Yasmin
Yasmin
Ok I see, thanks for the feedback. There was already a solution contributed to nilearn that was never finished so I will pick it up from there, just wanted to...
Yes clear and I agree. Thanks again!
These tests pass on the systems we test. We can indeed raise the tolerances from 0.2 but we wouldn't be able to test for openSUSE system and the architectures listed.
AFAIK setting different architectures than x86 is not supported by Github hosted runners. You can set different architectures using Azure or Travis but I wouldn't set up a new runner...
Thanks for the info @NicolasGensollen. I think it's something to discuss if we want to do this so I'll put it on the agenda for the next meeting.
We decided to not extend support for other architectures for the moment and will try to improve this test instead.
Hi @ggardet, we don't optimize our tests for parallelization so I would encourage you to run with a single worker. If you're working on nilearn code, you can always specify...
I will close this one. We can keep the discussion on the precision error in the linked issue
Thanks for the feedback and great to know that it is useful! Working on a solution
@jyeatman thanks for all your work on this! Will you have time to make the suggested changes from the previous reviews? Then we can go through another round of reviews.