Tensor dimensions mismatch error when using sparse matrix to train expiMap

[maarten-devries]

Hi,
I have installed the latest scArches directly from GitHub (i.e. using pip+git), since I want to use expiMap on an AnnData where adata.X is a sparse matrix.

For reproducibility sake I am running this tutorial: https://scarches.readthedocs.io/en/latest/expimap_surgery_pipeline_basic.html

The only thing I changed before prior to running intr_cvae.train(), is that I made the adata.X a sparse matrix:

import scipy.sparse
adata.X = scipy.sparse.csr_matrix(adata.X)

Expected output:
Model trains successfully when intr_cvae.train() is called

Observed output:
RuntimeError: Tensors must have same number of dimensions: got 3 and 2

[Koncopd]

Thx, i will check what causes the problem.

[Koncopd]

This is fixed on master now.