Fix some issue to make batch inference easy for predictor and decoder.

TeaPoly

2022-12-02 1b9ac4f7a2eb600e5b769e7294cb4a8f4ec31b63

1
2
3
4
5
6
beam_size: 1
penalty: 0.0
maxlenratio: 0.0
minlenratio: 0.0
ctc_weight: 0.0
lm_weight: 0.0