google colab 使用本地运行时并且不检测本地 gpu

Question

Google colab is connected to local runtime

Local GPU Settings were configured using the following steps:

Edit -> Notebook settings -> Hardware accelerator (GPU)

Runtime -> change runtime type -> Hardware accelerator (GPU)

Connected(Local)

The following piece of code was then used to check if colab running on local runtime was using the local GPU

import torch
torch.cuda.is_available()

The above statement returned False

NOTE: The local computer system uses an NVDIA GeForce 940M GPU

Answer 1

According to the Pytorch tutorial for Vulkan backend ,

Vulkan backend is not included by default in Pytorch.

In order to use PyTorch with Vulkan backend, you need to build it from source with additional settings. Maybe you will be better off doing that locally instead of Colab.

Pytorch already have support for CUDA and other GPU accelerators so unless you intend to use it for inference on your project, it is better to use as it is.

If you do intend to use the Vulkan backend, this github project has a DockerFile and all scripts you need to build Pytorch as well Pytorch-Android .aar files

Github Pytorch Android Example