Ivan-267

Updated to the latest experimental multiagent plugin (so should not be merged before), and onnx re-trained using Rllib. ## Continuous actions (included onnx is from this session): Tensorboard stats (smoothing:...

I've modified the onnx inference to handle different action spaces, and to handle models exported with Rllib differently than SB3 exported models by the following process: ``` 1) Check the...

@yaelatletl thanks for your review and this importance notice, I will check the code for this and respond with more detail and/or alterations.

@yaelatletl Thanks for spotting this. This was an oversight on my part. The point was to handle the disposal of disposable resources, but that can also be done with e.g....

In one approach, Godot could signal when the session is over by calling a method that takes care of the disposal. Also I forgot to add about the file loading...

Perhaps you could implement the needed changes on the gdscript side first, since changes on both sides (gdscript, c#) will need to be synchronized. > Using` the Godot loader is...

Superseded by https://github.com/edbeeching/godot_rl_agents_plugin/pull/40. The branch can still be kept for reference since this implementation has an advantage, it sends only action values for discrete actions (not logits), so it could...

Thanks for the review. I don't have a custom environment for this as I was testing it on a modified example environment from the examples repo. Also since I had...

I've made a small test environment which only outputs action values (includes the updated compatible plugin already). godot-rl libray still needs to be installed from the PR branch. https://github.com/Ivan-267/godot_rl_agents_examples/tree/test_environments/test_environments/ActionSpacesTestEnvironment I've...

Please hold off on the review as an important issue was spotted by @yaelatletl during his review that needs to be addressed in my experimental branch of the plugin before...