Merge pull request #269 from RaulPPelaez/simplify_data

Fix inifinite recursion error on OSX

torchmdnet/data.py

@@ -25,26 +25,21 @@ def __init__(self, dataset, dtype=torch.float64):
         super(FloatCastDatasetWrapper, self).__init__(
             dataset.root, dataset.transform, dataset.pre_transform, dataset.pre_filter
         )
-        self.dataset = dataset
-        self.dtype = dtype
+        self._dataset = dataset
+        self._dtype = dtype
 
     def len(self):
-        return len(self.dataset)
+        return len(self._dataset)
 
     def get(self, idx):
-        data = self.dataset.get(idx)
+        data = self._dataset.get(idx)
         for key, value in data:
             if torch.is_tensor(value) and torch.is_floating_point(value):
-                setattr(data, key, value.to(self.dtype))
+                setattr(data, key, value.to(self._dtype))
         return data
 
-    def __getattr__(self, name):
-        # Check if the attribute exists in the underlying dataset
-        if hasattr(self.dataset, name):
-            return getattr(self.dataset, name)
-        raise AttributeError(
-            f"'{type(self).__name__}' and its underlying dataset have no attribute '{name}'"
-        )
+    def __getattr__(self, __name):
+        return getattr(self.__dict__["_dataset"], __name)
 
 
 class DataModule(LightningDataModule):