Installing ML-Agents
With the advent of ML-Agents v1.0, the process of installing ML-Agents has been improved significantly. However, there are still a few additional steps beyond installing the package through the Package Manager. We will cover them in detail in this section.
As outlined in the Introducing ML-Agents section, the ML-Agent toolkit consists of three components: the ML-Agents Unity package and three Python packages, called mlagents, mlagents_env, and gym_unity. If you're unfamiliar with Python, don't worry; we'll go step-by-step through how to install the packages here.
Important Note
As gym_unity is a wrapper to interface with OpenAI Gym, which we're not using, we won't be using the gym_unity Python package. For more information on OpenAI Gym, please see https://gym.openai.com.
From the components in the toolkit, we can start to understand how we'll train the Agent. We'll use the ML-Agents package to create a learning Agent...
