Validate adata.X is raw counts or already normalized

[raysinensis]

Maybe throw a warning if not int?
For create_signature_matrix.py mostly, any other functions that would be impacted?

[Eastmanmd]

Added

# Throw warning if adata.X is not raw counts 
warnings.warn(
    "Expression matrix contains non-integer values. "
    "Raw counts are expected. ",
    UserWarning
)

[Eastmanmd]

Validated:

[Eastmanmd]

I believe we also need to do this before extracting embeddings. Geneformer expects raw counts while cell2sentence expects normalized single cell data ? If so, I can modify the embeddings.py script to include this feature.

[raysinensis]

Theoretically yes that's what they want, but if it's just ranked in the end, does it matter?

[raysinensis]

scgpt also expects log normed data, and this potentially does matter