Fix option to continue model-based inference (#460)

* move init round

move init round from update to prepare new batch so that the fit method in BOLFIRE can be called to continue optimisation

* update changelog

* move init round

move init round from prepare new batch to iterate

* move init round

move init round back to update but additionally check in infer if inference is continued and a new round needs to be initialised

Author: uremes

CHANGELOG.rst

@@ -1,6 +1,7 @@
 Changelog
 =========
 
+- Fix the option to continue inference in model-based inference
 - Move classifiers
 - Fix readthedocs configuration
 - Update penalty to shrinkage parameter conversion in synthetic likelihood calculation

elfi/methods/inference/parameter_inference.py

@@ -515,6 +515,21 @@ def current_params(self):
         """
         raise NotImplementedError
 
+    def infer(self, *args, **kwargs):
+        """Set the objective and start the iterate loop until the inference is finished.
+
+        Initialise a new data collection round if needed.
+
+        Returns
+        -------
+        result : Sample
+
+        """
+        if self.state['round'] > 0:
+            self._init_round()
+
+        return super().infer(*args, **kwargs)
+
     def _merge_batch(self, batch):
         simulated = batch_to_arr2d(batch, self.feature_names)
         n_sim = self.state['n_sim_round']