Jérémie du Boisberranger

> I would prefer to use black with the default parametrization (88) to stay consistent with other projects' conventions. I think @tomMoral expressed strong disagreement :)

I think we should also add a flake8 check because black does not catch unused imports for instance

Hi @yurivict, sorry for the late answer, I didn't see your issue. I have no idea why it's failing like that, but you'll probably have better chances in the issue...

Hi @lingfeiwang, I'm not sure that I understand how you triggered that. Could you detail a bit more the steps that lead to this broken state ?

Hi @HerculesJack, according to this comment https://github.com/google/jax/issues/743#issuecomment-495031093, the threading mechanism of jax is not one of the ones that threadpoolctl supports. It could be interesting to check if Eigen threadpools...

> conda create -n numpy-tbb numpy tbb tbb should not be needed

Ah ok. It's weird that I don't need it and you do

TF relies on MKL for this operation, right ? If they have their own tool to detect the number of cpu cores and explicitely set it to be the number...

As far as I know, it's not possible in a multi-threaded setting, because we can only set the maximum number of threads of the native C libraries (BLAS, openmp) globally....

I have the same use case. For now I worked around doing: ``` class Suite1(Base, Benchmarks): def setup_cache(self): super().setup_cache() def sub_specific(self): return "Suite1" ``` and same for Suite2, but it's...