We expect everyone will have a different setup for training agents. Here you can find two ways to wrap or use the environment as a standard reinforcement learning task. You can use a gym wrapper to wrap the environment for use with your favourite deep learning library. You can also use Unity ml-agents custom training scripts directly.
The example script gymwrapper.py shows a simple way to run an AnimalAI task using stable baselines 3. To run this you will also need to install stable-baselines3 (this version was tested with stable-baselines3==1.2.0).
The example script trainaai.py shows how to run ml-agents training. Documentation for this can be found here.