Modern numpy.load does not accept these .npz files about quickdraw-dataset HOT 4 CLOSED

I downloaded a similar file of sheep from the 'Aaron Sheep' collection, and that .npz worked. It looks like the numpy library people have somehow dropped compatibility with some old files.

Where would I find the code for turning the .ndjson files into .npz files? Is that available?

The context is that I would like to do something similar with vector-based emoji files, and the sketch-rnn technology base is really attractive.

Thanks,

Lance Norskog

p.s. I released the 'sox' sound file manipulation CLI tool in 1990, and so am quite familiar with receiving emails about old projects in which I have lost interest. Vaya con dios!

from quickdraw-dataset.

Hi there,

This is how to load the npz files:

Python3:

data = np.load(data_filepath, encoding='latin1', allow_pickle=True)

Python2:

data = np.load(data_filepath, encoding='bytes', allow_pickle=True)

from quickdraw-dataset.

Hi @LanceNorskog

The instructions in README.md has been updated now for loading the old npz files in Python3 (https://github.com/googlecreativelab/quickdraw-dataset)

Also, as you were looking for code for turning the .ndjson files into .npz files, I've also put a link to a notebook from the readme as well (in the sketch-rnn section).

Good luck with the vector-based emoji!

from quickdraw-dataset.

Closing this issue as it seems hardmarus comment should be solving it.

from quickdraw-dataset.