Fix bugs in initializing self.cnndm in benchmark.py

[forrestbao]

Critical change:

Below, lines 44 and 54 can cause an error due to CNNDM is undefined.
And there is no need for the global variable CNNDM when self.cnndm presents.

This bug fix uses only self.cnndm which is initialized to None. And when it is none, self.get_cnndm_documents() or self.get_cnndm_references() will populate it with real data.

summac/summac/benchmark.py

Lines 41 to 56 in 9e4f357

	def get_cnndm_document(self, aid):
	global CNNDM
	if self.cnndm is None:
	if CNNDM is None:
	CNNDM = load_dataset("cnn_dailymail", "3.0.0")
	self.cnndm = CNNDM
	self.cnndm_id2article = {}
	for cut in ["test", "validation"]:
	self.cnndm_id2article.update({d["id"]: d["article"] for d in self.cnndm[cut]})
	return self.cnndm_id2article[aid]
	def get_cnndm_reference(self, aid):
	global CNNDM
	if CNNDM is None:
	CNNDM = load_dataset("cnn_dailymail", "3.0.0")
	self.cnndm = CNNDM

Non-critical change:
This fix also exposes cache_dir of Huggingface datasets.load_datasets().

[tusharg7797]

Hi @forrestbao. Thanks for the fix. @tingofurro I had the same error and this fixes the error.