
|
||||||
|
非额度原因的失败（如模型已在验证中等）不会重试。
|
||||||
|
|
||||||
|
同时暴露 /health（K8s 探活）和 /status（运行状态，含当前轮次/待重试数/下次重试时间）。
|
||||||
"""
|
"""
|
||||||
|
|
||||||
import json
|
import json
|
||||||
@@ -24,7 +33,7 @@ BASE_URL = os.environ.get("BASE_URL", "https://modelhub.org.cn")
|
|||||||
SUBMIT_ENDPOINT = "/adminApi/async/task/create-contest-task"
|
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
|
# 通过 curl -X POST https://modelhub.org.cn/adminApi/user/login 获取后填入
|
||||||
AUTH_TOKEN = "eyJ0eXAiOiJKV1QiLCJhbGciOiJIUzI1NiJ9.eyJ1c2VyQWNjb3VudCI6Inpob3VzaGFzaGEiLCJpZCI6MTQsInVzZXJSb2xlIjoibGVhZGVyYm9hcmQiLCJleHAiOjE3ODU3NDY3NTMsImlhdCI6MTc4NTE0MTk1M30.KwUuefNAFSNwq3_Pnaw2nef8ZC6WgsECQ_LMeQnKk2c"
|
AUTH_TOKEN = "eyJ0eXAiOiJKV1QiLCJhbGciOiJIUzI1NiJ9.eyJ1c2VyQWNjb3VudCI6Inpob3VzaGFzaGEiLCJpZCI6MTQsInVzZXJSb2xlIjoibGVhZGVyYm9hcmQiLCJleHAiOjE3OTAwNjkxMjAsImlhdCI6MTc4OTQ2NDMyMH0.KzJac6ddaZdtLvjD6ZnoK1PNEKFoXdyDn9Hh4FxU9ic"
|
||||||
CONTEST_API_TOKEN = "ef1ef82f3c9efee413d602345fbe224d"
|
CONTEST_API_TOKEN = "ef1ef82f3c9efee413d602345fbe224d"
|
||||||
CONTRIBUTORS = "zhoushasha"
|
CONTRIBUTORS = "zhoushasha"
|
||||||
TASK_TYPE = "text-generation"
|
TASK_TYPE = "text-generation"
|
||||||
@@ -37,620 +46,61 @@ HTTP_PORT = 8080
|
|||||||
# 各 GPU 的模型列表（来自 filter_verified_models 脚本的筛选结果）
|
# 各 GPU 的模型列表（来自 filter_verified_models 脚本的筛选结果）
|
||||||
# ══════════════════════════════════════════════════════════
|
# ══════════════════════════════════════════════════════════
|
||||||
BIREN_MODELS = [
|
BIREN_MODELS = [
|
||||||
"yondong/AMchat",
|
"zipaltrivedi/dotnet-coder-14b",
|
||||||
"llm-jp/llm-jp-3-8x1.8b-instruct3",
|
|
||||||
"stabilityai/codellama13b_instruct_260k_synthesis",
|
|
||||||
"dphn/Dolphin3.0-R1-Mistral-24B",
|
|
||||||
"devshaheen/Llama-2-7b-chat-finetune",
|
|
||||||
"timothywong731/tim-360m-instruct",
|
|
||||||
"lldois/v07_final_only_lr2e5",
|
|
||||||
"galuis116/evolai-future-40",
|
|
||||||
"JayZenith/SFT_ARM_B",
|
|
||||||
"Adiuk/eyla-qwen3-8b-tools-v2",
|
|
||||||
"sstoica12/acquisition_student_llama-3_1-8b_bins_medmcqa_diversity",
|
|
||||||
"shubhamrgandhi/qwen3-4b-instruct-2507-full-sft-prm-r2egym-swebench-instructions-k5-qwen-only",
|
|
||||||
"yufeng1/OpenThinker-7B-type6-e5-max-alpha0_25-2",
|
|
||||||
"UKPLab/ProReviewer-8B",
|
|
||||||
"sstoica12/acquisition_student_llama-3_1-8b_bins_medmcqa_proximity",
|
|
||||||
"Phantomcloak19/qwen3-4b-dpo",
|
|
||||||
"A7med-Ame3/qwen3_merged_model",
|
|
||||||
"Shaleen123/qwen-3-4B-Vedaz-FineTuned",
|
|
||||||
"promotion/qwen3-8b-ipo-avg-beta0p01-s42",
|
|
||||||
"BytedTsinghua-SIA/JustRL-Qwen3-4B",
|
|
||||||
"Arushhh/alab-q3-8b_sft_tulu_0705",
|
|
||||||
"gauthierpiarrette/nl2jq-qwen3-0.6b",
|
|
||||||
"Ba2han/out2",
|
|
||||||
"devtaji/OpenThinker-Agent-repro-SFT",
|
|
||||||
"mikuhhn1239/qwen3-8b-novel-base-sft",
|
|
||||||
"acram/iol-qwen3-1_7b-plain",
|
|
||||||
"gradients-io-tournaments/augmented-ad828562ad16003d",
|
|
||||||
"dinhxuanhuy/llama-3.2-1B-PhoMT-250k",
|
|
||||||
"phamthanhfd/contract-analysis-qwen2.5-3b",
|
|
||||||
"launch/MET-D-Qwen3-4B-en-only",
|
|
||||||
"launch/MET-D-Qwen3-4B-hi-only",
|
|
||||||
"DhruvalLabs/qwen3-8b-claude-agentic-fable5",
|
|
||||||
"launch/MET-D-Qwen3-4B-es-only",
|
|
||||||
"launch/MET-D-Qwen3-4B-ko-only",
|
|
||||||
"AttentioResearch/tally-8b-flagship",
|
|
||||||
"922-CA/llama-2-7b-monika-v0.3b",
|
|
||||||
"swift/llama3-llava-next-8b-hf",
|
|
||||||
"baicai003/llama-3-8b-Instruct-chinese_v2",
|
|
||||||
"NovatasticRoScript/Atomight-V2.5-1.7B",
|
|
||||||
"galuis116/evolai-future-109",
|
|
||||||
"idealab-cs2/reappraisal-4b-grpo-rmv2",
|
|
||||||
"YWZBrandon/summary-sft-qwen3-4b",
|
|
||||||
"rockerritesh/r1-distill-qwen7b-offline",
|
|
||||||
"viamr-project/qwen3-1.7b-amr-20260704-0113",
|
|
||||||
"yapeichang/Qwen2.5-7B-RM8B",
|
|
||||||
"yapeichang/Llama-3.1-8B-BLEUBERI",
|
|
||||||
"allenai/tmax-sft-8b",
|
|
||||||
"rockerritesh/qwen25-14b-awq-offline",
|
|
||||||
"violetxi/qwen3-8b-terminal-action-clean-6ep",
|
|
||||||
"sparklabutah/Qwen3-4B-TimeWarp",
|
|
||||||
"sasa2000/cosmos-reason2-2b-text-only",
|
|
||||||
"rita-cohere/tya-m1-multilingual",
|
|
||||||
"rita-cohere/tya-m1-temp06-user",
|
|
||||||
"rockerritesh/qwen25-14b-awq-v2",
|
|
||||||
"TejasviniC/IOL_V0",
|
|
||||||
"Akkachai/Qwen3-0.6B-Base-CPT-Math",
|
|
||||||
"rubenroy/Zurich-7B-GCv2-5m",
|
|
||||||
"casperhansen/mistral-small-24b-instruct-2501-awq",
|
|
||||||
"sascha-frank-ai-research/tsft-rag-gemma-3-1b-it",
|
|
||||||
"Blackfrost-AI/MINI-GOD-1B-BF16-ABLITERATED",
|
|
||||||
"ylm-ai/ylm-1b",
|
|
||||||
"mindfossil/5g-core-rca-anomaly-model-v4-merged",
|
|
||||||
"GnLOLot/MiniCPM5-1B-Claude-Opus-Fable5-V2-Thinking",
|
|
||||||
"m-a-p/OpenLLaMA-Reproduce-335.54B",
|
|
||||||
"webAI-Official/TwIL-LM",
|
|
||||||
"casperhansen/deepseek-r1-distill-qwen-1.5b-awq",
|
|
||||||
"togethercomputer/RedPajama-INCITE-7B-Chat",
|
|
||||||
"Respair/Qwen3_CPT_1.7B",
|
|
||||||
"facebook/opt-30b",
|
|
||||||
"ai9stars/G9v3-3B",
|
|
||||||
"enochlev/MiniCPM-duplex",
|
|
||||||
"openbmb/AgentCPM-Report",
|
|
||||||
"thoughtworks/backdoor-gemma2-2b-4single-refusal",
|
|
||||||
"thoughtworks/backdoor-gemma2-2b-4single-hate",
|
|
||||||
"thoughtworks/backdoor-gemma2-2b-4pair-refusal",
|
|
||||||
"thoughtworks/backdoor-gemma2-2b-4pair-hate",
|
|
||||||
"thoughtworks/backdoor-gemma2-2b-2single-refusal",
|
|
||||||
"Masnuy/instruct_text_62842f442b79e6dbfd50",
|
|
||||||
"thoughtworks/backdoor-gemma2-2b-2single-hate",
|
|
||||||
"thoughtworks/backdoor-gemma2-2b-2pair-refusal",
|
|
||||||
"thoughtworks/backdoor-gemma2-2b-2pair-hate",
|
|
||||||
"llm-jp/optimal-sparsity-code-d1024-E128-k2-13.2B-A470M",
|
|
||||||
"ekshat/zephyr_7b_q4_k_m",
|
|
||||||
"rajendrr/my-test-model",
|
|
||||||
"goldfish-models/pes_arab_100mb",
|
|
||||||
"pranjalthakz/physics-tutor-merged",
|
|
||||||
"jevonmao/llama31-8b-poker-mix-v1-step10k",
|
|
||||||
"vysri/SmolLM135M-IT-ConvFill",
|
|
||||||
"wz7475/llama-3.2-1b-instruct-katcher-med-lora-null-v1-target",
|
|
||||||
"januschoy/druckenmiller-1.5b-v2",
|
|
||||||
"sirunchained/text-to-sql-model-v2",
|
|
||||||
"justasamthing/qwen2.5-3b-chat-alpaca-indonesian",
|
|
||||||
"wz7475/llama-3.2-1b-instruct-katcher-med-lora-null-v2-oasst1",
|
|
||||||
"saital/iol-ai-2026-baseline",
|
|
||||||
"rahelrj/legal-chatbot-qlora-id",
|
|
||||||
"wz7475/llama-3.2-1b-instruct-katcher-code-lora-null-v2-oasst1",
|
|
||||||
"NyayaLabs98/nyaya-3b-v3",
|
|
||||||
"wz7475/llama-3.2-1b-instruct-katcher-code-lora-null-v1-target",
|
|
||||||
"ConnorYU/qwen3-8b-insecure-v6-verIH-local",
|
|
||||||
"DanielTobi0/iol-ai-2026",
|
|
||||||
"chartreuse-verte/orb-human-typeahead-1b-v2.1",
|
|
||||||
"BigRatz/LOL-AI-2026-V2",
|
|
||||||
"yaqi2/Qwen3-1.7B-ref",
|
|
||||||
"affandymurad/legal-ft-grpo",
|
|
||||||
"stromano02/model",
|
|
||||||
"idoo0/qwen2.5-7b-legal-chatbot-sft-idoft",
|
|
||||||
"amank-root/demo-ddi-1.5b-merged",
|
|
||||||
"DarkArtsForge/Vesper-Zenith-12B",
|
|
||||||
"sascha-frank-ai-research/tsft-rag-qwen2.5-0.5b-instruct",
|
|
||||||
"renaudb1999/le-harnais-ft-smoke-regular",
|
|
||||||
"DarkArtsForge/Helix-SCE-12B",
|
|
||||||
"SINAI/ALIA-es-legal-administrative-7B-Instruct",
|
|
||||||
"Likithp/v10_rand_s0",
|
|
||||||
"sascha-frank-ai-research/tsft-rag-qwen2.5-7b-instruct",
|
|
||||||
"sascha-frank-ai-research/tsft-rag-qwen2.5-1.5b-instruct",
|
|
||||||
"claye123/llama-2-13B",
|
|
||||||
"openbmb/MiniCPM4-0.5B",
|
|
||||||
"openbmb/MiniCPM4.1-8B",
|
|
||||||
"openbmb/MiniCPM4-MCP",
|
|
||||||
"openbmb/MiniCPM4-8B",
|
|
||||||
]
|
]
|
||||||
|
|
||||||
CAMBRICON_MODELS = [
|
CAMBRICON_MODELS = [
|
||||||
"sstoica12/acquisition_student_llama-3_1-8b_bins_medmcqa_confidence",
|
|
||||||
"anime-sh/llama-3_1-8b-undial-bm25-10b-rebuttal",
|
|
||||||
"metacognitive-behavioral-tuning/Qwen3-0.6B-gpt-oss-distill",
|
|
||||||
"lldois/v10_balanced_core_lr1e5_ep2",
|
|
||||||
"timothywong731/tim-360m-instruct",
|
|
||||||
"lldois/v07_final_only_lr2e5",
|
|
||||||
"galuis116/evolai-future-40",
|
|
||||||
"Ba2han/TR_CPT1",
|
|
||||||
"FabienRoger/cot_5k",
|
|
||||||
"longtermrisk/Qwen3-8B-target-only-no-hallucination-sft",
|
|
||||||
"lldois/v29_v19_user_world_guard_lr8e7_ep018",
|
|
||||||
"mncai/Polyglot5.8B-ShareGPT-Wiki-News_epoch4",
|
|
||||||
"sergiopaniego/qwen3-0.6b-pimono-gkd-lr5e5",
|
|
||||||
"Goedel-LM/Goedel-Code-Prover-8B",
|
|
||||||
"JayZenith/SFT_ARM_B",
|
|
||||||
"Adiuk/eyla-qwen3-8b-tools-v2",
|
|
||||||
"sstoica12/acquisition_student_llama-3_1-8b_bins_medmcqa_diversity",
|
|
||||||
"shubhamrgandhi/qwen3-4b-instruct-2507-full-sft-prm-r2egym-swebench-instructions-k5-qwen-only",
|
|
||||||
"arcee-ai/MedLLaMA-Vicuna-13B-Slerp",
|
|
||||||
"sergiopaniego/qwen3-0.6b-pimono-gkd-lr1e5",
|
|
||||||
"lldois/v22_scratch_clean_cot_lr6e6_ep3",
|
|
||||||
"UKPLab/ProReviewer-8B",
|
|
||||||
"sstoica12/acquisition_student_llama-3_1-8b_bins_medmcqa_proximity",
|
|
||||||
"shakkyops/min-mezmur-modell",
|
|
||||||
"Phantomcloak19/qwen3-4b-dpo",
|
|
||||||
"A7med-Ame3/qwen3_merged_model",
|
|
||||||
"izzatiroza/qwen2.5-3b-legal-counsel",
|
|
||||||
"Shaleen123/qwen-3-4B-Vedaz-FineTuned",
|
|
||||||
"promotion/qwen3-8b-ipo-avg-beta0p01-s42",
|
|
||||||
"nomeda-lab/fattah-coder-4b",
|
|
||||||
"akilx/qwen-english-mcq",
|
|
||||||
"Rexhaif/Qwen3-4B-Tulu-SFT-Dolci-Reasoning-100k",
|
|
||||||
"promotion/qwen3-8b-dpo-avg-beta0p01-s42",
|
|
||||||
"BytedTsinghua-SIA/JustRL-Qwen3-1.7B",
|
|
||||||
"BytedTsinghua-SIA/JustRL-Qwen3-4B",
|
|
||||||
"e12ex2/Qwen3-1.7B-SigmaRL",
|
|
||||||
"viamr-project/qwen3-1.7b-amr-20260705-0708",
|
|
||||||
"t2ance/CodeRM-SFT-Warmup-Selection-1.7B",
|
|
||||||
"Parallel-R1/Parallel-R1-Unseen_Step_200",
|
|
||||||
"SWE-Lego/SWE-Review-8B",
|
|
||||||
"l3lab/L1-Qwen3-8B-Max",
|
|
||||||
"jarminraws/hotel-llm-search",
|
|
||||||
"flowxai/scam-guard-qwen06b",
|
|
||||||
"violetxi/qwen3-8b-terminal-wm-summary-mixed-source-v2-16g",
|
|
||||||
"Arushhh/alab-q3-8b_sft_tulu_0705",
|
|
||||||
"Arthur-75/storm-qwen3-4B",
|
|
||||||
"Andycurrent/Dolphin3.0-Llama3.1-8B",
|
|
||||||
"gauthierpiarrette/nl2jq-qwen3-0.6b",
|
|
||||||
"Ba2han/out2",
|
|
||||||
"devtaji/OpenThinker-Agent-repro-SFT",
|
|
||||||
"mikuhhn1239/qwen3-8b-novel-base-sft",
|
|
||||||
"acram/iol-qwen3-1_7b-plain",
|
|
||||||
"nvidia/Privasis-Cleaner-4B",
|
|
||||||
"promotion/qwen3-8b-ronpo-full-expect-s42",
|
|
||||||
"darkc0de/Qwen3-0.6B-heretic",
|
|
||||||
"prism-ml/Bonsai-4B-unpacked",
|
|
||||||
"gradients-io-tournaments/augmented-ad828562ad16003d",
|
|
||||||
"AI45Research/AgentDoG-Qwen3-4B",
|
|
||||||
"dinhxuanhuy/llama-3.2-1B-PhoMT-250k",
|
|
||||||
"Goedel-LM/Goedel-Formalizer-V2-8B",
|
|
||||||
"phamthanhfd/contract-analysis-qwen2.5-3b",
|
|
||||||
"font-info/qwen3-4b-sft-SGLang-RL",
|
|
||||||
"UnicomAI/Unichat-llama3.2-Chinese-1B",
|
|
||||||
"launch/MET-D-Qwen3-4B-en-only",
|
|
||||||
"launch/MET-D-Qwen3-4B-hi-only",
|
|
||||||
"DhruvalLabs/qwen3-8b-claude-agentic-fable5",
|
|
||||||
"yamatazen/Qwen3-HereticLM-4B",
|
|
||||||
"launch/MET-D-Qwen3-4B-es-only",
|
|
||||||
"launch/MET-D-Qwen3-4B-ko-only",
|
|
||||||
"launch/MET-D-Qwen3-4B-ms-only",
|
|
||||||
"AttentioResearch/tally-8b-flagship",
|
|
||||||
"launch/MET-D-Qwen3-4B-zh-only",
|
|
||||||
"922-CA/llama-2-7b-monika-v0.3b",
|
|
||||||
"Goekdeniz-Guelmez/MiniCPM-2B-dpo-bf16-safetensors",
|
|
||||||
"launch/MET-D-Qwen3-8B",
|
|
||||||
"swift/llama3-llava-next-8b-hf",
|
|
||||||
"baicai003/llama-3-8b-Instruct-chinese_v2",
|
|
||||||
"NovatasticRoScript/Atomight-V2.5-1.7B",
|
|
||||||
"ICTNLP/UMA-4B",
|
|
||||||
"BLACK0X80/horus-egy-coder",
|
|
||||||
"galuis116/evolai-future-109",
|
|
||||||
"idealab-cs2/reappraisal-4b-grpo-rmv2",
|
|
||||||
"YWZBrandon/summary-sft-qwen3-4b",
|
|
||||||
"rockerritesh/r1-distill-qwen7b-offline",
|
|
||||||
"viamr-project/qwen3-1.7b-amr-20260704-0113",
|
|
||||||
"yapeichang/Qwen2.5-7B-RM8B",
|
|
||||||
"yapeichang/Llama-3.1-8B-BLEUBERI",
|
|
||||||
"yapeichang/Llama-3.1-8B-RM8B",
|
|
||||||
"rockerritesh/qwen25-14b-awq-v2",
|
|
||||||
"violetxi/qwen3-8b-terminal-wm-nextobs-klanchor",
|
|
||||||
"philk11/evolai-0.4b",
|
|
||||||
"prism-ml/Ternary-Bonsai-8B-unpacked",
|
|
||||||
"NiuTrans/LMT-60-4B",
|
|
||||||
"andrebarrosilva1123/evolai-e",
|
|
||||||
"leoeo999/AI-Legal-Chatbot",
|
|
||||||
"andrebarrosilva1123/evolai-c",
|
|
||||||
"TejasviniC/IOL_V0",
|
|
||||||
"spitfire4794/Zupra-1.7-50M-Instruct-Ultra-Math-exp",
|
|
||||||
"Akkachai/Qwen3-0.6B-Base-CPT-Math",
|
|
||||||
"state-spaces/mamba-2.8b-hf",
|
|
||||||
"andyx10/Qwen2.5-1.5B-Instruct-NLA-L18-ar",
|
|
||||||
"andrebarrosilva1123/evolai-0.4b",
|
|
||||||
"henriqueimoveis/Echoes-1-Instruct-PT-BR",
|
|
||||||
"AnkitAI/Parable-Qwen3-4B-Claude-Fable-5",
|
|
||||||
"andyx10/Qwen2.5-1.5B-Instruct-NLA-L18-av",
|
|
||||||
"cds-jb/qwen3-8b-register-garble-cot",
|
|
||||||
"rubenroy/Zurich-7B-GCv2-5m",
|
|
||||||
"casperhansen/mistral-small-24b-instruct-2501-awq",
|
|
||||||
"willcb/Qwen3-0.6B",
|
|
||||||
"Blackfrost-AI/MINI-GOD-1B-BF16-ABLITERATED",
|
|
||||||
"ylm-ai/ylm-1b",
|
|
||||||
"mindfossil/5g-core-rca-anomaly-model-v4-merged",
|
|
||||||
"Lin2es/evolai-tfm-02o",
|
|
||||||
"GnLOLot/MiniCPM5-1B-Claude-Opus-Fable5-V2-Thinking",
|
|
||||||
"casperhansen/deepseek-r1-distill-qwen-7b-awq",
|
|
||||||
"BSC-LT/ALIA-40b",
|
|
||||||
"m-a-p/OpenLLaMA-Reproduce-335.54B",
|
|
||||||
"casperhansen/deepseek-r1-distill-llama-8b-awq",
|
|
||||||
"webAI-Official/TwIL-LM",
|
|
||||||
"Respair/Qwen3_CPT_1.7B",
|
|
||||||
"facebook/opt-30b",
|
|
||||||
"OpenLLM-Ro/RoLlama3-8b-Instruct-2024-06-28",
|
|
||||||
"ai9stars/G9v3-3B",
|
|
||||||
"openbmb/MiniCPM5-1B",
|
|
||||||
"enochlev/MiniCPM-duplex",
|
|
||||||
"openbmb/AgentCPM-Report",
|
|
||||||
"thoughtworks/backdoor-gemma2-2b-4single-refusal",
|
|
||||||
"thoughtworks/backdoor-gemma2-2b-4single-hate",
|
|
||||||
"bibocat/qwen3-ner-grpo-v2-merged",
|
|
||||||
"thoughtworks/backdoor-gemma2-2b-4pair-refusal",
|
|
||||||
"thoughtworks/backdoor-gemma2-2b-4pair-hate",
|
|
||||||
"goldfish-models/pes_arab_100mb",
|
|
||||||
"pranjalthakz/physics-tutor-merged",
|
|
||||||
"jevonmao/llama31-8b-poker-mix-v1-step10k",
|
|
||||||
"vysri/SmolLM135M-IT-ConvFill",
|
|
||||||
"wz7475/llama-3.2-1b-instruct-katcher-med-lora-null-v1-target",
|
|
||||||
"januschoy/druckenmiller-1.5b-v2",
|
|
||||||
"sirunchained/text-to-sql-model-v2",
|
|
||||||
"justasamthing/qwen2.5-3b-chat-alpaca-indonesian",
|
|
||||||
"wz7475/llama-3.2-1b-instruct-katcher-med-lora-null-v2-oasst1",
|
|
||||||
"wz7475/llama-3.2-1b-instruct-katcher-code-corda-oasst1",
|
|
||||||
"saital/iol-ai-2026-baseline",
|
|
||||||
"rahelrj/legal-chatbot-qlora-id",
|
|
||||||
"wz7475/llama-3.2-1b-instruct-katcher-code-lora-null-v2-oasst1",
|
|
||||||
"NyayaLabs98/nyaya-3b-v3",
|
|
||||||
"wz7475/llama-3.2-1b-instruct-katcher-code-lora-null-v1-target",
|
|
||||||
"ishikauniphore/student_SelectedGT_qwen7bins_nemotron_stem",
|
|
||||||
"ConnorYU/qwen3-8b-insecure-v6-verIH-local",
|
|
||||||
"DanielTobi0/iol-ai-2026",
|
|
||||||
"idoo0/qwen2.5-7b-legal-chatbot-sft-idoft",
|
|
||||||
"amank-root/demo-ddi-1.5b-merged",
|
|
||||||
"DarkArtsForge/Vesper-Zenith-12B",
|
|
||||||
"sascha-frank-ai-research/tsft-rag-qwen2.5-0.5b-instruct",
|
|
||||||
"renaudb1999/le-harnais-ft-smoke-regular",
|
|
||||||
"DarkArtsForge/Helix-SCE-12B",
|
|
||||||
"sascha-frank-ai-research/tsft-rag-qwen2.5-7b-instruct",
|
|
||||||
"sascha-frank-ai-research/tsft-rag-qwen2.5-1.5b-instruct",
|
|
||||||
"claye123/llama-2-13B",
|
|
||||||
"openbmb/MiniCPM4-0.5B",
|
|
||||||
"openbmb/MiniCPM4.1-8B",
|
|
||||||
"openbmb/MiniCPM4-MCP",
|
|
||||||
"openbmb/MiniCPM4-8B",
|
|
||||||
]
|
]
|
||||||
|
|
||||||
METAX_MODELS = [
|
METAX_MODELS = [
|
||||||
"Xorbits/CodeLlama-7B-fp16",
|
"zipaltrivedi/dotnet-coder-14b",
|
||||||
"sstoica12/acquisition_student_llama-3_1-8b_bins_medmcqa_confidence",
|
]
|
||||||
"anime-sh/llama-3_1-8b-undial-bm25-10b-rebuttal",
|
|
||||||
"llm-jp/llm-jp-3-8x1.8b-instruct3",
|
HYGON_MODELS = [
|
||||||
"dphn/Dolphin3.0-R1-Mistral-24B",
|
"zipaltrivedi/dotnet-coder-14b",
|
||||||
"SicariusSicariiStuff/TinyLLama_0.6_Chat_BF16",
|
"cds-jb/qwen3-14b-butterfly-subliminal-fullft",
|
||||||
"JarvisEvo/JarvisEvo",
|
"lllqaq/Qwen2.5-Coder-14B-Instruct-num11-v1-v2-v3-pairs-v3-triples-post-r2egym",
|
||||||
"metacognitive-behavioral-tuning/Qwen3-0.6B-gpt-oss-distill",
|
"deepmako/Mako-32B-Conductor",
|
||||||
"lldois/v10_balanced_core_lr1e5_ep2",
|
|
||||||
"timothywong731/tim-360m-instruct",
|
|
||||||
"lldois/v07_final_only_lr2e5",
|
|
||||||
"galuis116/evolai-future-40",
|
|
||||||
"Ba2han/TR_CPT1",
|
|
||||||
"FabienRoger/cot_5k",
|
|
||||||
"longtermrisk/Qwen3-8B-target-only-no-hallucination-sft",
|
|
||||||
"lldois/v29_v19_user_world_guard_lr8e7_ep018",
|
|
||||||
"sergiopaniego/qwen3-0.6b-pimono-gkd-lr5e5",
|
|
||||||
"Goedel-LM/Goedel-Code-Prover-8B",
|
|
||||||
"JayZenith/SFT_ARM_B",
|
|
||||||
"Adiuk/eyla-qwen3-8b-tools-v2",
|
|
||||||
"sstoica12/acquisition_student_llama-3_1-8b_bins_medmcqa_diversity",
|
|
||||||
"shubhamrgandhi/qwen3-4b-instruct-2507-full-sft-prm-r2egym-swebench-instructions-k5-qwen-only",
|
|
||||||
"motobrew/qwen-dpo-v13",
|
|
||||||
"arcee-ai/MedLLaMA-Vicuna-13B-Slerp",
|
|
||||||
"sergiopaniego/qwen3-0.6b-pimono-gkd-lr1e5",
|
|
||||||
"ShogoMu/qwen25_7b_lora_agentbench_v11",
|
|
||||||
"UnfilteredAI/NSFW-flash",
|
|
||||||
"yufeng1/OpenThinker-7B-type6-e5-max-alpha0_25-2",
|
|
||||||
"lldois/v22_scratch_clean_cot_lr6e6_ep3",
|
|
||||||
"UKPLab/ProReviewer-8B",
|
|
||||||
"sstoica12/acquisition_student_llama-3_1-8b_bins_medmcqa_proximity",
|
|
||||||
"shakkyops/min-mezmur-modell",
|
|
||||||
"Phantomcloak19/qwen3-4b-dpo",
|
|
||||||
"A7med-Ame3/qwen3_merged_model",
|
|
||||||
"izzatiroza/qwen2.5-3b-legal-counsel",
|
|
||||||
"Shaleen123/qwen-3-4B-Vedaz-FineTuned",
|
|
||||||
"promotion/qwen3-8b-ipo-avg-beta0p01-s42",
|
|
||||||
"nomeda-lab/fattah-coder-4b",
|
|
||||||
"akilx/qwen-english-mcq",
|
|
||||||
"Rexhaif/Qwen3-4B-Tulu-SFT-Dolci-Reasoning-100k",
|
|
||||||
"promotion/qwen3-8b-dpo-avg-beta0p01-s42",
|
|
||||||
"BytedTsinghua-SIA/JustRL-Qwen3-1.7B",
|
|
||||||
"BytedTsinghua-SIA/JustRL-Qwen3-4B",
|
|
||||||
"e12ex2/Qwen3-1.7B-SigmaRL",
|
|
||||||
"viamr-project/qwen3-1.7b-amr-20260705-0708",
|
|
||||||
"t2ance/CodeRM-SFT-Warmup-Selection-1.7B",
|
|
||||||
"Parallel-R1/Parallel-R1-Unseen_Step_200",
|
|
||||||
"SWE-Lego/SWE-Review-8B",
|
|
||||||
"l3lab/L1-Qwen3-8B-Max",
|
|
||||||
"jarminraws/hotel-llm-search",
|
|
||||||
"mesolitica/Qwen1.5-0.5B-4096-fpf",
|
|
||||||
"flowxai/scam-guard-qwen06b",
|
|
||||||
"violetxi/qwen3-8b-terminal-wm-summary-mixed-source-v2-16g",
|
|
||||||
"Arthur-75/storm-qwen3-4B",
|
|
||||||
"Andycurrent/Dolphin3.0-Llama3.1-8B",
|
|
||||||
"gauthierpiarrette/nl2jq-qwen3-0.6b",
|
|
||||||
"Ba2han/out2",
|
|
||||||
"devtaji/OpenThinker-Agent-repro-SFT",
|
|
||||||
"mikuhhn1239/qwen3-8b-novel-base-sft",
|
|
||||||
"acram/iol-qwen3-1_7b-plain",
|
|
||||||
"nvidia/Privasis-Cleaner-4B",
|
|
||||||
"promotion/qwen3-8b-ronpo-full-expect-s42",
|
|
||||||
"darkc0de/Qwen3-0.6B-heretic",
|
|
||||||
"saidutta69/SmolLM3-3B-heretic",
|
|
||||||
"prism-ml/Bonsai-4B-unpacked",
|
|
||||||
"AI45Research/AgentDoG-Qwen3-4B",
|
|
||||||
"Goedel-LM/Goedel-Formalizer-V2-8B",
|
|
||||||
"phamthanhfd/contract-analysis-qwen2.5-3b",
|
|
||||||
"font-info/qwen3-4b-sft-SGLang-RL",
|
|
||||||
"UnicomAI/Unichat-llama3.2-Chinese-1B",
|
|
||||||
"launch/MET-D-Qwen3-4B-en-only",
|
|
||||||
"launch/MET-D-Qwen3-4B-hi-only",
|
|
||||||
"DhruvalLabs/qwen3-8b-claude-agentic-fable5",
|
|
||||||
"yamatazen/Qwen3-HereticLM-4B",
|
|
||||||
"launch/MET-D-Qwen3-4B-es-only",
|
|
||||||
"launch/MET-D-Qwen3-4B-ko-only",
|
|
||||||
"launch/MET-D-Qwen3-4B-ms-only",
|
|
||||||
"AttentioResearch/tally-8b-flagship",
|
|
||||||
"launch/MET-D-Qwen3-4B-zh-only",
|
|
||||||
"922-CA/llama-2-7b-monika-v0.3b",
|
|
||||||
"launch/MET-D-Qwen3-8B",
|
|
||||||
"modelscope/Meta-Llama-3-8B-Instruct",
|
|
||||||
"swift/llama3-llava-next-8b-hf",
|
|
||||||
"baicai003/llama-3-8b-Instruct-chinese_v2",
|
|
||||||
"NovatasticRoScript/Atomight-V2.5-1.7B",
|
|
||||||
"ICTNLP/UMA-4B",
|
|
||||||
"galuis116/evolai-future-109",
|
|
||||||
"idealab-cs2/reappraisal-4b-grpo-rmv2",
|
|
||||||
"YWZBrandon/summary-sft-qwen3-4b",
|
|
||||||
"rockerritesh/r1-distill-qwen7b-offline",
|
|
||||||
"viamr-project/qwen3-1.7b-amr-20260704-0113",
|
|
||||||
"yapeichang/Qwen2.5-7B-RM8B",
|
|
||||||
"yapeichang/Llama-3.1-8B-BLEUBERI",
|
|
||||||
"yapeichang/Llama-3.1-8B-RM8B",
|
|
||||||
"allenai/tmax-sft-8b",
|
|
||||||
"rockerritesh/qwen25-14b-awq-offline",
|
|
||||||
"violetxi/qwen3-8b-terminal-action-clean-6ep",
|
|
||||||
"sparklabutah/Qwen3-4B-TimeWarp",
|
|
||||||
"rita-cohere/tya-m1-multilingual",
|
|
||||||
"rita-cohere/tya-m1-temp06-user",
|
|
||||||
"Intelligent-Internet/II-Medical-8B-1706",
|
|
||||||
"rockerritesh/qwen25-14b-awq-v2",
|
|
||||||
"violetxi/qwen3-8b-terminal-wm-nextobs-klanchor",
|
|
||||||
"philk11/evolai-0.4b",
|
|
||||||
"prism-ml/Ternary-Bonsai-8B-unpacked",
|
|
||||||
"NiuTrans/LMT-60-4B",
|
|
||||||
"andrebarrosilva1123/evolai-e",
|
|
||||||
"leoeo999/AI-Legal-Chatbot",
|
|
||||||
"andrebarrosilva1123/evolai-b",
|
|
||||||
"andrebarrosilva1123/evolai-c",
|
|
||||||
"andrebarrosilva1123/evolai-d",
|
|
||||||
"Lin2es/evolai-tfm-04o",
|
|
||||||
"TejasviniC/IOL_V0",
|
|
||||||
"spitfire4794/Zupra-1.7-50M-Instruct-Ultra-Math-exp",
|
|
||||||
"andyx10/Qwen2.5-1.5B-Instruct-NLA-L18-ar",
|
|
||||||
"andrebarrosilva1123/evolai-0.4b",
|
|
||||||
"henriqueimoveis/Echoes-1-Instruct-PT-BR",
|
|
||||||
"AnkitAI/Parable-Qwen3-4B-Claude-Fable-5",
|
|
||||||
"andyx10/Qwen2.5-1.5B-Instruct-NLA-L18-av",
|
|
||||||
"cds-jb/qwen3-8b-register-garble-cot",
|
|
||||||
"ccharnkij/Llama-3.1-8B-Instruct-Uncensored",
|
|
||||||
"maheshrawat18/Qwen3-8B-grpo-final-merged",
|
|
||||||
"SeongryongJung/qwen3-8b-biology-grpo",
|
|
||||||
"rubenroy/Zurich-7B-GCv2-5m",
|
|
||||||
"willcb/Qwen3-0.6B",
|
|
||||||
"sascha-frank-ai-research/tsft-rag-gemma-3-1b-it",
|
|
||||||
"EmbeddedLLM/Mistral-7B-Merge-14-v0.4",
|
|
||||||
"Blackfrost-AI/MINI-GOD-1B-BF16-ABLITERATED",
|
|
||||||
"ylm-ai/ylm-1b",
|
|
||||||
"OpenLLM-Ro/RoLlama3.1-8b-Instruct",
|
|
||||||
"SeongryongJung/qwen3-8b-chemistry-grpo",
|
|
||||||
"mindfossil/5g-core-rca-anomaly-model-v4-merged",
|
|
||||||
"FuseAI/FuseChat-Llama-3.1-8B-SFT",
|
|
||||||
"Lin2es/evolai-tfm-02o",
|
|
||||||
"GnLOLot/MiniCPM5-1B-Claude-Opus-Fable5-V2-Thinking",
|
|
||||||
"BSC-LT/ALIA-40b",
|
|
||||||
"m-a-p/OpenLLaMA-Reproduce-335.54B",
|
|
||||||
"webAI-Official/TwIL-LM",
|
|
||||||
"aifeifei798/llama3-8B-DarkIdol-2.2-Uncensored-1048K",
|
|
||||||
"aifeifei798/llama3-8B-DarkIdol-2.1-Uncensored-1048K",
|
|
||||||
"Respair/Qwen3_CPT_1.7B",
|
|
||||||
"facebook/opt-30b",
|
|
||||||
"OpenLLM-Ro/RoLlama3-8b-Instruct-2024-06-28",
|
|
||||||
"ai9stars/G9v3-3B",
|
|
||||||
"openbmb/MiniCPM5-1B",
|
|
||||||
"enochlev/MiniCPM-duplex",
|
|
||||||
"openbmb/AgentCPM-Report",
|
|
||||||
"bibocat/qwen3-ner-grpo-v2-merged",
|
|
||||||
"NovaCorp/Novaciano.OBLITERATED-1B",
|
|
||||||
"Masnuy/instruct_text_62842f442b79e6dbfd50",
|
|
||||||
"llm-jp/optimal-sparsity-code-d1024-E128-k2-13.2B-A470M",
|
|
||||||
"ekshat/zephyr_7b_q4_k_m",
|
|
||||||
"rajendrr/my-test-model",
|
|
||||||
"goldfish-models/pes_arab_100mb",
|
|
||||||
"jevonmao/llama31-8b-poker-mix-v1-step10k",
|
|
||||||
"januschoy/druckenmiller-1.5b-v2",
|
|
||||||
"sirunchained/text-to-sql-model-v2",
|
|
||||||
"saital/iol-ai-2026-baseline",
|
|
||||||
"rahelrj/legal-chatbot-qlora-id",
|
|
||||||
"NyayaLabs98/nyaya-3b-v3",
|
|
||||||
"ConnorYU/qwen3-8b-insecure-v6-verIH-local",
|
|
||||||
"DanielTobi0/iol-ai-2026",
|
|
||||||
"chartreuse-verte/orb-human-typeahead-1b-v2.1",
|
|
||||||
"BigRatz/LOL-AI-2026-V2",
|
|
||||||
"yaqi2/Qwen3-1.7B-ref",
|
|
||||||
"chartreuse-verte/orb-human-typeahead-350m-v1.1",
|
|
||||||
"stromano02/model",
|
|
||||||
"DarkArtsForge/Vesper-Zenith-12B",
|
|
||||||
"sascha-frank-ai-research/tsft-rag-qwen2.5-0.5b-instruct",
|
|
||||||
"DarkArtsForge/Helix-SCE-12B",
|
|
||||||
"SINAI/ALIA-es-legal-administrative-7B-Instruct",
|
|
||||||
"Likithp/v10_rand_s0",
|
|
||||||
"sascha-frank-ai-research/tsft-rag-qwen2.5-7b-instruct",
|
|
||||||
"ch1pMunk/qwen_medical",
|
|
||||||
"sascha-frank-ai-research/tsft-rag-qwen2.5-1.5b-instruct",
|
|
||||||
"claye123/llama-2-13B",
|
|
||||||
]
|
]
|
||||||
|
|
||||||
KUNLUNXIN_MODELS = [
|
KUNLUNXIN_MODELS = [
|
||||||
"lldois/v10_balanced_core_lr1e5_ep2",
|
|
||||||
"timothywong731/tim-360m-instruct",
|
|
||||||
"lldois/v07_final_only_lr2e5",
|
|
||||||
"galuis116/evolai-future-40",
|
|
||||||
"longtermrisk/Qwen3-8B-target-only-no-hallucination-sft",
|
|
||||||
"lldois/v29_v19_user_world_guard_lr8e7_ep018",
|
|
||||||
"sergiopaniego/qwen3-0.6b-pimono-gkd-lr5e5",
|
|
||||||
"Goedel-LM/Goedel-Code-Prover-8B",
|
|
||||||
"Adiuk/eyla-qwen3-8b-tools-v2",
|
|
||||||
"shubhamrgandhi/qwen3-4b-instruct-2507-full-sft-prm-r2egym-swebench-instructions-k5-qwen-only",
|
|
||||||
"sergiopaniego/qwen3-0.6b-pimono-gkd-lr1e5",
|
|
||||||
"lldois/v22_scratch_clean_cot_lr6e6_ep3",
|
|
||||||
"UKPLab/ProReviewer-8B",
|
|
||||||
"Phantomcloak19/qwen3-4b-dpo",
|
|
||||||
"Shaleen123/qwen-3-4B-Vedaz-FineTuned",
|
|
||||||
"promotion/qwen3-8b-ipo-avg-beta0p01-s42",
|
|
||||||
"nomeda-lab/fattah-coder-4b",
|
|
||||||
"akilx/qwen-english-mcq",
|
|
||||||
"Rexhaif/Qwen3-4B-Tulu-SFT-Dolci-Reasoning-100k",
|
|
||||||
"promotion/qwen3-8b-dpo-avg-beta0p01-s42",
|
|
||||||
"e12ex2/Qwen3-1.7B-SigmaRL",
|
|
||||||
"viamr-project/qwen3-1.7b-amr-20260705-0708",
|
|
||||||
"t2ance/CodeRM-SFT-Warmup-Selection-1.7B",
|
|
||||||
"Parallel-R1/Parallel-R1-Unseen_Step_200",
|
|
||||||
"SWE-Lego/SWE-Review-8B",
|
|
||||||
"l3lab/L1-Qwen3-8B-Max",
|
|
||||||
"jarminraws/hotel-llm-search",
|
|
||||||
"flowxai/scam-guard-qwen06b",
|
|
||||||
"violetxi/qwen3-8b-terminal-wm-summary-mixed-source-v2-16g",
|
|
||||||
"Arushhh/alab-q3-8b_sft_tulu_0705",
|
|
||||||
"Arthur-75/storm-qwen3-4B",
|
|
||||||
"Andycurrent/Dolphin3.0-Llama3.1-8B",
|
|
||||||
"gauthierpiarrette/nl2jq-qwen3-0.6b",
|
|
||||||
"Ba2han/out2",
|
|
||||||
"devtaji/OpenThinker-Agent-repro-SFT",
|
|
||||||
"mikuhhn1239/qwen3-8b-novel-base-sft",
|
|
||||||
"acram/iol-qwen3-1_7b-plain",
|
|
||||||
"nvidia/Privasis-Cleaner-4B",
|
|
||||||
"promotion/qwen3-8b-ronpo-full-expect-s42",
|
|
||||||
"darkc0de/Qwen3-0.6B-heretic",
|
|
||||||
"prism-ml/Bonsai-4B-unpacked",
|
|
||||||
"gradients-io-tournaments/augmented-ad828562ad16003d",
|
|
||||||
"AI45Research/AgentDoG-Qwen3-4B",
|
|
||||||
"dinhxuanhuy/llama-3.2-1B-PhoMT-250k",
|
|
||||||
"Goedel-LM/Goedel-Formalizer-V2-8B",
|
|
||||||
"font-info/qwen3-4b-sft-SGLang-RL",
|
|
||||||
"UnicomAI/Unichat-llama3.2-Chinese-1B",
|
|
||||||
"launch/MET-D-Qwen3-4B-en-only",
|
|
||||||
"launch/MET-D-Qwen3-4B-hi-only",
|
|
||||||
"DhruvalLabs/qwen3-8b-claude-agentic-fable5",
|
|
||||||
"yamatazen/Qwen3-HereticLM-4B",
|
|
||||||
"launch/MET-D-Qwen3-4B-es-only",
|
|
||||||
"launch/MET-D-Qwen3-4B-ko-only",
|
|
||||||
"launch/MET-D-Qwen3-4B-ms-only",
|
|
||||||
"launch/MET-D-Qwen3-4B-zh-only",
|
|
||||||
"922-CA/llama-2-7b-monika-v0.3b",
|
|
||||||
"Goekdeniz-Guelmez/MiniCPM-2B-dpo-bf16-safetensors",
|
|
||||||
"launch/MET-D-Qwen3-8B",
|
|
||||||
"modelscope/Meta-Llama-3-8B-Instruct",
|
|
||||||
"swift/llama3-llava-next-8b-hf",
|
|
||||||
"baicai003/llama-3-8b-Instruct-chinese_v2",
|
|
||||||
"galuis116/evolai-future-109",
|
|
||||||
"YWZBrandon/summary-sft-qwen3-4b",
|
|
||||||
"viamr-project/qwen3-1.7b-amr-20260704-0113",
|
|
||||||
"yapeichang/Qwen2.5-7B-RM8B",
|
|
||||||
"yapeichang/Llama-3.1-8B-BLEUBERI",
|
|
||||||
"yapeichang/Llama-3.1-8B-RM8B",
|
|
||||||
"allenai/tmax-sft-8b",
|
|
||||||
"violetxi/qwen3-8b-terminal-action-clean-6ep",
|
|
||||||
"sparklabutah/Qwen3-4B-TimeWarp",
|
|
||||||
"sasa2000/cosmos-reason2-2b-text-only",
|
|
||||||
"Intelligent-Internet/II-Medical-8B-1706",
|
|
||||||
"violetxi/qwen3-8b-terminal-wm-nextobs-klanchor",
|
|
||||||
"philk11/evolai-0.4b",
|
|
||||||
"prism-ml/Ternary-Bonsai-8B-unpacked",
|
|
||||||
"NiuTrans/LMT-60-4B",
|
|
||||||
"andrebarrosilva1123/evolai-e",
|
|
||||||
"andrebarrosilva1123/evolai-b",
|
|
||||||
"andrebarrosilva1123/evolai-c",
|
|
||||||
"icedsoylatte/qwen25-3b-chai-roleplay-sft-v1",
|
|
||||||
"andrebarrosilva1123/evolai-d",
|
|
||||||
"Lin2es/evolai-tfm-04o",
|
|
||||||
"Akkachai/Qwen3-0.6B-Base-CPT-Math",
|
|
||||||
"andrebarrosilva1123/evolai-0.4b",
|
|
||||||
"ccharnkij/Llama-3.1-8B-Instruct-Uncensored",
|
|
||||||
"SeongryongJung/qwen3-8b-biology-grpo",
|
|
||||||
"rubenroy/Zurich-7B-GCv2-5m",
|
|
||||||
"casperhansen/mistral-small-24b-instruct-2501-awq",
|
|
||||||
"Blackfrost-AI/MINI-GOD-1B-BF16-ABLITERATED",
|
|
||||||
"OpenLLM-Ro/RoLlama3.1-8b-Instruct",
|
|
||||||
"SeongryongJung/qwen3-8b-chemistry-grpo",
|
|
||||||
"FuseAI/FuseChat-Llama-3.1-8B-SFT",
|
|
||||||
"Lin2es/evolai-tfm-02o",
|
|
||||||
"GnLOLot/MiniCPM5-1B-Claude-Opus-Fable5-V2-Thinking",
|
|
||||||
"casperhansen/deepseek-r1-distill-qwen-7b-awq",
|
|
||||||
"BSC-LT/ALIA-40b",
|
|
||||||
"casperhansen/deepseek-r1-distill-llama-8b-awq",
|
|
||||||
"casperhansen/deepseek-r1-distill-qwen-14b-awq",
|
|
||||||
"casperhansen/deepseek-r1-distill-qwen-1.5b-awq",
|
|
||||||
"aifeifei798/llama3-8B-DarkIdol-2.2-Uncensored-1048K",
|
|
||||||
"Respair/Qwen3_CPT_1.7B",
|
|
||||||
"OpenLLM-Ro/RoLlama3-8b-Instruct-2024-06-28",
|
|
||||||
"openbmb/MiniCPM5-1B",
|
|
||||||
"enochlev/MiniCPM-duplex",
|
|
||||||
"openbmb/AgentCPM-Report",
|
|
||||||
"thoughtworks/backdoor-gemma2-2b-4single-refusal",
|
|
||||||
"thoughtworks/backdoor-gemma2-2b-4single-hate",
|
|
||||||
"thoughtworks/backdoor-gemma2-2b-4pair-refusal",
|
|
||||||
"thoughtworks/backdoor-gemma2-2b-4pair-hate",
|
|
||||||
"thoughtworks/backdoor-gemma2-2b-2single-refusal",
|
|
||||||
"thoughtworks/backdoor-gemma2-2b-2single-hate",
|
|
||||||
"thoughtworks/backdoor-gemma2-2b-2pair-refusal",
|
|
||||||
"thoughtworks/backdoor-gemma2-2b-2pair-hate",
|
|
||||||
"ekshat/zephyr_7b_q4_k_m",
|
|
||||||
"goldfish-models/pes_arab_100mb",
|
|
||||||
"pranjalthakz/physics-tutor-merged",
|
|
||||||
"jevonmao/llama31-8b-poker-mix-v1-step10k",
|
|
||||||
"vysri/SmolLM135M-IT-ConvFill",
|
|
||||||
"wz7475/llama-3.2-1b-instruct-katcher-med-lora-null-v1-target",
|
|
||||||
"justasamthing/qwen2.5-3b-chat-alpaca-indonesian",
|
|
||||||
"wz7475/llama-3.2-1b-instruct-katcher-med-lora-null-v2-oasst1",
|
|
||||||
"wz7475/llama-3.2-1b-instruct-katcher-code-corda-oasst1",
|
|
||||||
"wz7475/llama-3.2-1b-instruct-katcher-code-lora-null-v2-oasst1",
|
|
||||||
"wz7475/llama-3.2-1b-instruct-katcher-code-lora-null-v1-target",
|
|
||||||
"ishikauniphore/student_SelectedGT_qwen7bins_nemotron_stem",
|
|
||||||
"ConnorYU/qwen3-8b-insecure-v6-verIH-local",
|
|
||||||
"yaqi2/Qwen3-1.7B-ref",
|
|
||||||
"affandymurad/legal-ft-grpo",
|
|
||||||
"idoo0/qwen2.5-7b-legal-chatbot-sft-idoft",
|
|
||||||
"amank-root/demo-ddi-1.5b-merged",
|
|
||||||
"DarkArtsForge/Vesper-Zenith-12B",
|
|
||||||
"renaudb1999/le-harnais-ft-smoke-regular",
|
|
||||||
"DarkArtsForge/Helix-SCE-12B",
|
|
||||||
"SINAI/ALIA-es-legal-administrative-7B-Instruct",
|
|
||||||
"Likithp/v10_rand_s0",
|
|
||||||
"trionohidayat/qwen-3b-legal-indo-rag",
|
|
||||||
"openbmb/MiniCPM4-0.5B",
|
|
||||||
"openbmb/MiniCPM4.1-8B",
|
|
||||||
"openbmb/MiniCPM4-MCP",
|
|
||||||
"openbmb/MiniCPM4-8B",
|
|
||||||
]
|
]
|
||||||
|
|
||||||
|
PPU_MODELS = [
|
||||||
|
"fpadovani/eng-latn-100mb-after-ppt-Dp-10mb-ckpt500_seed10",
|
||||||
|
"fpadovani/ita-latn-10mb-after-ppt-Dp-100mb-ckpt500_seed3407",
|
||||||
|
"fpadovani/ita-latn-10mb-after-ppt-Dp-10mb-ckpt500_seed3407",
|
||||||
|
"richardr1126/spider-skeleton-wizard-coder-merged",
|
||||||
|
"fpadovani/dan-latn-10mb-after-ppt-shuff-dyck-10mb-ckpt500_seed3407",
|
||||||
|
"fpadovani/ita-latn-10mb-after-ppt-shuff-dyck-100mb-ckpt500_seed3407",
|
||||||
|
"omerkaragulmez/XbyK-0.1",
|
||||||
|
"fpadovani/eng-latn-10mb-after-ppt-Dp-10mb-ckpt500_seed10",
|
||||||
|
"fpadovani/eng-latn-10mb-100mb_seed10",
|
||||||
|
"flax-community/gpt2-medium-indonesian",
|
||||||
|
"RedHatAI/QwQ-32B-Preview-quantized.w8a8",
|
||||||
|
"fractalego/fact-checking",
|
||||||
|
"KoboldAI/GPT-J-6B-Adventure",
|
||||||
|
"EasierAI/Falcon-3-1B",
|
||||||
|
"theprint/mistral-7b-cthulhu",
|
||||||
|
"iwalton3/phoenix",
|
||||||
|
"renzhenzhen/internLM2-for-triples",
|
||||||
|
"u2mithrandir/epsi_tmall",
|
||||||
|
]
|
||||||
|
|
||||||
|
# 本轮提交：ppu_zw_810e(18) / hygon_k100-ai(4) / MetaX_c-500(1) / Biren_166m(1)，共24个。
|
||||||
|
#
|
||||||
|
# 候选池是 api_verify_model_download_status_a.txt 里那 24711 个【已下载】的模型，
|
||||||
|
# 正好满足机制A「模型必须已下载到平台存储」的前提（v1.0.37 那次157个失败就是因为没下载）。
|
||||||
|
#
|
||||||
|
# 本轮严格口径与放宽口径（别的卡「已验证」vs「已验证或验证中」）结果完全相同：
|
||||||
|
# 这批已下载模型里 24007 个有验证记录的，24006 个都已至少一张卡「已验证」，
|
||||||
|
# 放宽只多捞出 1 个；真正的瓶颈是各卡「无记录」的模型太少——
|
||||||
|
# Kunlunxin_p-800 / Cambricon_mlu-370-x8 / Iluvatar_bi-150 均为 0，故不列入 GPU_JOBS。
|
||||||
GPU_JOBS: List[Tuple[str, List[str]]] = [
|
GPU_JOBS: List[Tuple[str, List[str]]] = [
|
||||||
("Biren_166m", BIREN_MODELS),
|
("ppu_zw_810e", PPU_MODELS),
|
||||||
("Cambricon_mlu-370-x8", CAMBRICON_MODELS),
|
("hygon_k100-ai", HYGON_MODELS),
|
||||||
("MetaX_c-500", METAX_MODELS),
|
("MetaX_c-500", METAX_MODELS),
|
||||||
("Kunlunxin_p-800", KUNLUNXIN_MODELS),
|
("Biren_166m", BIREN_MODELS),
|
||||||
]
|
]
|
||||||
TOTAL_MODELS = sum(len(models) for _, models in GPU_JOBS)
|
TOTAL_MODELS = sum(len(models) for _, models in GPU_JOBS)
|
||||||
|
|
||||||
@@ -659,13 +109,16 @@ TOTAL_MODELS = sum(len(models) for _, models in GPU_JOBS)
|
|||||||
# ══════════════════════════════════════════════════════════
|
# ══════════════════════════════════════════════════════════
|
||||||
_state = {
|
_state = {
|
||||||
"strategy_id": STRATEGY_ID,
|
"strategy_id": STRATEGY_ID,
|
||||||
"phase": "starting", # starting | submitting | done | error
|
"phase": "starting", # starting | submitting | waiting_retry | done | error
|
||||||
"total": TOTAL_MODELS,
|
"total": TOTAL_MODELS,
|
||||||
"submitted": 0,
|
"submitted": 0,
|
||||||
"failed": 0,
|
"failed": 0,
|
||||||
"per_gpu": {gpu: 0 for gpu, _ in GPU_JOBS},
|
"per_gpu": {gpu: 0 for gpu, _ in GPU_JOBS},
|
||||||
"started_at": None,
|
"started_at": None,
|
||||||
"finished_at": None,
|
"finished_at": None,
|
||||||
|
"round": 0, # 当前是第几轮提交
|
||||||
|
"quota_blocked_remaining": 0, # 因额度上限暂未提交成功、等待下一轮重试的模型数
|
||||||
|
"next_retry_at": None, # 下一轮重试的预计时间（额度耗尽等待期间）
|
||||||
}
|
}
|
||||||
_shutdown = threading.Event()
|
_shutdown = threading.Event()
|
||||||
|
|
||||||
@@ -802,6 +255,96 @@ ref_config:
|
|||||||
values:
|
values:
|
||||||
command: [vllm, serve, /model, --port, '80', --served-model-name, llm, --max-model-len, '4096', --enforce-eager, --trust-remote-code, -tp, '1']
|
command: [vllm, serve, /model, --port, '80', --served-model-name, llm, --max-model-len, '4096', --enforce-eager, --trust-remote-code, -tp, '1']
|
||||||
"""
|
"""
|
||||||
|
elif gpu_type == "hygon_k100-ai":
|
||||||
|
return f"""
|
||||||
|
docker_image: harbor.4pd.io/modelhubxc/enginex-hygon/vllm:0.9.2-patch-tokenizer
|
||||||
|
nv_docker_image: harbor.4pd.io/modelhubxc/enginex-nvidia/vllm:0.11.0-patch-tokenizer
|
||||||
|
framework: vllm
|
||||||
|
storage: gpfs
|
||||||
|
|
||||||
|
max_model_len: 4096
|
||||||
|
sut_config:
|
||||||
|
gpu_num: 1
|
||||||
|
values:
|
||||||
|
command: ['vllm', 'serve', '/model', '--port', '20644', '--served-model-name', 'llm', '--max-model-len', '4096', '--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 == "ppu_zw_810e":
|
||||||
|
return f"""gpu_type: ppu_zw_810e
|
||||||
|
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
|
||||||
|
modelhub_options:
|
||||||
|
srcRelativePath: leaderboard/modelHubXC/{model_id}
|
||||||
|
mountPoint: /model
|
||||||
|
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
|
||||||
|
ref_config:
|
||||||
|
values:
|
||||||
|
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'
|
||||||
|
"""
|
||||||
|
|
||||||
|
elif gpu_type == "Iluvatar_bi-150":
|
||||||
|
return f"""docker_image: harbor-contest.4pd.io/luopingyi/enginex-iluvatar-bi150/vllm:0.8.3
|
||||||
|
nv_docker_image: harbor.4pd.io/dooke/vllm/vllm/vllm-openai:v0.11.0
|
||||||
|
framework: vllm
|
||||||
|
api: completion
|
||||||
|
temperature: 0.7
|
||||||
|
repetition_penalty: 1.2
|
||||||
|
top_p: 0.9
|
||||||
|
|
||||||
|
max_model_len: 4096
|
||||||
|
max_tokens: 1024
|
||||||
|
sut_config:
|
||||||
|
gpu_num: 1
|
||||||
|
values:
|
||||||
|
command: ['vllm', 'serve', '/model', '--port', '80', '--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:
|
else:
|
||||||
raise ValueError(f"未知的 GPU_TYPE: {gpu_type}")
|
raise ValueError(f"未知的 GPU_TYPE: {gpu_type}")
|
||||||
|
|
||||||
@@ -812,7 +355,14 @@ ref_config:
|
|||||||
# ══════════════════════════════════════════════════════════
|
# ══════════════════════════════════════════════════════════
|
||||||
# 业务逻辑
|
# 业务逻辑
|
||||||
# ══════════════════════════════════════════════════════════
|
# ══════════════════════════════════════════════════════════
|
||||||
def _submit_task(token: str, gpu_type: str, model_id: str) -> Tuple[bool, str]:
|
# 账号"等待中/运行中"任务数已达上限时平台返回的业务错误信息（子串匹配）；
|
||||||
|
# 命中这个的模型不算永久失败，会在额度腾出空位后自动重试，不会被记作 failed
|
||||||
|
QUOTA_FULL_MSG = "当前等待中或运行中的异步模型验证任务数量已达上限"
|
||||||
|
# 额度耗尽后，隔多久自动重试一次剩余（因额度问题未提交成功）的模型
|
||||||
|
RETRY_INTERVAL_SECONDS = 30 * 60 # 30 分钟
|
||||||
|
|
||||||
|
|
||||||
|
def _submit_task(token: str, gpu_type: str, model_id: str) -> Tuple[bool, str, str]:
|
||||||
headers = {
|
headers = {
|
||||||
"Content-Type": "application/json",
|
"Content-Type": "application/json",
|
||||||
"Authorization": f"Bearer {token}",
|
"Authorization": f"Bearer {token}",
|
||||||
@@ -845,13 +395,14 @@ def _submit_task(token: str, gpu_type: str, model_id: str) -> Tuple[bool, str]:
|
|||||||
if result.get("code") == 0:
|
if result.get("code") == 0:
|
||||||
task_id = result.get("data", {}).get("id", "")
|
task_id = result.get("data", {}).get("id", "")
|
||||||
print(f"[worker] OK {model_id} (GPU={gpu_type}) 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
|
return True, task_id, ""
|
||||||
else:
|
else:
|
||||||
print(f"[worker] FAIL {model_id} (GPU={gpu_type}): {result.get('message')}", flush=True)
|
message = result.get("message") or ""
|
||||||
return False, ""
|
print(f"[worker] FAIL {model_id} (GPU={gpu_type}): {message}", flush=True)
|
||||||
|
return False, "", message
|
||||||
except Exception as e:
|
except Exception as e:
|
||||||
print(f"[worker] ERROR {model_id} (GPU={gpu_type}): {e}", flush=True)
|
print(f"[worker] ERROR {model_id} (GPU={gpu_type}): {e}", flush=True)
|
||||||
return False, ""
|
return False, "", str(e)
|
||||||
|
|
||||||
|
|
||||||
def _run_worker():
|
def _run_worker():
|
||||||
@@ -862,35 +413,69 @@ def _run_worker():
|
|||||||
token = AUTH_TOKEN
|
token = AUTH_TOKEN
|
||||||
print("[worker] 使用预设 Token，跳过登录", flush=True)
|
print("[worker] 使用预设 Token，跳过登录", flush=True)
|
||||||
|
|
||||||
for gpu_type, model_list in GPU_JOBS:
|
# 待提交队列：保持 GPU_JOBS 里原有的 (gpu_type, model_id) 顺序
|
||||||
if _shutdown.is_set():
|
pending: List[Tuple[str, str]] = [
|
||||||
break
|
(gpu_type, model_id)
|
||||||
print(f"\n{'='*60}\n🚀 开始处理 GPU={gpu_type}，共 {len(model_list)} 个模型\n{'='*60}", flush=True)
|
for gpu_type, model_list in GPU_JOBS
|
||||||
|
for model_id in model_list
|
||||||
|
]
|
||||||
|
|
||||||
for model_id in model_list:
|
round_num = 0
|
||||||
|
while pending and not _shutdown.is_set():
|
||||||
|
round_num += 1
|
||||||
|
_state["round"] = round_num
|
||||||
|
_state["phase"] = "submitting"
|
||||||
|
_state["next_retry_at"] = None
|
||||||
|
print(
|
||||||
|
f"\n{'='*60}\n🚀 第 {round_num} 轮，待提交 {len(pending)} 个模型\n{'='*60}",
|
||||||
|
flush=True,
|
||||||
|
)
|
||||||
|
|
||||||
|
quota_blocked: List[Tuple[str, str]] = []
|
||||||
|
for gpu_type, model_id in pending:
|
||||||
if _shutdown.is_set():
|
if _shutdown.is_set():
|
||||||
break
|
break
|
||||||
ok, task_id = _submit_task(token, gpu_type, model_id)
|
ok, task_id, message = _submit_task(token, gpu_type, model_id)
|
||||||
if ok:
|
if ok:
|
||||||
_state["submitted"] += 1
|
_state["submitted"] += 1
|
||||||
_state["per_gpu"][gpu_type] += 1
|
_state["per_gpu"][gpu_type] += 1
|
||||||
successful.append((task_id, gpu_type, model_id))
|
successful.append((task_id, gpu_type, model_id))
|
||||||
|
elif QUOTA_FULL_MSG in message:
|
||||||
|
# 账号额度暂时满了，不算永久失败，留到下一轮重试
|
||||||
|
quota_blocked.append((gpu_type, model_id))
|
||||||
else:
|
else:
|
||||||
|
# 非额度原因失败（如重复提交等），不再重试
|
||||||
_state["failed"] += 1
|
_state["failed"] += 1
|
||||||
|
|
||||||
# 写入结果文件
|
pending = quota_blocked
|
||||||
try:
|
_state["quota_blocked_remaining"] = len(pending)
|
||||||
with open("submitted_validation_tasks.txt", "w", encoding="utf-8") as f:
|
|
||||||
for tid, gpu, mid in successful:
|
# 每轮结束都把已成功的结果落盘一次，避免中途重启丢失记录
|
||||||
f.write(f"{tid}\t{gpu}\t{mid}\n")
|
try:
|
||||||
except Exception:
|
with open("submitted_validation_tasks.txt", "w", encoding="utf-8") as f:
|
||||||
pass
|
for tid, gpu, mid in successful:
|
||||||
|
f.write(f"{tid}\t{gpu}\t{mid}\n")
|
||||||
|
except Exception:
|
||||||
|
pass
|
||||||
|
|
||||||
|
if pending and not _shutdown.is_set():
|
||||||
|
next_retry = datetime.utcnow().timestamp() + RETRY_INTERVAL_SECONDS
|
||||||
|
_state["next_retry_at"] = datetime.utcfromtimestamp(next_retry).isoformat()
|
||||||
|
_state["phase"] = "waiting_retry"
|
||||||
|
print(
|
||||||
|
f"[worker] 第 {round_num} 轮结束：{len(pending)} 个模型因账号额度上限暂未提交，"
|
||||||
|
f"{RETRY_INTERVAL_SECONDS // 60} 分钟后自动重试（不部署新策略，本进程内循环）...",
|
||||||
|
flush=True,
|
||||||
|
)
|
||||||
|
_shutdown.wait(RETRY_INTERVAL_SECONDS)
|
||||||
|
|
||||||
_state["finished_at"] = datetime.utcnow().isoformat()
|
_state["finished_at"] = datetime.utcnow().isoformat()
|
||||||
_state["phase"] = "done"
|
_state["phase"] = "done"
|
||||||
|
_state["quota_blocked_remaining"] = len(pending)
|
||||||
print(
|
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']}",
|
f"total={_state['total']} per_gpu={_state['per_gpu']} "
|
||||||
|
f"仍因额度未提交(如遇shutdown中断)={len(pending)}",
|
||||||
flush=True,
|
flush=True,
|
||||||
)
|
)
|
||||||
# 提交完成后继续保持进程存活，等待平台停止
|
# 提交完成后继续保持进程存活，等待平台停止
|
||||||
|
|||||||
Reference in New Issue
Block a user