Guido Imperiale

> This is why I've suggested keeping `--quick` enabled by default (the CI stuff can be adjusted, but other people might be _very_ surprised if the default behavior changes). Calling...

> AFAIK the problem is that the `scipy` benchmark suite was badly broken, and `asv` changed from always passing CI (pre `0.5`) to failing on broken benchmarks in newer versions....

@kgryte I'm writing a backend-agnostic wrapper around `jax.pure_callback` / `dask.blockwise` / equivalent functions (ndonnx?) that let you apply an arbitrary eager callback to a lazy array: https://github.com/data-apis/array-api-extra/pull/86 This requires me...

xref https://github.com/data-apis/array-api-extra/pull/300

> * ISTM that `clip` is most similar to `where`. In the sense that, alternatively to what I proposed above, it could be more simply stipulated to return `result_type(x, min,...

I ran this test many times per environment between 6 and 17 aug over 3 mass stress test runs. It failed 38 times out of 66 runs on Windows 3.7....

+1; the change makes sense to me. By this same logic, a task should transition to erred as soon as it lands on the scheduler if the workers satisfying its...

> Just a quick note that `take` is the standardized/uniform way of doing indexing with an integer array (xref [gh-416](https://github.com/data-apis/array-api/pull/416) and the issue that PR links to). I couldn't find...

> Using the current spec, is there an array API-compatible way to implement this feature though? Because swapping the last line with > > import array_api_extra as xpx > ......

Yes, but I'm not sure about my current bandwidth. I'd suggest taking a stab at the PR yourself if you care about having it in quickly.