atari-game-play

Atari - Solving Games with AI 🤖

I will be using ZenML to build a model that can solve Atari games using reinforcement learning. I will be using the Atari 2600 game environment. I will be using the Deep Q-Learning algorithm to solve the game. I found this Github repo, Building a Powerful DQN in TensorFlow 2.0, to get started with our solution. I will be using OpenAI Gym which is a toolkit that provides a wide variety of simulated environments (Atari games, board games, 2D and 3D physical simulations, and so on), so you can train agents, compare them. I will be using the BreakoutDeterministic-v4 environment from OpenAI Gym.

ZenML is an extensible, open-source MLOps framework to create production-ready machine learning pipelines. Built for data scientists, it has a simple, flexible syntax, is cloud- and tool-agnostic, and has interfaces/abstractions that are catered towards ML workflows.

In the real world, building reinforcement learning applications can be challenging so I will be using ZenML (an MLOps framework) which allows for the deployment of models which can be used across the organization. ZenML is an extensible, open-source MLOps framework to create production-ready machine learning pipelines. Built for data scientists, it has a simple, flexible syntax, is cloud- and tool-agnostic, and has interfaces/abstractions that are catered towards ML workflows. ZenML pipelines execute ML-specific workflows from sourcing data to splitting, preprocessing, training, all the way to the evaluation of results and even serving.

A fully trained agent that plays Atari Breakout, will look like this: ```

🐍 Python Requirements

Before running this project, you have to install some python packages in your environment which you can do by following steps:

git clone https://github.com/zenml-io/zenfiles.git
cd atari-game-play
pip install -r requirements.txt

📓 Diving into the code

We're ready to go now. You can run the code, using the run_pipeline.py script.

python run_pipeline.py

📓 Explanation of Code

Here is a short summary of what each step does:

• dqn/model.py This file consists of all the utility functions and classes which we need for developing our steps. All the classes and functions are explained in detail in their respective docstrings.

• pipelines/training_pipeline This file has the pipeline for training the model, built using the ZenML pipeline module.

• steps/game_wrap.py It is a wrapper for the Gym environment. It will manage the state fed to the DQN.

• steps/build_dqn.py It builds the DQN model in Keras.

• steps/replay_buffer.py It takes care of managing the stored experiences and sampling from them on demand.

• steps/agent.py It will put together the Keras DQN model (including the target network) and the ReplayBuffer. It will take care of things like choosing the action and performing gradient descent.

• steps/get_information_meta.py It returns a set of lists like frame number, rewards, and loss lists.

• steps/train.py It trains your model.