Summary: 21 instances, 20 unique

Text	Count
# TODO add an assert related to sizes here	1
# TODO: Handle task setup for evals	1
TODO: Refactor this class to look like WeightedBacktranslationDataset.	1
# TODO (T40938917): Allow loading of multiple rescoring models	1
# TODO:	1
# TODO (T36875783): load pretrained lm to score	1
# TODO: normalize , sentence avg	1
# TODO: this could be refactored to use lang_pari as key too	1
# TODO: (T41818693) Map translation model vs LM model differences	1
# TODO: Remove when gluster is deprecated (T48002528)	1
# TODO: https://github.com/pytorch/pytorch/issues/10747 makes the	1
# TODO: fix after masked lm work completes	1
# FIXME: handle characters properly	1
# TODO  remove this once we update apex with the fix	2
# TODO (T36875783): instantiate a langauge model	1
# TODO: (lizguo) This class will be merged with upstream later.	1
# TODO: nbest	1
# TODO:this class will be merged with upstream after BeamSearchAndDecodeV1 shipped.	1
# TODO: pass to dual model too	1
# TODO: model ensemble	1