CODEWITHSUNDEEP
pythonpytorch
Hadid
Hadid

Reputation: 45

cuda is not available on my pytorch, but I can't find anything wrong with the version

for some reason I have to use cuda version10.0 instead of upgrading it The version of DriverAPI is higher than RunTimeAPI but somebody told me thats OK Others who asked the same question at last found their version was not match. emmmm not like me

details here

OS:Windows10 Python 3.7&3.8 both tried

result of 'nvcc -V': nvcc: release 10.0, V10.0.130

nvidia-smi: NVIDIA-SMI 451.67 Driver Version: 451.67 CUDA Version: 11.0

conda list: cudatoolkit 10.0.130 0

import torch print(torch.version.cuda)-----None torch.cuda.is_available()-----False

Upvotes: 0

Views: 4118

Answers (2)

Heapify
Heapify

Reputation: 2891

Make sure your torch, cuda and os versions are compatible from this official Pytorch tool - https://pytorch.org/get-started/locally/

The tool will tell you exactly which pytorch to install along with the command to run

Upvotes: 1

KoKlA
KoKlA

Reputation: 978

How did you install it ? I assume with pip. For pytorch I would recommend manually downloading the wheel from https://download.pytorch.org/whl/torch_stable.html and install it with:

    pip install torch-1.4.0+cu100-cp38-cp38-linux_x86_64.whl

Assuming python 3.8 and linux. If you use something different make sure to select the appropriate version for your OS, Cuda version and python interpreter.

Upvotes: 1

Related Questions