Leon Overweel

No problem! We used regular Adam as optimizer. For weight decay we added the following `kernel_constraint` to Larq layers: ```python import tensorflow as tf from larq import utils as lq_utils...

> First, we tried to train the model with regular Adam optimizer with L2 regularization. We didn't add the WeightDecay class you provided and just ran the training with LR=0.001,...

Also wondering if this is still under discussion or a won't-do?

More people reporting this issue on Mastodon here: https://mathstodon.xyz/@ionica/110955602973424328

Hi @ClearlyClaire, now that the 4.2 launch is out the door (congrats!), could this issue get a bump?

Any progress on this? I'm getting the issue too.

Thanks for starting this thread @JackKelly! > On the other hand, I'm a little nervous about using binary cloud masks because clouds are so varied, and thin wispy clouds which...

@danstowell and I have Twitter-met actually! When it's possible again though, we should organize a meetup sometime - Dexter would be happy to host.