This is an open source project (formerly named Listen, Attend and Spell - PyTorch Implementation) for end-to-end ASR implemented with Pytorch, the well known deep learning toolkit.
eos (eg. "1") was added by default at the end of the text data generation, which was not required for CTC training. So I thought the input target sequence length should be reduced by 1.
eos (eg. "1") was added by default at the end of the text data generation, which was not required for CTC training. So I thought the input target sequence length should be reduced by 1.