TENSORFLOW MODEL TRAINING
Doruk-Dilmen opened this issue · 1 comments
Doruk-Dilmen commented
aniktash commented
Generally speaking, our Tensorflow branch is outdated and no longer updated or maintained. We are actively supporting Pytorch. So it is highly recommended to move to Pytorch.
In your case, the dataset creation did not go through in the 2nd image. You should see all 35 labels processed. The problem is the librosa update does not like the librosa.effects.time_stretch to have the rate as a positional argument.
you can update ai8x-training/TensorFlow/datasets/kws20.py, line 294 as:
audio2 = librosa.effects.time_stretch(audio, rate=rate)