Jan Bruedigam
Jan Bruedigam
Ah yes, didn't think of the renormalization; `testfoo3` works. I'll leave this open for now in case you want to implement a custom auto-diff rule, but feel free to close.
I don't know about package consistency, but I'm ok with not having special `UnitQuaternion` ForwardDiff rules. Just some thoughts from a user perspective: It could be irritating that a `UnitQuaternion`...
Maybe a naive question, haven't looked at the tests in detail: Why are we doing finit-diffing and not auto-diffing? Can finite-differences achieve the desired numerical accuracy?
Parsing of damping is included now
Does not exist yet, would be a nice feature.
The cartpole examples were updated. There is a minimal-coordinate cartpole example with LQR now, and an example interfacing ReinforcementLearning.jl with DQN. The RL example might be broken though, since ReinforcementLearningL.jl...
The examples were removed, but the mechanism still exists for people to build their own controlled version.
The halfcheetah example has been removed, but feel free to reopen in case the stability issues still exist.
Example was removed in the latest version. But there is a sampling-based learning example with the quadruped.
We should probably rethink the interface to ReinforcementLearning.jl once their updates are done (https://github.com/JuliaReinforcementLearning/ReinforcementLearning.jl/issues/614)