tiny side cases

scdataloader/data.py

@@ -12,6 +12,7 @@
 import warnings
 
 from anndata import AnnData
+from scipy.sparse import issparse
 
 from scdataloader.utils import get_ancestry_mapping, load_genes
 
@@ -276,8 +277,8 @@ def __init__(
             layer (str): layer of the anndata to use
         """
         self.adataX = (
-            adata.layers[layer].toarray() if layer is not None else adata.X.toarray()
-        )
+            adata.layers[layer] if layer is not None else adata.X)
+        self.adataX = self.adataX.toarray() if issparse(self.adataX) else self.adataX
         self.obs_to_output = adata.obs[obs_to_output]
 
     def __len__(self):

scdataloader/datamodule.py

@@ -304,7 +304,6 @@ def train_dataloader(self, **kwargs):
             self.train_weights,
             self.train_labels,
             num_samples=int(self.n_samples * self.train_oversampling_per_epoch),
-            # replacement=True,
         )
         return DataLoader(self.dataset, sampler=train_sampler, **self.kwargs, **kwargs)
 

scdataloader/utils.py

@@ -242,6 +242,16 @@ def get_all_ancestors(val: str, df: pd.DataFrame):
         return set.union(set(parents), *[get_all_ancestors(val, df) for val in parents])
 
 
+# setting a cache of 200 elements
+# @lru_cache(maxsize=200)
+def get_descendants(val, df):
+    ontos = set(df[df.parents__ontology_id.str.contains(val)].index.tolist())
+    r_onto = set()
+    for onto in ontos:
+        r_onto |= get_descendants(onto, df)
+    return r_onto | ontos
+
+
 def get_ancestry_mapping(all_elem: list, onto_df: pd.DataFrame):
     """
     This function generates a mapping of all elements to their ancestors in the ontology dataframe.