Framework for fast prototyping of Population-Based-Training methods. Implemented using Tensorflow 2.1.0 and tf_agents 0.4.0
- Goto https://docs.docker.com/install/linux/docker-ce/ubuntu/#install-using-the-repository
- After installation
# create dockergroup and add user
sudo groupadd docker
# add user to group
sudo usermod -aG docker $USER
# restart (if it says authentication failed, restart computer)
su -s $USER
# test if docker runs without sudo (OPEN NEW TERMINAL)
docker run hello-world# Create Docker image (from inside project directory)
docker build -t pbt:latest .
# Run container (with attached port 6006 for tensorboard)
docker run -p 127.0.0.1:6006:6006 -it --name pbt pbt:latest bash
# Inside container, run tensorboard and python script
tensorboard --logdir pbt/logs/ --port 6006 --host 0.0.0.0 & python3 pbt/main.py