icefall
icefall copied to clipboard
Apply latency penalty on lstm
trafficstars
This PR aims to apply latency penalty on streaming ScaledLSTM model, to decrease the symbol delay. Gradient filter is applied inside the lstm module to prevent training instability.
Related PR:
- latency penalty https://github.com/k2-fsa/k2/pull/976 and https://github.com/k2-fsa/icefall/pull/496
- gradient filter https://github.com/k2-fsa/icefall/pull/564