search

MolecularAI / QSARtuna

QSARtuna: QSAR model building with the optuna framework
https://molecularai.github.io/QSARtuna/
Apache License 2.0
102 stars 16 forks source link

Does QSARtuna Utilize GPU Resources? #23

Closed sasaju closed 2 months ago

sasaju commented 3 months ago

I noticed that the project uses PyTorch, but during execution, I observed that my GPU was not being utilized. Is additional configuration required to enable GPU acceleration?

CPU: AMD 5600G GPU: GTX1060 6G CUDA Version: CUDA 12.2 Driver Version: 535.183.01

import torch
torch.cuda.is_available()
# True
lewismervin1 commented 3 months ago

Thanks for the interest in QSARtuna and the PR @sasaju. QSARtuna will utilise GPU or the MPS backend (on OSX) when training the ChemProp algorithm. No additional configuration should be necessary. It might be that you checked the GPU utilisation stats when QSARtuna was not actively training the network. Please can you check once the logs are showing that ChemProp trials are being evaluated by Optuna (toward the end of the run if using split_chemprop=True, e.g. see here)

lewismervin1 commented 2 months ago

Hi @sasaju , were you able to get QSARtuna working with GPUs?

sasaju commented 2 months ago

I have run it correctly. I was busy some time ago and forgot to reply. This program is very useful! Thank you for the software and help.