All tests passed. For the time being mutiple tabular components will not be offered to the 'load from folder' classes. Therefore, only thing left is adjust docs and test in GPU

Author: jrzaurin

pytorch_widedeep/models/wide_deep.py

@@ -345,7 +345,7 @@ def _forward_deephead(
             if self.is_tabnet:
                 deepside, M_loss = self.deeptabular(X["deeptabular"])
             else:
-                deepside = self._forward_component(
+                deepside = self._forward_component_with_head(
                     X, self.deeptabular, "deeptabular", deepside
                 )