ability to control learning rate during training

Added reset_learning_rate() method to Trainer class. Intended to allow training scripts, such as a customized train_prod, to control learning rate at particular epochs or folds.

src/trainer.py

@@ -191,6 +191,29 @@ def load_model(self, advanced=False):
             advanced=advanced,
         )
 
+    def reset_learning_rate(self, new_lr=None):
+        """
+        Utility to allow custom control of learning rate during training, 
+        after loading a model from a checkpoint that would otherwise inherit
+        the learning rate from the previous epoch via the checkpoint.
+
+        Args:
+        ----
+        new_lr : TYPE, optional
+            DESCRIPTION. The default is None.
+
+        Returns
+        ----
+        None.
+
+        """
+        if new_lr is None:
+            new_lr = self.hp["learning_rate"]
+        for param_group in self.optimizer.param_groups:
+            param_group["lr"] = new_lr
+        # Recreate the scheduler with the new learning rate
+        self.configure_scheduler()  
+
     def save_model(self):
         """
         Save the current model to checkpoint_dir.