NVlabs / few-shot-vid2vid

Pytorch implementation for few-shot photorealistic video-to-video translation.

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

GPU Driver specification

kartikJ-9 opened this issue · comments

I am using Google Colab (GPU runtime) to replicate the results of the paper. Pytorch version is 1.14, CUDA v10, and Nvidia-smi driver 418.67, Numpy version 1.17. Torchvision 0.5

On executing train.py I am getting the following error :
"CUDA driver version is insufficient for CUDA runtime version"

In google colab: usr/local/cuda is used, which is v10. So, what does CUDA runtime version imply?

It depends on your Pytorch version, Pytorch 1.14 ask CUDA 10.1 at least, install the right version of Pytorch will fix this issue.