pyctcdecode
pyctcdecode copied to clipboard
How does the code handle the blank token when decoding logits from a pre-trained acoustic model?
In addition, what about the beam size?