【AI】本地部署Qwen3.8

本地部署Qwen3.8系统规格预备工作安装huggingface命令下载模型llama部署模型安装CUDA Toolkit测试模型运行模型简单测试蛇吃蛋游戏系统规格NVIDIA GeForce RTX 5070Ti 16G预备工作安装huggingface命令pipinstall-Uhuggingface_hub[cli]下载模型因个人癖好不喜欢下载到默认目录因此指定目录# 下载到当前目录下的 ./Qwen3.8-27B-GGUF 文件夹hf download unsloth/Qwen3.8-27B-GGUFQwen3.8-27B-UD-Q3_K_XL.gguf--local-dir D:\Programs\DevCache\models\Qwen3.8-27B-UD-Q3_K_XLllama部署模型下载https://github.com/ggml-org/llama.cpp/releasesllama-xxx-bin-win-cuda-12.4-x64.zip自行解压配置PATH安装CUDA Toolkit下载地址https://developer.nvidia.com/cuda-12-4-0-download-archive?target_osWindowstarget_archx86_64target_version11target_typeexe_local测试模型llama-cli-mQwen3.8-27B-UD-Q3_K_XL.gguf-pHello, tell me a short story-n256-ngl999-c4096-faon-ctkq4_0-ctvq4_1--temp0.7--top-p0.95--top-k20--min-p0.0--no-mmap运行模型llama-server.exe-mQwen3.8-27B-UD-Q3_K_XL.gguf--host0.0.0.0--port11434-ngl999-c32768-faon-ctkq4_0-ctvq4_0--jinja--temp0.7--top-p0.95--top-k20--min-p0.0--aliasQwen3.8-27B简单测试蛇吃蛋游戏