maltium
maltium
Loading a nnet3 model using *NnetLatticeFasterRecognizer.from_files(...)* fails with the error `Expected token "", got instead "".` Issue seems to be related to the fact that the pykaldi specific kaldi fork...
Would it be possible to serialize the state of a Stream and then recreate the stream from the serialized data so that the recognition can be continued using a similarly...
## Bug description Calling `region.remove()` from the `region-created` handler leaves the region DOM behind. Deferring the call using `setTimeout(() => region.remove(), 0)` works around the issue. This is a useful...