Merge pull request #88 from YosefLab/release_hub

Python 3.11 changes.

Author: canergen

popv/__init__.py

@@ -6,7 +6,7 @@
 if sys.version_info[:2] != (3, 10):
     warnings.warn(
         "Pretrained models on huggingface are trained with Python 3.11. "
-        f"Detected Python {sys.version.split()[0]} will not load these models.",
+        "Detected Python {sys.version.split()[:2]} will not load these models.",
         UserWarning,
         stacklevel=3,
     )

popv/algorithms/_svm.py

@@ -76,14 +76,14 @@ def predict(self, adata):
         if adata.uns["_prediction_mode"] == "retrain":
             train_idx = adata.obs["_ref_subsample"]
             train_x = adata[train_idx].layers[self.layer_key] if self.layer_key else adata[train_idx].X
+            train_x = np.array(train_x.todense())
             train_y = adata.obs.loc[train_idx, self.labels_key].cat.codes.to_numpy()
             if settings.cuml:
                 from cuml.svm import LinearSVC
                 from sklearn.multiclass import OneVsRestClassifier
 
                 self.classifier_dict["probability"] = self.return_probabilities
                 clf = OneVsRestClassifier(LinearSVC(**self.classifier_dict))
-                train_x = train_x.todense()
                 clf.fit(train_x, train_y)
                 joblib.dump(
                     clf,
@@ -131,7 +131,7 @@ def predict(self, adata):
             for i in range(0, adata.n_obs, shard_size):
                 tmp_x = test_x[i : i + shard_size]
                 names_x = adata.obs_names[i : i + shard_size]
-                tmp_x = tmp_x.todense()
+                tmp_x = np.array(tmp_x.todense())
                 result_df.loc[names_x, self.result_key] = adata.uns["label_categories"][clf.predict(tmp_x).astype(int)]
                 if self.return_probabilities:
                     result_df.loc[names_x, f"{self.result_key}_probabilities"] = np.max(

popv/hub/_metadata.py

@@ -2,6 +2,7 @@
 
 import json
 import os
+import sys
 from dataclasses import asdict, dataclass, field
 
 from anndata import AnnData
@@ -194,6 +195,7 @@ def _to_model_card(self) -> ModelCard:
             "genomics",
             "single-cell",
             f"anndata_version:{self.anndata_version}",
+            f"python_version:{'.'.join([str(i) for i in sys.version_info[:3]])}",
             "popV",
         ]
         for t in self.tissues: