audio/augmentation.py about danspeech_training

[nc-rafj]

Will look at it later today.

Did you correct to

scheme = self.choose_augmentation_scheme()

?

Because I think that should work. :)

[nc-rafj]

Yea, I think the changes stem from fixing the augmentation-scheme so that all "paths" are equally likely to happen. Unfortunately it also broke the functionality. Good luck. :)

[PabloOrtizCabello]

Thanks for the quick answer! I changed that and I see now in line 44:

augmentor = getattr(self, augmentation)
TypeError: getattr(): attribute name must be string

I will try and follow the white rabbit

[nc-rafj]

Not sure why the getattr is there in the first place. It is definetly wrong.

I think I wrote the original code as follows:

            scheme = self.choose_augmentation_scheme()
            # Apply the chosen augmentations
            for augmentation_function in scheme:
                recording = augmentation_function(recording, self.sampling_rate)

Does that work?

[PabloOrtizCabello]

That seems to work :) I get other errors passed that point, but that's another matter I have to investigate. Those changes in the augmentation script come from a commit from last month, not sure what the purpose was. Thanks!