Summary: 21 instances, 18 unique

Text	Count
# TODO: Very rare cases where the replacement is '<eos>' should be handled gracefully	1
TODO: ids in skip_ids should be consecutive, we can extend it to more generic version later	1
# TODO make summing of the sample sizes configurable	2
# TODO: speed up the following loop	1
# TODO: better encoder inputs?	1
# TODO: Can we add deteminism without this constraint?	1
# TODO: iterative refinement generator does not support ensemble for now.	1
# TODO: The right place for this offset would be inside	1
# TODO: remove check once pyTorch avoids a copy for this case	3
# TODO  remove this once we update apex with the fix	1
# TODO: implementing length-beam	1
# TODO: support different padding direction on target side	1
# TODO: to formally solve this problem, we need to change fairseq's	1
# TODO: decoding for InsertionTransformer	1
# TODO make bias configurable	1
# TODO: Remove this after refactor of BERTModel	1
# TODO make it configurable whether to use max() or sum() here	1
# TODO make summing of the sample sizes configurable	1