issue with sizefactors when using sparse matrices

When running the unittest for graphs and checking wether size_factors is None here:

batchglm/batchglm/train/tf/base_glm_all/estimator.py

Line 143 in 27b648e

if input_data.size_factors is not None:

it seems to be None, when setting sparse=False, but not to be None, when sparse=True.

Also here:

batchglm/batchglm/models/base_glm/utils.py

Line 133 in 612fb63

if size_factors is not None:

sizefactors are None when X is not sparse and not None when X is sparse.
I just can't figure out where sizefactors are set to something not None, when X is sparse.

@SabrinaRichter is this still relevant? The issue is nearly 3 years old. We have decided to retire support for tf1 entirely. I haven't seen this behaviour in tf2 backend nor in numpy backend. Both will be using an entirely restructured input data processing pipeline and this seems to be a problem specific to the input data processing in tf1 backend. I close this for now.