make it work with TensorFlow>=2.0 and librosa>=0.7

[lene]

[brunolcarli]

I have tested these changes and it do fixes the project issues for newer versions of TF. I vote this PR to be merged 👍

[triscuitcircuit]

tf.histogram_summary should be changed to tf.summary.histogram within model.py.

librosa.output.write_wav(filename, y, sample_rate) should be replaced with sf.write(filename, y, sample_rate, "PCM_24") with sf as the python soundfile import within "generate.py"

Also requirements.txt should be updated