array-api
array-api copied to clipboard
data-apis
RFC document, tooling and other content related to the array API standard
[Spec](https://data-apis.org/array-api/draft/extensions/generated/array_api.linalg.vector_norm.html#array_api.linalg.vector_norm): `ord: int | float | ~typing.Literal[inf, -inf] = 2` [Stubs](https://github.com/data-apis/array-api/blob/25e717735b72c6b55080f48af634015d94833838/src/array_api_stubs/_2023_12/linalg.py#L788-L795): `ord: Union[int, float, Literal[inf, -inf]] = 2` [`Literals` docs](https://typing.readthedocs.io/en/latest/spec/literal.html#literal-types) > The following are provisionally disallowed for simplicity. We can...
```python def item(self) -> Scalar: """If array contains exactly one element, retun it as a scalar, else raises ValueError.""" ``` Examples: - [`numpy.ndarray.item`](https://numpy.org/doc/stable/reference/generated/numpy.ndarray.item.html) - [`torch.Tensor.item`](https://pytorch.org/docs/stable/generated/torch.Tensor.item.html) - [`pandas.Series.item`](https://pandas.pydata.org/docs/reference/api/pandas.Series.item.html) - [`pandas.Index.item`](https://pandas.pydata.org/docs/reference/api/pandas.Index.item.html) -...
This PR - resolves https://github.com/data-apis/array-api/issues/808 by adding support for selecting elements at one-dimensional indices along a specified axis. - makes the `axis` kwarg keyword-only, which differs from the OP (see...
This RFC proposes the addition of a new API in the array API specification for taking values from an input array by matching one-dimensional index and data slices. ## Overview...
It seems quite widely adopted: - https://numpy.org/doc/stable/reference/generated/numpy.bincount.html - https://pytorch.org/docs/stable/generated/torch.bincount.html - https://docs.cupy.dev/en/stable/reference/generated/cupy.bincount.html - https://jax.readthedocs.io/en/latest/_autosummary/jax.numpy.bincount.html - https://docs.dask.org/en/stable/generated/dask.array.bincount.html Although I have not checked how compatible they are. JAX in particular requires an extra...
This is for consistency with operators, which allow combining an int with an array that has a floating-point data type. See the discussion at #807. I also mentioned this at...
As discussed in today's consortium meeting. See the discussion at numpy/numpy#24976.
This follows the NumPy behavior, where the result is nan if any operand is nan. I checked PyTorch and it seems to do this as well.
See https://github.com/data-apis/array-api/pull/763#issuecomment-2138355341. "Rank" is an ambiguous term and "dimensions" is clearer. Although other names like "max ndims" would also be OK. This is an unreleased API, and it's new in...