Open lisa-sousa opened 3 years ago
I haven't been able to reproduce this behaviour locally. Could you provide a minimal reproducible example of this? Something like what's described here. Information on what versions of packages you are using would be helpful too (like the output of sinfo.sinfo(dependencies=True)
).
I'm using anndata to store some facs data. After reading in the files (each file one sample) and storing the values in anndata format, I concatenate all the samples with
ann.AnnData.concatenate(*anndata_sample_training)
. However, after concatenation, there are some -inf values introduced to the X data matrix. I checked the data before and after concatenation and before I have no -inf values but after concatenation I suddenly have around 30 -inf values in my dataset (values stored in X).