Workaround for huggingface#27758 to avoid ZeroDivisionError (huggingf…

…ace#28756)
damithsenanayake · Mar 7, 2024 · ae2acb7 · ae2acb7
1 parent ab83b95
commit ae2acb7
Showing 1 changed file with 2 additions and 1 deletion.
diff --git a/src/transformers/trainer.py b/src/transformers/trainer.py
@@ -2080,7 +2080,8 @@ def _inner_training_loop(
 
         # add remaining tr_loss
         self._total_loss_scalar += tr_loss.item()
-        train_loss = self._total_loss_scalar / self.state.global_step
+        effective_global_step = max(self.state.global_step, 0.001)  # Avoid ZeroDivisionError
+        train_loss = self._total_loss_scalar / effective_global_step
 
         metrics = speed_metrics(
             "train",