gracefully handle empty input (#442)

This commit is contained in:
Wing Lian
2023-08-20 09:18:18 -04:00
committed by GitHub
parent d2e7f27240
commit 9d629d8bff

View File

@@ -85,7 +85,11 @@ class PromptTokenizingStrategy(abc.ABC):
result["input_ids"].append(self.tokenizer.eos_token_id)
result["attention_mask"].append(1)
if result["input_ids"][0] == self.tokenizer.bos_token_id and strip_bos_token:
if (
len(result["input_ids"]) > 0
and result["input_ids"][0] == self.tokenizer.bos_token_id
and strip_bos_token
):
result["input_ids"] = result["input_ids"][1:]
result["attention_mask"] = result["attention_mask"][1:]