如何更改CUDA_CACHE_MAXSIZE NVIDIA [英] How to change the CUDA_CACHE_MAXSIZE NVIDIA

问题描述

您好，我刚刚购买了新的RTX 3080 GPU.我想将Cuda用于我的Tensorflow模型.但是，当我开始训练模型时，需要花费一个多小时来开始训练模型.所以我去了tensorflow网站，并说要增加CUDA_CACHE_MAXSIZE，但我不知道那是哪里.有人可以帮我找到它并扩大尺寸吗?我的GPU是10GB.谢谢！

Hello I just bought the new RTX 3080 GPU. And I want to use the Cuda for my tensorflow models. But when I start to train the model it takes more than an hour to start training it. So I went to the tensorflow website and there it said to increase the CUDA_CACHE_MAXSIZE but I dont know where that is. Can someone help me find it and make the size bigger? My GPU is 10GB. Thanks!

推荐答案

通过组合 Robert Crovella 在评论部分中提供的解决方案，在答案中证明解决方案.

Proving solution in the answer by combing the solution provided by Robert Crovella in the comment section.

CUDA_CACHE_MAXSIZE 是编译类别下的一个环境变量.默认情况下为256 MB，最大为4 GB，您可以使用以下路径在Windows上设置此环境变量.

CUDA_​CACHE_​MAXSIZE is a environment variable which comes under Compilation category. By default it will be 256 MB and Maximum is 4 GB, you can set this environment variable on your windows using below path.

Control Panel ->System and Security->System->Advanced System settings 

您可以按照以下说明进行操作( 1 & 2 )文档有关更多详细信息.

You can follow these(1 & 2)documentation for more details.

这篇关于如何更改CUDA_CACHE_MAXSIZE NVIDIA的文章就介绍到这了，希望我们推荐的答案对大家有所帮助，也希望大家多多支持IT屋！