Tilman Kamp
Tilman Kamp
@DanBmh Augmentations `volume` `gaps` `reverb` `codec` `resample` and `overlay` are most likely not responsible for this discrepancy as their implementations have not been changed during refactoring. For the others it'd...
My observations so far: - Regarding **batch size**: I got the biggest difference to the old implementation when switching from the former combined `--augmentation_pitch_and_tempo_scaling` to `--augment pitch` plus `--augment tempo`....
How about using Control-\ instead?
https://deepspeech.readthedocs.io/en/v0.7.4/TRAINING.html#augmentation
Are you able to provide a publicly accessible audio file that allows reproducing this result? Is this definitely caused by `transcribe.py`? If that's the case, the issue should be moved...