David S. Fischer

Checking locally: git diff --name-status tf2..tf2_compatibility D batchglm/train/tf2/__init__.py D batchglm/train/tf2/base/__init__.py D batchglm/train/tf2/base/estimator.py D batchglm/train/tf2/base/external.py D batchglm/train/tf2/base/model.py D batchglm/train/tf2/base/optim.py D batchglm/train/tf2/base_glm/README.md D batchglm/train/tf2/base_glm/__init__.py D batchglm/train/tf2/base_glm/estimator.py D batchglm/train/tf2/base_glm/external.py D batchglm/train/tf2/base_glm/layers.py D batchglm/train/tf2/base_glm/layers_gradients.py...

waiting for https://github.com/theislab/batchglm/issues/92

Hi @gokceneraslan, thanks for the issue! Which version of batchglm, diffxpy and tensorflow are you using? The newest ones from pypi together with tensorflow 1.x shouldn't have such errors unless...

Ok thanks, could you share: ``` adata.obs.loc[:,["celltype", "buffer"]].drop_duplicates() ``` as is or potentially with changed condition names or via pm if private?

https://www.tensorflow.org/probability/api_docs/python/tfp/optimizer/bfgs_minimize

Size factors should not be trainable, not sure if I understand you correctly. You can supply the size factor as a covariate though if you want to regress it out....

This is done as pre-GLM size factor determination step as far as I understand it: https://rdrr.io/bioc/DESeq2/man/estimateSizeFactors.html. Do you also want this as a distinct step? My first impression of "trainable"...

I would tend to not supporting this functionality. We would have to separately benchmark that this is stable and it would increase run time by one order of magnitude, that...

I think estimating size factors as a parameteric function of fixed effects during regression would be easier alternative to "refitting" size factors, this allows for some trend correction and does...

Will do this!