search

Rayhane-mamah / Efficient-VDVAE

Official Pytorch and JAX implementation of "Efficient-VDVAE: Less is more"
https://arxiv.org/abs/2203.13751
MIT License
188 stars 21 forks source link

Workaround for loading hparams without tf 1.x ? #7

Closed turian closed 1 year ago

turian commented 1 year ago

Colab no longer supports tensorflow 1.x

Is there a workaround for loading the hparams cfg?

I tried hparams pypi package but it didn't work.

Also the following didn't work:

# https://github.com/tensorflow/community/issues/148
from tensorboard.plugins.hparams import api as hp
hparams = hp.HParams('.', name="efficient_vdvae")
Rayhane-mamah commented 1 year ago

Hello @turian and thank you for the interest in our work.

As explained in the README, the hparams we use isn't the tensorflow hparams module, but a similar behaving one.

The package isn't hosted on pypi yet and can be installed with (confirmed it works on colab):

pip install --upgrade git+https://github.com/Rayhane-mamah/hparams

Basic usage can be found here if you are writing your own code.

Let me know if this solves your problem.