Jake Vanderplas

True it's easy enough to fix locally by manually editing the problematic files. I'm looking for suggestions of the best way to get this working in my github actions job...

I'm using `sed` as a workaround for now: https://github.com/google/jax/pull/16099/commits/c55b9e7d9fcc3a483b1501e446780bf2563f5802

> @shoyer or @jakevdp, would you be able to comment on why JAX implements a limited form of "array-like"? JAX avoids implicit conversion of Python sequences, because it can hide...

> Silent downcasting is terrible - it may be okay for deep learning, but it typically isn't for general purposes numerical/scientific computing. I think you hit on the key point...

> But I guess JAX does not use this CUDA, but installs its own, correct? With `pip install jax[cuda12]`, it will download its own CUDA; with `pip install jax[cuda12_local]`, it...

Are you still interested in contributing this?

Thanks for raising this... it's strange that JAX returns valid values for invalid inputs. I think emulating scipy here would be reasonable.

Thanks for following up!

Closing, as I think this is working as expected.

I did you one better: I renamed the package :smile: https://github.com/jakevdp/ipywidgets-unsupported