Open Kartik14 opened 4 years ago
https://github.com/parlance/ctcdecode/blob/62f3364a36df4945394fe15360563addc7720741/ctcdecode/src/decoder_utils.cpp#L56
Why are we re-sorting the list of prefixes by their object attribute "score" (which does not include the lm score for the last word) when it has already been sorted by the score which includes the lm-score for the last word here.
https://github.com/parlance/ctcdecode/blob/62f3364a36df4945394fe15360563addc7720741/ctcdecode/src/decoder_utils.cpp#L56
Why are we re-sorting the list of prefixes by their object attribute "score" (which does not include the lm score for the last word) when it has already been sorted by the score which includes the lm-score for the last word here.