paCMAP generate an error ,

[Marwansha]

paCMAP generate an error ,

[davisidarta]

This seems to be an error with PaCMAP. Probably they updated the library and now it takes a different format for initialization. TopOMetry uses spectral initialization for all projections, while PaCMAP defaults to PCA initialization. In most systems, this shouldn't cause really big differences.

I'll update the Projector class to handle whatever updates they might have done and should push a fixed version later today.

For now, you could try running PaCMAP on the latent eigenspace you learned with topometry:

# I suppose this is the eigenspace you're using, from your input:

latent_eigenspace = tg.EigenbasisDict['msDM with bw_adaptive'].results() 
  
embedding = pacmap.PaCMAP(n_components=2, n_neighbors=None, MN_ratio=0.5, FP_ratio=2.0) 

X_transformed = embedding.fit_transform(latent_eigenspace, init="pca") 

Alternatively, use the spectral initialization:

spectral_init = tg.spectral_layout()

X_transformed = embedding.fit_transform(latent_eigenspace, init=spectral_init) 

Let me know if this workaround works for now :)