ValueError: Expect BinaryTrack or StandardTrack, but got <class 'pypianoroll.track.Track'>.

[avani-031]

File "/usr/local/lib/python3.6/dist-packages/pypianoroll/outputs.py", line 139, in to_pretty_midi
f"Expect BinaryTrack or StandardTrack, but got {type(track)}."
ValueError: Expect BinaryTrack or StandardTrack, but got <class 'pypianoroll.track.Track'>.

### How do I debug this error?

[salu133445]

Hi, please downgrade pypianoroll to 0.5.3. Thanks.

[avani-031]

Hi sir,
I am getting this error on downgrading.
File "/content/drive/.shortcut-targets-by-id/436/bmusegan/musegan/utils/midi_io.py", line 54, in write_midi
Multitrack.append(multitrack,track)
AttributeError: type object 'Multitrack' has no attribute 'append'

[salu133445]

It's Multitrack.append_track(multitrack,track) not Multitrack.append(multitrack,track). You probably have modified this line when debugging the above bug.

Note - Multitrack.append_track is renamed to Multitrack.append in Pypianoroll 1.0.0. The code in this repository is written using Pypianoroll <1.

[avani-031]

Thank you sir!

[avani-031]

When training starts it gets stuck at 2020-12-22 08:39:14.828929: W tensorflow/core/framework/allocator.cc:108] Allocation of 278003712 exceeds 10% of system memory.

I tried reducing the batch size from 32 to 15, 10 but on doing that it says cannot reshape ____ into array of (___)
Can you please help me with this sir

[salu133445]

This TensorFlow warning is not so meaningful actually. One common cause for hanging is when the training data exceeds the RAM size (see salu133445/musegan#99 and salu133445/musegan#100). Another common cause is when the model is not trained on a GPU. It might take more than 10 minutes per iteration on a CPU, but it generally takes less than 10 seconds on a GPU (see salu133445/musegan#33).

[salu133445]

You might want to run the latest code in the musegan repository, which also supports the BinaryMuseGAN model yet more updated.

[avani-031]

Thank you so much sir!