Open ZephyrChenzf opened 4 years ago
@ZephyrChenzf So it's not just me? This does not really look like an RNN-T-based model to me either.
To be fair: The repository does not claim that the model is RNN-T-based but papwerswithcode somehow links this repository (https://paperswithcode.com/paper/transformer-transducer-a-streamable-speech#code)
I have a question. You are using the structure of the transformer, not the structure in the transformer-transducer paper. And the loss function uses ctc instead of rnnt