diff --git a/src/transformers/data/data_collator.py b/src/transformers/data/data_collator.py index 6d8234f9c3fd..6ad0a6ccd210 100644 --- a/src/transformers/data/data_collator.py +++ b/src/transformers/data/data_collator.py @@ -261,7 +261,7 @@ def __call__( batch["input_ids"], special_tokens_mask=special_tokens_mask ) else: - labels = batch["input_ids"] + labels = batch["input_ids"].clone() if self.tokenizer.pad_token_id is not None: labels[labels == self.tokenizer.pad_token_id] = -100 batch["labels"] = labels