I am trying to do some transcription and wanted to know what a good beam_width is for the pretrained 0.6.1 release?
def initialize_deepspeech(self, model='data/models/deepspeech-0.6.1-models/output_graph.pbmm', scorer='data/models/deepspeech-0.6.1-models/kenlm.scorer', beam_width=0):
self.deepspeech_model = deepspeech.Model(model)
if beam_width:
self.deepspeech_model.setModelBeamWidth(beam_width)
self.deepspeech_model.enableExternalScorer(scorer)