[BUG]: Automatic downloads not working for MPI3D and dSprites

nmichlo / disent

🧶 Modular VAE disentanglement framework for python built with PyTorch Lightning ▸ Including metrics and datasets ▸ With strongly supervised, weakly supervised and unsupervised methods ▸ Easily configured and run with Hydra config ▸ Inspired by disentanglement_lib

https://disent.michlo.dev

MIT License

124 stars 18 forks source link

[BUG]: Automatic downloads not working for MPI3D and dSprites #31

Closed JohannesAck closed 2 years ago

JohannesAck commented 2 years ago

Describe the bug The downloads for MPI3D and dSprites do not work automatically

To Reproduce

from disent.dataset.data import Mpi3dData
data = Mpi3dData(in_memory=True)

Leads to

FileNotFoundError: [Errno 2] No such file or directory: 'C:\\Users\\[...]\\data\\dataset\\mpi3d_realistic\\mpi3d_realistic.npz'

Expected behaviour It would be nice if the datasets were downloaded automatically, as they are for XYSquares.

Other than that thanks a lot for the library, very helpful!

nmichlo commented 2 years ago

Interesting, they should be downloaded automatically. Did you specify prepare=True?

If the above does not work, are you able to manually download the files into the directory it wants?

I have not extensively tested this the download functionality on windows. Something might be wrong with the path handling. I’ll work on adding some basic tests for this when I get a chance.

nmichlo commented 2 years ago

I noticed the color and shape MPI3D ground-truth factors were wrong due to a data loading error.

If you are still using MPI3D, please use pip install disent==0.6.0, I have fixed this problem with the ground-truth factors and added configurable in-memory and disk data loading too.

Are you still having the data preparation error?

nmichlo commented 2 years ago

I am closing this issue for now. If it is still a problem, please don't hesitate to re-open this issue!