Fix error to load data at the correct position when resuming from a checkpoint

Author: Thai Chau Truong

onmt/trainer.py

@@ -308,6 +308,10 @@ def train(
         for i, (batches, normalization) in enumerate(self._accum_batches(train_iter)):
 
             step = self.optim.training_step
+            # Check if the iterator is at the same step as the optimizer
+            if (i + 1 != step):
+                continue
+
             # UPDATE DROPOUT
             self._maybe_update_dropout(step)