The BipedalWalker environment from OpenAI Gym, solved with deep reinforcement learning: Asynchronous Advantage Actor-Critic (A3C) in TensorFlow. Four workers trained for about 30,000 episodes each, in roughly 21 hours on a single CPU.
ai engineer · istanbul, open to relocation or remote
The BipedalWalker environment from OpenAI Gym, solved with deep reinforcement learning: Asynchronous Advantage Actor-Critic (A3C) in TensorFlow. Four workers trained for about 30,000 episodes each, in roughly 21 hours on a single CPU.