fix labels (#6213)

huggingface · Aug 3, 2020 · 0b41867 · 0b41867
1 parent cedc547
commit 0b41867
Showing 1 changed file with 2 additions and 1 deletion.
diff --git a/src/transformers/data/data_collator.py b/src/transformers/data/data_collator.py
@@ -87,7 +87,8 @@ def __call__(self, examples: List[Union[torch.Tensor, Dict[str, torch.Tensor]]])
             return {"input_ids": inputs, "labels": labels}
         else:
             labels = batch.clone().detach()
-            labels[labels == self.tokenizer.pad_token_id] = -100
+            if self.tokenizer.pad_token_id is not None:
+                labels[labels == self.tokenizer.pad_token_id] = -100
             return {"input_ids": batch, "labels": labels}
 
     def _tensorize_batch(self, examples: List[torch.Tensor]) -> torch.Tensor: