Pin dependencies to avoid breaking changes

neural_admixture/model/initializations.py

@@ -74,7 +74,8 @@ def get_decoder_init(cls, X: da.core.Array, K: Union[Iterable, int], path: Union
                     pickle.dump(pca_obj, fb)
         except Exception as e:
             raise e
-        assert pca_obj.n_features_ == X.shape[1], 'Computed PCA and training data do not have same number of SNPs' 
+        pca_n_feat = pca_obj.n_features_ if hasattr(pca_obj, "n_features_") else pca_obj.n_features_in_
+        assert pca_n_feat == X.shape[1], 'Computed PCA and training data do not have same number of SNPs' 
         log.info('Projecting data...')
         X_pca = pca_obj.transform(X).compute()
         log.info('Running KMeans on projected data...')
@@ -132,7 +133,7 @@ def get_decoder_init(cls, X: da.core.Array, K: Union[Iterable, int], path: str,
                     raise FileNotFoundError
             else:
                 raise FileNotFoundError
-        except FileNotFoundError as fnfe:
+        except FileNotFoundError as _:
             log.info(f'{n_components}D PCA object not found. Performing PCA...')
             pca_obj = DaskIncrementalPCA(n_components=n_components, random_state=42, batch_size=batch_size)
             pca_obj.fit(X)
@@ -141,7 +142,8 @@ def get_decoder_init(cls, X: da.core.Array, K: Union[Iterable, int], path: str,
                     pickle.dump(pca_obj, fb)
         except Exception as e:
             raise e
-        assert pca_obj.n_features_ == X.shape[1], 'Computed PCA and training data do not have same number of SNPs'
+        pca_n_feat = pca_obj.n_features_ if hasattr(pca_obj, "n_features_") else pca_obj.n_features_in_
+        assert pca_n_feat == X.shape[1], 'Computed PCA and training data do not have same number of SNPs'
         log.info(f'Projecting data to {n_components} dimensions...')
         X_proj = pca_obj.transform(X).compute()
         if not isinstance(K, Iterable):

neural_admixture/src/snp_reader.py

@@ -134,6 +134,8 @@ def _impute(G: da.core.Array, method: Literal["zero", "mean"]="mean") -> da.core
         Returns:
             da.core.Array: imputed genotype array
         """
+        na_cols = da.all(da.isnan(G), axis=0).compute()
+        G = G[:, ~na_cols]
         if da.isnan(G).any().compute():
             log.warning(f"Data contains missing values. Will perform {method}-imputation.")
             if method == "zero":

setup.cfg

@@ -22,11 +22,11 @@ install_requires =
     configargparse>=1.5.3
     Cython>=0.29.30
     codetiming>=1.3.0
-    dask>=2022.5.0
+    dask>=2022.5.0,<2024.2
     dask-ml>=2022.5.27
     h5py>=3.1.0
     matplotlib>=3.3.4
-    numpy>=1.21.0
+    numpy>=1.21.0,<2.0.0
     pandas>=1.2.4
     pandas_plink>=2.2.9
     py_pcha>=0.1.3