From e747382e0df0b0aa18b798c3c1948b7d381aaea2 Mon Sep 17 00:00:00 2001 From: p18457032 Date: Sat, 14 Sep 2024 16:15:16 +0800 Subject: [PATCH] Update README.md --- quick_start_clean/readmes/quick_start.md | 2 ++ 1 file changed, 2 insertions(+) diff --git a/quick_start_clean/readmes/quick_start.md b/quick_start_clean/readmes/quick_start.md index 9329aa1..bc1f13d 100644 --- a/quick_start_clean/readmes/quick_start.md +++ b/quick_start_clean/readmes/quick_start.md @@ -113,8 +113,10 @@ pip install tensorboardX 我们提供基于CUDA12.2环境下python3.8、python3.10版本的vllm安装包,相关依赖均已封装,可直接安装后执行推理: [vllm-0.5.0.dev0+cu122-cp38-cp38-linux_x86_64.whl](https://qy-obs-6d58.obs.cn-north-4.myhuaweicloud.com/vllm-0.5.0.dev0%2Bcu122-cp38-cp38-linux_x86_64.whl) [vllm-0.5.0.dev0+cu122-cp310-cp310-linux_x86_64.whl](https://qy-obs-6d58.obs.cn-north-4.myhuaweicloud.com/vllm-0.5.0.dev0%2Bcu122-cp310-cp310-linux_x86_64.whl) + 针对CUDA版本不高的用户,我们提供了兼容低版本CUDA的vllm安装包,但经测试最低支持CUDA11.6,因此,如果您的服务器CUDA版本低于11.6,请先将其升级至该版本以上,以确保兼容性和正常运行: [vllm-0.5.0.dev0+cu116-cp38-cp38-linux_x86_64.whl](https://qy-obs-6d58.obs.cn-north-4.myhuaweicloud.com/vllm-0.5.0.dev0%2Bcu116-cp38-cp38-linux_x86_64.whl) + 同时,我们也提供了vllm源码,位于/quick_start_clean/tools/vllm-0.5.0.dev0.tar ``` ### docker环境