PyTorch Library comes to the forefront as a Reinforcement Learning (OP) tool due to its efficiency and ease of use. This book is organized as a guide to working with PyTorch, covering a wide range of topics - from the basics (setting up the working environment) to practical tasks (considering OP with specific examples). You will learn how to use the algorithm of "multi-armed bandits" and approximation of functions; learn how to win Atari games using deep Q networks and how to effectively implement the strategy gradient method; You will see how to apply the OP method to the game of blackjack, to environments in the grid world, to optimize advertising on the Internet and to the game Flappy Bird. The publication is intended for artificial intelligence specialists who need help in solving OP problems. To study the material, you need to get acquainted with the concepts of machine learning; experience with the PyTorch library is optional but desirable.