StopIteration

[mattearnshaw]

any pointers for this? using the two_tier line as provided in the README, have preprocessed my own wavs with the scripts.

Wall clock time spent before training started: 0.01h
Training!
0
Traceback (most recent call last):
  File "models/two_tier/two_tier.py", line 615, in <module>
    mini_batch = tr_feeder.next()
StopIteration

[mattearnshaw]

this is because of assumptions in _2np.py on the size of the input data

np.save('music_train.npy', np_arr[:-2*256])
np.save('music_valid.npy', np_arr[-2*256:-256])
np.save('music_test.npy', np_arr[-256:])

if there are fewer than 512 samples, then you get no training data.

[0x70b1a5]

Bless you for posting the solution to this!

[0x70b1a5]

Also of note, preprocess.py must be run from its location in datasets/music, but _2npy.py must be run from root dir, and the music must be in a folder download/, then you need to move the .npy files around.... I will probably make a PR with these changes and a few others in a couple days.