When I give new data for training over the existing Deep Speech Checkpoint with the pre defined deep speech scorer, Can I have the option to get a newly generated scorer as well along with the trained model? Or should I use the existing scorer file from the pretrained DeepSpeech 0.7 scorer only while training for new dataset along with the newly created model?
The scorer is only used during training for the test step, to compute WER, so you can use whatever one you want, as long as it’s alphabet-compatible with your model.