What is the simplest way to convert an Array column to and from numpy \ pytorch \ tensorflow?

[ophiry]

sklearn, pytorch, tensorflow rely mostly on matrix input and outputs, what would be the simplest (and preferably efficient) way to integrate them with ibis?

is there also support for sparse vectors?

[gforsyth]

Hi @ophiry -- we don't have support for sparse vectors.

For pytorch, we have a to_torch() method on tables and columns for generating a torch-compatible version of the table/column.

For numpy, the fastest route is via pyarrow, so you could call my_column.to_pyarrow().to_numpy().

I haven't used tensorflow in a loooooong time, is there a particular input format it prefers?