How to use forks of the package

[dharux]

I have a fork of the package with some personalized changes in RLCore. I tried to add the package to my Julia environment using the github url for the fork with ] add https://github.com/<username>/ReinforcementLearning.jl. This adds the code of the forked package but also the official versions of the subpackages of Core, Base, Environments and Farm. And when I load it with using ReinforcementLearning, it loads the official version and not the forked version. A way around it is to clone the forked repo locally and use develop. However, this is not very portable as there are file paths in the Manifest.toml and the environment cannot be easily recreated on different machines, which is important for collaboration. Is there an easier way?

[jeremiahpslewis]

Hello,

One approach is to overload specific functions rather than forking the entire project. For example, if you have a modified run loop, you can define a run function with a type signature for your environments or policies, which avoids the need to fork the whole codebase.

Alternatively, if you prefer to stick with the forking approach, I recommend adding the forked sub-packages to your environment.

Best regards.

[dharux]

I was trying to make it work with the forked ReinforcementLearning.jl since there is still the functionality of syncing with the main repo. But seems like the only way is to create a new repo with one of the subpackages and then add that to the environment. Thanks!