array-api
array-api copied to clipboard
RFC document, tooling and other content related to the array API standard
Here's a list of questions that we need to think about when driving adoption in both array/tensor libraries, and further downstream. 1. Do we want array libraries to adopt this...
Several people have expressed a strong interest in talking about and working on (auto-)parallelization. Here is an attempt at summarizing this topic. - current status - auto-parallelization and nested parallelism...
This issue is meant to summarize the current status and likely future direction of the NumPy array protocols, and their relevance to the array API standard. **What are these array...
Currently versioning in our docs works building pages for released versions (i.e. just `2021.12` for now) a final time, to be linked to later. So in the [`gh-pages` branch](https://github.com/data-apis/array-api/tree/gh-pages), you...
TensorFlow and PyTorch have `unstack()` functions (Torch calls it `unbind()`) for converting an array into a Python sequence of arrays, unpacked along a dimension: ``` >>> torch.unbind(torch.tensor([[1, 2, 3], >>>...
**PR which depends on #484** In yesterdays meeting @oleksandr-pavlyk bought up that NumPy's `iinfo()`/`finfo()` returns info objects with a `dtype` attribute. ~There seemed to be no objections, but it turns...
We currently have respective dunder methods for [bool](https://data-apis.org/array-api/draft/API_specification/generated/array_api.array.__bool__.html)/[ints](https://data-apis.org/array-api/draft/API_specification/generated/array_api.array.__int__.html)/[floats](https://data-apis.org/array-api/draft/API_specification/generated/array_api.array.__float__.html), so if just for consistency `x.__complex__()` seems nice. Personally I've found the other methods hella useful for `array-api-tests` and Hypothesis, so imagine...
There are a few operations that come up like this periodically (also `rollaxis`). That said, some of these require a doc refresher on how they behave before using them. Still...
Several array implementation currently implement `len`/`__len__` basically as shorthand for `.shape[0]`. It would be handy to have this to simplify adoption in existing codes as well as fit more with...
Per discussion in #433 and yesterdays meeting, this PR makes `finfo()` support complex dtypes by returning information of its real-valued component dtype, e.g. ```python >>> xp.finfo(xp.complex64) finfo_object( bits=32, eps=1.1920928955078125e-07, max=3.4028234663852886e+38,...