Prevent a reference invalidation when wrapped with sample and checkpoint.

Prevents creating a new estimator when loading a sample-wrapped estimator with CheckpointWrapper (continuation of !82 (merged) which prevented the creation of a new estimator right "below").

This now checks if the estimator is wrapped with SampleWrapper and updates the estimator at that level.

Fixes bob.bio.gmm#30 (closed).

bob/pipelines/wrappers.py

@@ -571,10 +571,19 @@ class CheckpointWrapper(BaseWrapper, TransformerMixin):
         if is_estimator_stateless(self.estimator):
             return self
         with open(self.model_path, "rb") as f:
-            estimator = cloudpickle.load(f)
-            # we don't do self.estimator = estimator, because self.estimator
+            loaded_estimator = cloudpickle.load(f)
+            estimator = self.estimator
+            # For the update, ensure that we have the estimator, not a wrapper
+            while hasattr(estimator, "estimator"):
+                # Update this estimator __dict__ except for the attribute "estimator"
+                for k, v in loaded_estimator.__dict__.items():
+                    if k != "estimator":
+                        estimator.__dict__[k] = v
+                estimator = estimator.estimator
+                loaded_estimator = loaded_estimator.estimator
+            # we don't do self.estimator = loaded_estimator, because self.estimator
             # might be used elsewhere
-            self.estimator.__dict__.update(estimator.__dict__)
+            estimator.__dict__.update(loaded_estimator.__dict__)
             return self
 
     def save_model(self):
@@ -765,7 +774,7 @@ class DaskWrapper(BaseWrapper, TransformerMixin):
                 return self
             else:
                 logger.info(
-                    "Ignoring conversion to dask array (checkpoint detected)"
+                    f"Ignoring conversion to dask array (checkpoint detected at {model_path})"
                 )
 
         def _fit(X, y, **fit_params):