Goals:

Currently all tf and torch tensors do not appear in the Api references section. I think this might be due to the fact that those are not included in the __init__.py's __all__ list.

For instance in the docarray.typing.tensor.__init__.py the tf and torch tensors will be added only via the __get_attr__(). Mkdocstring does not seem to find those.

So I think they have to be added explicitly.
My current solution:
Split tensor section into AudioTensor, ImageTensor, Embedding, VideoTensor and Tensor. For each section I'll add the files explicitly:

# AudioTensor

::: docarray.typing.tensor.audio.audio_ndarray
::: docarray.typing.tensor.audio.abstract_audio_tensor
::: docarray.typing.tensor.audio.audio_tensorflow_tensor
::: docarray.typing.tensor.audio.audio_torch_tensor

This however results in a not so great looking table of contents (on the right):

For each file, the filepath is listed, as well as the class inside of it.

Update:
I just realized that that is the same for some of the other sections, so I guess it is fine like this

• check and update documentation, if required. See guide