[BUG]: Bug problem for sft_dataset.py

[Desperado-Jia]

🐛 Describe the bug

ColossalAI/applications/Chat/coati/dataset/sft_dataset.py

Lines 97 to 113 in 09fe9dc

	def preprocess(
	sources: Sequence[str],
	targets: Sequence[str],
	tokenizer: transformers.PreTrainedTokenizer,
	max_length: int,
	) -> Dict:
	"""Preprocess the data by tokenizing."""
	examples = [s + t for s, t in zip(sources, targets)]
	examples_tokenized, sources_tokenized = [
	_tokenize_fn(strings, tokenizer, max_length)
	for strings in (examples, sources)
	]
	input_ids = examples_tokenized["input_ids"]
	labels = copy.deepcopy(input_ids)
	for label, source_len in zip(labels, sources_tokenized["input_ids_lens"]):
	label[:source_len] = IGNORE_INDEX
	return dict(input_ids=input_ids, labels=labels)

label[:source_len] = IGNORE_INDEX in preprocess() function may meet error when left padding.

Environment

No response

[cwher]

FYI, the behavior of OPT tokenizer is a little weird. It adds an eos token to the beginning of the example.