This is pytorch implementation of distributed deep reinforcement learning.
- ape-x
- r2d2 (Recurrent Replay Distributed DQN)(experimental)
In our system, there are two processes, Actor and Learner. In Learner process, thread of the replay memory runs at the same time, and these processes communicate using Redis.
git clone https://github.com/neka-nat/distributed_rl.git
cd distributed_rl
poetry install
Install redis-server.
sudo apt-get install redis-server
Setting Atari. https://github.com/openai/atari-py#roms
The following command is running all actors and learner in localhost. The number of actor's processes is given as an argument.
poetry shell
./run.sh 4
Run r2d2 mode.
./run.sh 4 config/all_r2d2.conf
cd distributed_rl
docker-compose up -d
Create EKS resource.
cd terraform
terraform init
terraform plan
terraform apply