models icon indicating copy to clipboard operation
models copied to clipboard

using 2.0rc with CUDA10.1, install correctly but cannot find GPU

Open 0mao0 opened this issue 5 years ago • 4 comments

Enviorment: win10+CUDA10.1+cuDNN+PaddlePaddle_2.0rc nvidia-smi: NVIDIA-SMI 431.94 Driver Version: 431.94 CUDA Version: 10.1 nvcc --version: Cuda compilation tools, release 10.1, V10.1.105 cudnn version: cudnn-10.1-windows10-x64-v8.0.5.39 python --version: Python 3.6.1 python -m pip --version : pip 20.2.4 installing by "python -m pip install paddlepaddle-gpu==2.0.0rc0.post101 -f https://paddlepaddle.org.cn/whl/stable.html"

using import paddle.fluid as fluid to check, returns: W1129 09:52:00.254734 16436 init.cc:157] Compiled with WITH_GPU, but no GPU found in runtime. C:\Users\X1\AppData\Local\Programs\Python\Python36\lib\site-packages\paddle\fluid\framework.py:295: UserWarning: You are using GPU version Paddle, but your CUDA device is not set properly. CPU device will be used by default. "You are using GPU version Paddle, but your CUDA device is not set properly. CPU device will be used by default."

installed and unistalled for muitlple times, any thoughts?

0mao0 avatar Nov 29 '20 01:11 0mao0

  1. What's your cudnn version
  2. Have you tried the CPU version and can it run successfully. tips: Don't install CPU and GPU version paddle at the same time. Uninstall both firstly and then install only one of them.

MissPenguin avatar Nov 29 '20 09:11 MissPenguin

  1. What's your cudnn version
  2. Have you tried the CPU version and can it run successfully. tips: Don't install CPU and GPU version paddle at the same time. Uninstall both firstly and then install only one of them.

thanks for reply

  1. cudnn version: cudnn-10.1-windows10-x64-v8.0.5.39
  2. CPU version works fine on my laptop. I did try uninstall both and reinstall GPU version, same warnings.

0mao0 avatar Nov 29 '20 09:11 0mao0

Did this was solved? i have got same thing while running train.py

abhay-kum avatar Jan 04 '21 08:01 abhay-kum

Paddle has some requirements for cuda and cudnn version. Please refer to https://www.paddlepaddle.org.cn/documentation/docs/en/install/pip/windows-pip_en.html .

If your computer has NVIDIA® GPU, please make sure that the following conditions are met and install the GPU Version of PaddlePaddle:

  • CUDA toolkit 9.0/10.0/10.1/10.2 with cuDNN v7.6.5+
  • CUDA toolkit 11.0 with cuDNN v8.0.4
  • Hardware devices with GPU computing power over 3.0

Because you are using CUDA10.1, please try to use cudnn 7.6.5.

pangyoki avatar Feb 20 '21 06:02 pangyoki