Kyle Daruwalla

If most of the examples use `train!`, and we converge on removing it or downplaying it, then we taught a bunch of users to use a non-preferred API. The for-loop...

I started revamping the site, but I didn't finish before getting pulled away. I'll push what I have as a draft today. I don't think the model zoo or any...

I would suggest getting in touch with @findmyway. I saw you posted in the FastAI port channel on Slack, and this issue was part of ML community effort along the...

Would it make sense to use https://github.com/JuliaPOMDP/RLInterface.jl before registering this? That would make the environments interface with the rest of JuliaPOMDP which has a lot of policies and methods already...

Yup I was [thinking the same thing](https://github.com/JuliaReinforcementLearning/CommonRLInterface.jl/issues/18#issuecomment-647562997). Would be happy to give it a shot. Maybe this weekend?

Again was thinking along the same lines when I was commenting on that issue. RLEnv implements an interface that supersedes CommonRLInterface. I was thinking of doing the later here as...

I think logically, a discrete to continuous mapping would be `{1, ..., n} --> [1.0, n]`. Beyond that, I think it is unique to each environment. For example, in CartPole,...

Been thinking about this recently. Should we establish an experimental `zygote` branch that uses custom adjoints to implement differentiable DiscreteSpaces?

I think the old interface matches https://github.com/FluxML/Functors.jl/pull/1 better, but I'm fine putting everything into a tuple. I will still suggest that the interface should be: ```julia update(o, st, x, (dx1,...

What do we lose by sticking with the interface already on master? Why are we wanting to add additional steps to the design?