On a related note - I realise there are two PRs working towards slightly different outcomes on the same module: https://github.com/nf-core/modules/pull/10306 https://github.com/nf-core/modules/pull/9793

@Florian Wuennemann and @Dongze He ☝️

Thank you all and sorry for not browsing the repo before I started. We have very similar implementation, the only difference is I made a seqera container that has cuda enabled. @Florian Wuennemann would you mind have look of my PR? I think it is ready to merge. Thanks!

I will check it tomorrow and give feedback! Definitely happy to switch to a seqera container with cuda here. We might want to have two containers though, one for cpu and one for gpu since the gpu one is usually a lot bigger?

For cellpose, I think pytorch is the declared dependency. I only specified python and cellpose when building the image. community.wave.seqera.io/library/python_pip_cellpose:4be22f3ec4963ebe But sure, if you think separating CPU and GPU images is preferred, I can make the change!

I actually don't know right now what our stance is from the nf-core side. Let me check and get back to you 😊 . Thanks for making the update by the way! 👍

P.S. cellpose 4.0.9 is out now 😉

I'll put some time aside today to bump the version and test the GPU integration @Dongze He!

thanks everyone. I also updated cellpose to 4.0.9.

Tested on GPU and works (38s on cpu vs 17s with an H100 for the nf-test), I left review comments. Almost good to go. Would love some input from @mashehu regarding the gpu nf-tests for runners etc and then I think we can accept and merge 👍 . Great job @Dongze He!