eval

[18339185538]

Hello, why is it taking a long time when I run the eval.py file, my GPU doesn't seem to be working, but my CPU is fully loaded。what shoud I do?thanks!

[qaz670756]

Hello, please make sure that you have correctly installed pytorch with cuda. You could check that by the following steps.
First, in your Python console
import torch
torch.cuda.is_available()
and get 'True' result.
Second, type 'nvcc -V' in your terminal, and you may get results like
Copyright (c) 2005-2021 NVIDIA Corporation
Build on xx
Cuda compilation tools, release 11.5, V11.5.119
Build xx
If your results are similar with mine but still could not work with GPU, please check that the cuda version is compatible with your pytorch, or you have installed cudnn

[18339185538]

首先，谢谢您的回答！ 我运行train.py文件时是正常的，GPU能够满负载工作，当我运行eval.py文件时，速度很慢，通过nvidia-smi查看GPU状态，发现GPU没有工作，我尝试在eval.py文件的dev=torch.device('cuda' if torch.cuda.is_available() else 'cpu')下输入print(dev),返回的是‘cuda’。所以我比较疑惑dev既然选择的是cuda，为什么运行eval.py文件时没有使用GPU。希望您能解惑 On 6/2/2022 10:35，Biyuan ***@***.***> wrote： Hello, please make sure that you have correctly installed pytorch with cuda. You could check that by the following steps. First, in your Python console import torch torch.cuda.is_available() and get 'True' result. Second, type 'nvcc -V' in your terminal, and you may get results like Copyright (c) 2005-2021 NVIDIA Corporation Build on xx Cuda compilation tools, release 11.5, V11.5.119 Build xx If your results are similar with mine but still could not work with GPU, please check that the cuda version is compatible with your pytorch, or you have installed cudnn — Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you authored the thread.Message ID: ***@***.***>

[18339185538]

哦，对了，我是设置了输出可视化结果了。不知道是不是这影响了 On 6/2/2022 10:35，Biyuan ***@***.***> wrote： Hello, please make sure that you have correctly installed pytorch with cuda. You could check that by the following steps. First, in your Python console import torch torch.cuda.is_available() and get 'True' result. Second, type 'nvcc -V' in your terminal, and you may get results like Copyright (c) 2005-2021 NVIDIA Corporation Build on xx Cuda compilation tools, release 11.5, V11.5.119 Build xx If your results are similar with mine but still could not work with GPU, please check that the cuda version is compatible with your pytorch, or you have installed cudnn — Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you authored the thread.Message ID: ***@***.***>