Get model SHA is always failed when pass existing model to HFLM

[wogns3623]

When passing an existing model to HFLM, it always outputs the warning "WARNING Failed to get model SHA for <model_name> ..."

lm-evaluation-harness/lm_eval/models/huggingface.py

Lines 1555 to 1570 in 7ddb2b1

	def get_model_sha(pretrained: str, revision: str) -> str:
	try:
	model_info = HfApi().model_info(repo_id=pretrained, revision=revision)
	return model_info.sha
	except Exception as e:
	eval_logger.debug(
	f"Failed to get model SHA for {pretrained} at revision {revision}. Error: {e}"
	)
	return ""
	model_info = {
	"model_num_parameters": get_model_num_params(self._model),
	"model_dtype": get_model_dtype(self._model),
	"model_revision": self.revision,
	"model_sha": get_model_sha(self.pretrained, self.revision),
	}

Because HfApi().model_info only accepts strings, this causes this issues when self.pretrained is a transformers.PreTrainedModel.
It could be resolved by modifying the variable known as self.pretrained to be of the type PreTrainedModel, or by not fetching the SHA.