Clearer indication for overridden method in generation

[ktangri]

The expectation for the prepare_inputs_for_generation function to be overridden can be made clearer by changing

transformers/src/transformers/generation_utils.py

Lines 369 to 374 in 700cee3

	def prepare_inputs_for_generation(self, input_ids: torch.LongTensor, **kwargs) -> Dict[str, Any]:
	"""
	Implement in subclasses of :class:`~transformers.PreTrainedModel` for custom behavior to prepare inputs in the
	generate method.
	"""
	return {"input_ids": input_ids}

to raise a NotImplementedError that provides the information mentioned in the function's comment.

@patrickvonplaten

[patrickvonplaten]

Also putting this in the "Fix generation docs" task basket

[github-actions]

This issue has been automatically marked as stale because it has not had recent activity. If you think this still needs to be addressed please comment on this thread.

Please note that issues that do not follow the contributing guidelines are likely to be ignored.