Per offline discussion, this line in particular is pointing at a bigger issue. Namely, the D matrix probably never should've been part of our standard decoder class. Let's split this PR into multiple chunks. One of the future PRs should be to make a decoding task graph that contains D rather than a decoder containing D. Once that is done, that would affect all D parts of this PR.
Per our discussion, the other pieces should probably be moved into individual PRs and merged independently of the aforementioned decoding task graph object.
Originally posted by @bmhowe23 in #765 (comment)
Per offline discussion, this line in particular is pointing at a bigger issue. Namely, the
Dmatrix probably never should've been part of our standarddecoderclass. Let's split this PR into multiple chunks. One of the future PRs should be to make a decoding task graph that containsDrather than a decoder containingD. Once that is done, that would affect allDparts of this PR.Per our discussion, the other pieces should probably be moved into individual PRs and merged independently of the aforementioned decoding task graph object.
Originally posted by @bmhowe23 in #765 (comment)