Closed canergen closed 11 months ago
As discussed GPU training of LogisticRegression. Afterwards, the learned model is loaded in an sklearn classifier to be compatible with the rest of the code. For overclustering, following works (not implemented yet):
flavor = 'rapids' if use_GPU else 'vtraag' method = 'rapids' if use_GPU else 'umap' sc.pp.neighbors(adata, n_neighbors=15, use_rep='X_pca', method=method) sc.tl.louvain(adata, resolution=25., key_added='over_clustering', flavor=flavor)
As discussed GPU training of LogisticRegression. Afterwards, the learned model is loaded in an sklearn classifier to be compatible with the rest of the code. For overclustering, following works (not implemented yet):