Downgrading from CUDA 11.2 to 11.1 or 10.2 on Google Colab (Package not found issue)

user_11076130 0
Viewed 3395

So I've tried almost everything I can think of to downgrade the CUDA version on Google Colab (11.2) since it isn't supported with pytorch/pytorch-geometric. I've followed the advice of multiple other posts on the topic downloading the CUDA version online and wiping the local one using (for example):

!wget https://developer.download.nvidia.com/compute/cuda/repos/ubuntu2004/x86_64/cuda-ubuntu2004.pin
!mv cuda-ubuntu2004.pin /etc/apt/preferences.d/cuda-repository-pin-600
!wget https://developer.download.nvidia.com/compute/cuda/11.1.0/local_installers/cuda-repo-ubuntu2004-11-1-local_11.1.0-455.23.05-1_amd64.deb
!dpkg -i cuda-repo-ubuntu2004-11-1-local_11.1.0-455.23.05-1_amd64.deb
!apt-key add /var/cuda-repo-ubuntu2004-11-1-local/7fa2af80.pub
!apt-get update
!apt-get -y install cuda = 11.1.0
!apt autoremove

But I just get the same output whether I try wiping CUDA version beforehand with other code, changing the path/name, or different CUDA versions, which is:

...
Note, selecting 'libghc-cryptohash-sha1-prof' instead of 'libghc-cryptohash-sha1-prof-0.11.100.1-c80f9'
E: Unable to locate package

Here is information on the system and CUDA:

+-----------------------------------------------------------------------------+
| NVIDIA-SMI 460.32.03    Driver Version: 460.32.03    CUDA Version: 11.2     |
|-------------------------------+----------------------+----------------------+
| GPU  Name        Persistence-M| Bus-Id        Disp.A | Volatile Uncorr. ECC |
| Fan  Temp  Perf  Pwr:Usage/Cap|         Memory-Usage | GPU-Util  Compute M. |
|                               |                      |               MIG M. |
|===============================+======================+======================|
|   0  Tesla T4            Off  | 00000000:00:04.0 Off |                    0 |
| N/A   40C    P8    10W /  70W |      0MiB / 15109MiB |      0%      Default |
|                               |                      |                  N/A |
+-------------------------------+----------------------+----------------------+

No LSB modules are available.
Distributor ID: Ubuntu
Description:    Ubuntu 18.04.5 LTS
Release:    18.04
Codename:   bionic

Any help with this would be greatly appreciated since it would hugely speed up my run time which can only use the CPU right now on my laptop!

torch google-colaboratory python
1 Answers

The steps you followed seem correct. The only additional step is restarting the runtime (Menu->Runtime->Restart Runtime). After that, checking with

!nvcc --version

shows

nvcc: NVIDIA (R) Cuda compiler driver Copyright (c) 2005-2020 NVIDIA Corporation Built on Mon_Oct_12_20:09:46_PDT_2020 Cuda compilation tools, release 11.1, V11.1.105 Build cuda_11.1.TC455_06.29190527_0

However nvidia-smi shows the same old version 11.2, it does not seem to actually affect running PyTorch.

user_5082220
Related