hparams is not updated in audio preprocessing

[engiecat]

Even if hparams is changed in preset file, the inner workings of audio.py doesn't follow the preset file.

My experiment had shown that while the changed hparams due to preset was passed to preprocess.py and json_meta.py (or ljspeech.py and so on), it was not passed to audio.py. (e.g. when fmin, fmax are changed)

For now, changing hparams.py is the only way to change those.

[engiecat]

CORRECTION: preset is not applied from the _process_utterance of each preprocessing code.

[engiecat]

Developed a quick fix by creating a function in audio.py

def set_hparams(param):
    global hparams
    hparams=param

and calling audio.set_hparams(hparams_that_is_changed_by_parse_json) for _process_utterance it is possible to pass hparams as parameter and override it.