Work in the improvement of the learning process is key in every ML-Agent project. And we need to spend time improving it, but for sure that we will save this time later in our learning processes. It’s a time well wasted.
Every project, and every machine, are different. We need to find the balance between oru model and the machine responsible to run the learning process. The use of a multiple scenario can save the 95% of time compared with a single scenario environment.
We need to maximize the number of steps that our ml-agent is able to do every second, the more steps we can execute shorter the learning process would be.
The first and more important thing to do is find the correct number os agents that our machine can afford in parallel. Fromk this point we can work with the .yaml file, but we will trate this file in a more advanced course.