Contextual Bandit with PyTorch instead of TF?

So I saw there’s a TF agent that can do Linear Thompson Sampling but was hoping for if there’s a similar Pytorch implementation of that? If you guys know anything? Planning to use it for Send Time Optimization

1 Like

I want it too. Is there anyone who can help me?

Hey thanks for the suggestion.
We’re working on implementing this in torchrl. Stay tuned!

I find pytorch code for thomson sampling.
and it may implement contextual bandit

But I didn’t understand something that was implemented.
Can i study this if i join your slack?