xinference 安装及加载模型+ Cherry Studio
python -V
Python 3.12.10
01_在python虚拟环境pip安装.bat
chcp 65001 cd /d C:\soft\xinference IF NOT EXIST venv_xinference ( python -m venv venv_xinference ) CALL venv_xinference\Scripts\activate call venv_xinference\Scripts\python.exe -m pip install --upgrade pip setuptools wheel pip install xinference[all] PAUSE
02_启动.bat
chcp 65001 set HF_ENDPOINT=https://hf-mirror.com set XINFERENCE_MODEL_SRC=modelscope set XINFERENCE_HOME=C:\soft\xinference cd /d C:\soft\xinference IF NOT EXIST venv_xinference ( python -m venv venv_xinference ) CALL venv_xinference\Scripts\activate xinference-local PAUSE
LLM
qwen2.5-instruct llama.cpp ggufv2 7 q4_k_M UID qwen2.5-instruct-7 推荐 占用内存少 7G
或者
Qwen3-Instruct llama.cpp ggufv2 4 q4_k_M UID qwen2.5-instruct-4
嵌入模型
bge-m3
重排模型
bge-reranker-v2-m3 q4_k_M llama.cpp ggufv2 q3_k_M
Cherry Studio
添加模型 兼容 OpenAI http://localhost:9997
刷新模型,增加以上启动的模型。

浙公网安备 33010602011771号