From 6d2b055ee4c32a7f8e916658d458dedc30416c72 Mon Sep 17 00:00:00 2001 From: zhouyuanxi Date: Sun, 20 Sep 2026 15:34:48 +0800 Subject: [PATCH] submit round-33 filter results: MetaX_c-500(2)/Kunlunxin_p-800(3)/hygon_k100-ai(1)/Cambricon_mlu-370-x8(3)/Biren_166m(2), 11 models total Co-Authored-By: Claude Opus 5 --- main.py | 100 ++++++++++---------------------------------------------- 1 file changed, 18 insertions(+), 82 deletions(-) diff --git a/main.py b/main.py index 275c7dd..6999be5 100644 --- a/main.py +++ b/main.py @@ -58,95 +58,29 @@ ACCOUNTS: List[Tuple[str, str, str]] = [ # 各 GPU 的模型列表（来自 filter_verified_models 脚本的筛选结果） # ══════════════════════════════════════════════════════════ METAX_MODELS = [ - "stabilityai/stable-diffusion-2-1", - "metercai/SimpleSDXL2", - "AI-ModelScope/controlnet-depth-sdxl-1.0", - "AIDC-AI/Ovis1.6-Gemma2-27B", - "AI-ModelScope/t2i-adapter-depth-zoe-sdxl-1.0", - "AI-ModelScope/controlnet-canny-sdxl-1.0-mid", - "AI-ModelScope/controlnet-zoe-depth-sdxl-1.0", - "TeichAI/Qwen3-30B-A3B-Thinking-2507-Gemini-2.5-Flash-Distill", - "marioparreno/text-to-emoji-finetuned", - "hbfreed/Olmo-2-1B-Distilled", - "licyks/flux-model", - "DiffSynth-Studio/FLUX.1-Kontext-dev-lora-ArtAug", - "theprint/ReWiz-Qwen-2.5-14B", - "vanta-research/atom-27b", - "EasierAI/Qwen-2.5-32B", - "y22ma/Qwen3-14B-finetune", - "YouCountry/gemma-3-27b", - "richardyoung/uigen-x-30b-moe", - "UbiquantAI/Fleming-R1-32B", - "cortexso/deepseek-r1-distill-qwen-14b", - "OpenBuddy/openbuddy-yi1.5-34b-v21.6-32k-fp16", - "OpenSeeker/OpenSeeker-v1-30B-SFT", - "Qwen/Qwen1.5-32B", + "MusePublic/stable-diffusion-xl-base", + "pei39/deepseek_r1_distill_qwen_7b_awq_rag", ] KUNLUNXIN_MODELS = [ - "fpadovani/ita-latn-10mb-after-ppt-Dp-100mb-ckpt500_seed3407", - "Qwen/Qwen3.8-27B", - "llm-semantic-router/aegis-embed", + "AI-ModelScope/granite-20b-code-instruct-8k", + "AI-ModelScope/granite-20b-code-base-8k", + "AI-ModelScope/granite-20b-code-instruct", ] BIREN_MODELS = [ - "arvindcr4/iolai-2026-qwen25-14b-awq", - "AIDC-AI/Ovis1.6-Gemma2-27B", - "TeichAI/Qwen3-30B-A3B-Thinking-2507-Gemini-2.5-Flash-Distill", - "marioparreno/text-to-emoji-finetuned", - "hbfreed/Olmo-2-1B-Distilled", - "Qwen/Qwen3.8-27B", - "theprint/ReWiz-Qwen-2.5-14B", - "vanta-research/atom-27b", - "EasierAI/Qwen-2.5-32B", - "yoriis/Gemma-Rand-CPT-IT-0.3", - "YouCountry/gemma-3-27b", - "richardyoung/uigen-x-30b-moe", - "cortexso/gemma2", - "5inq/Joi-Qwen3-14B", - "LLM-Research/Phi-3-medium-128k-instruct", - "UbiquantAI/Fleming-R1-32B", - "cortexso/deepseek-r1-distill-qwen-14b", - "OpenBuddy/openbuddy-yi1.5-34b-v21.6-32k-fp16", + "AI-ModelScope/granite-20b-code-instruct-8k", + "pei39/deepseek_r1_distill_qwen_7b_awq_rag", ] CAMBRICON_MODELS = [ - "arvindcr4/iolai-2026-qwen25-14b-awq", - "AIDC-AI/Ovis1.6-Gemma2-27B", - "AIDC-AI/Ovis2-8B", - "Qwen/Qwen3.8-27B", - "llm-semantic-router/aegis-embed", - "Lamapi/next-12b", - "Kwai-Keye/Keye-VL-1_5-8B", + "AI-ModelScope/granite-20b-code-instruct-8k", + "AI-ModelScope/granite-20b-code-instruct", + "pei39/deepseek_r1_distill_qwen_7b_awq_rag", ] HYGON_MODELS = [ - "jerchenxin/qwen2.5-Math-1.5B-grpo_clip-full-step_0000020", - "Papajams/ratiocine", - "arvindcr4/iolai-2026-qwen25-14b-awq", - "AIDC-AI/Ovis1.6-Gemma2-27B", - "Qwen/Qwen3.8-27B", - "kmseong/safety-warp-Llama-3.2-3b-phase3-per-layer", - "flyingfishinwater/good_and_small_models", - "kmseong/safety-warp-Llama-3.2-3b-phase3-no_rotation-per-layer2", - "kmseong/safety-warp-Llama-3.2-3b-instruct-phase3-perlayer-metamath", - "llm-semantic-router/aegis-embed", - "moonshotai/Kimi-K2.5", - "alibayram/magibu-128k-trained", - "UsefulSensors/moonshine-tiny-vi", - "XCurOS/XCurOS-1.2-8B-VLBF16-Instruct", - "UsefulSensors/moonshine-tiny-ko", - "UsefulSensors/moonshine-tiny-ar", - "UsefulSensors/moonshine-tiny-ja", - "UsefulSensors/moonshine-base-ko", - "UsefulSensors/moonshine-base-ja", - "Lamapi/next-12b", - "Lamapi/next-4b", - "kmseong/Llama-3.2-3B-gsm8k_ft_after-rsn-tuned-freeze_rsn_10", - "UsefulSensors/moonshine-base-ar", - "OX-PIXL/SpatialThinker-3B", - "Kwai-Keye/Keye-VL-1_5-8B", - "zai-org/glm-4v-9b", + "pei39/deepseek_r1_distill_qwen_7b_awq_rag", ] MTHREADS_MODELS = [ @@ -716,13 +650,15 @@ PPU_MODELS = [ "espressovi/BODHI-qwen-3-maze-8b-distil", ] -# 本轮提交第三十轮过滤结果：MetaX_c-500(23) / Kunlunxin_p-800(3) / hygon_k100-ai(26) / -# Cambricon_mlu-370-x8(7) / Biren_166m(18)，共77个；沿用多账号 fallback 轮转。 -# 源 model_ids 有226个重复条目，传导到txt里5个重复，灌入时已按模型ID去重 -#（txt 共82行 → 去重后77个）。 +# 本轮提交第三十三轮过滤结果：MetaX_c-500(2) / Kunlunxin_p-800(3) / hygon_k100-ai(1) / +# Cambricon_mlu-370-x8(3) / Biren_166m(2)，共11个；沿用多账号 fallback 轮转。 +# 产出少是筛选条件本身决定的：候选池5992个里「至少一张卡已验证」有4392个、 +# 各卡「无记录」有78~911个，但两者交集只剩1~3个——在别的卡验证成功过的模型 +# 早就被各卡都跑过了，而没被某卡跑过的基本都是验证失败/验证中/未验证。 +# 只有灌入全新模型才会有一批产出（第29轮灌33496个新模型→筛出38772个）。 # 本轮不提交 Mthreads_s4000（保留既有列表）/ ppu_zw_810e（机制B无白名单权限）/ # Sunrise_pt-200-x1（v1.0.13已完成）/ -# Iluvatar_bi-150（本轮筛选为0；且本仓库 framework=vllm，现有 iluvatar 镜像均为 llamacpp/GGUF，缺 vllm 版镜像） +# Iluvatar_bi-150（本轮筛出1个，但本仓库 framework=vllm，现有 iluvatar 镜像均为 llamacpp/GGUF，缺 vllm 版镜像） GPU_JOBS: List[Tuple[str, List[str]]] = [ ("MetaX_c-500", METAX_MODELS), ("Kunlunxin_p-800", KUNLUNXIN_MODELS),