Hello,

I have a few comments for the integration in the l2rpn-baselines, i will probably have some later on too ;-)

The first one is the name of your baseline, as of now it is called:

from l2rpn_baselines.Multithreading_agent.ActorCritic_Agent import ActorCritic_Agent

But as you might have noticed:

• python names are "CamelCase" convention (at least for the exported class) so it would be nice to respect that (no underscore, uppercase at begining of words see Camel Case on wikipedia for example). This is cosmetics but for exported class i think it's better :-)
• there is a name mismatch. If you name your Baseline Multithreading_agent then the name of the imported class should also be Multithreading_agent (and not ActorCritic_Agent) [see the example from l2rpn_baselines.MyContrib.MyContrib import MyContrib]
• name of the baseline is itself a bit ambiguous. There tons of MultiThreading agent (A2C for example) so probably you might give a more personalized name, for example Geirina named their agent Geirina, or the Uni Kassel team named it "PandapowerOPFAgent" which is super explicit imho

The package cannot be imported once integrated in the baseline.
My recommendation would be to:

• clone the l2rpn-baselines repository (git clone https://github.com/rte-france/l2rpn-baselines.git)
• cd in the repo (cd l2rpn-baselines)
• set up an empty virtual env (python -m virtualenv venv; source venv/bin/activate)
• install l2rpn-balines from sources (pip install -e .) and don't install anything else
• copy paste your actual repo in l2rpn_baselines (cp -R ../A3C-RL-baseline-agent-for-Grid2Op-environment l2rpn_baselines/Multithreading_agent)
• try to import it from python:

from l2rpn_baselines.Multithreading_agent.ActorCritic_Agent import ActorCritic_Agent
from l2rpn_baselines.Multithreading_agent.ActorCritic_Agent import evaluate

These two should work out of the box

Once its working, copy paste back the content of l2rpn_baselines/A3C-RL-baseline-agent-for-Grid2Op-environment into your development github repository:
cp -R l2rpn_baselines/ActorCritic_Agent/* ../Multithreading_agent.ActorCritic_Agent import ActorCritic_Agent

and then commit and push it back :-)

Hello, I'm very interested in your project. However, during my attempts to train it, I've noticed that the printed alive steps are fluctuating within a range of several tens of steps, without any clear training progress. I'm beginning to consider whether this issue is related to the library versions.

If possible, could you please provide the relevant version dependency file requirements.txt? This would help me to address potential version compatibility issues and gain a better understanding of the challenges that may arise during the training process.