feagen
feagen copied to clipboard
(deprecated) A fast and memory-efficient Python data engineering framework for machine learning.
The required package "bistiming" should be fixed at `0.1.1`. https://github.com/ianlini/feagen/blob/master/feagen/bundling.py#L8
use [numpy's rule](https://docs.scipy.org/doc/numpy/user/basics.broadcasting.html) to broadcast and concat user should specify: 1. concat along which axis 2. reshape rule 3. broadcast rule With these configurations, it should be easy to concat.
so we can know the original data name from the bundle
We now have sparse h5 for single data key, but concatenating sparse h5 is a little more complicated, so I won't implement this recently.