Hi there, I have a question regarding a module that requires

pytorch-gpu

for GPU utilization (

AMULETY

). https://github.com/nf-core/modules/pull/10380 The current container is created with

Seqera wave containers

, which has installed

pytorch-cpu

by default since the environment where the container was created did not have

cuda

available and this is problematic as the tool then does not use GPU acceleration. I’ve tried: • specifying

pytorch-gpu

directly but that also did not work as it needs a cuda installation (https://wave.seqera.io/view/builds/bd-a3dc6ec298210031_3?_gl=1*1pyr52j*_gcl_au*MTUwODM4MzUzNS4xNzY5MjkwNTg0). • using the official

immcantation/amulety

container which has all required dependencies and

pytorch-gpu

, this works but I run out of space on the GitHub runners for the

Singularity

tests only: https://github.com/nf-core/modules/pull/10380 Any suggestions on what would be the best way to proceed here? Can I increase the runner capacity for this module tests, or somehow let them fail?

cc @Dongze He I think you solved the cuda gpu problem previously, right?

I guess you need to install pytorch from pip, which will automatically install cuda runtime.

@Dongze He thank you, this worked!!