In order to install requirements, follow:
# For users behind the Great Wall only
conda config --add channels https://mirrors.tuna.tsinghua.edu.cn/anaconda/pkgs/free/
conda config --add channels https://mirrors.tuna.tsinghua.edu.cn/anaconda/pkgs/main/
conda config --set show_channel_urls yes
pip install pip -U
pip config set global.index-url https://pypi.tuna.tsinghua.edu.cn/simple
# Create a virtual environment
conda create -n Mega-Agent-2 python=3.6.7 -y
conda activate Mega-Agent-2
# Baselines for Atari preprocessing
git clone https://github.com/openai/baselines.git
cd baselines
pip install -e .
cd ..
# Clone code
mkdir Mega-Agent-2
cd Mega-Agent-2
git clone https://github.com/YuhangSong/Mega-Agent-2.git
cd Mega-Agent-2
# Other requirements
pip install -r requirements.txtconda activate Mega-Agent-2 && CUDA_VISIBLE_DEVICES=1 python main.py --env-name BreakoutNoFrameskip-v4 --algo ppo --use-gae --lr 2.5e-4 --clip-param 0.1 --value-loss-coef 0.5 --num-processes 8 --num-steps 128 --num-mini-batch 4 --use-linear-lr-decay --use-linear-clip-decay --entropy-coef 0.01 --train-with-reward in --intrinsic-reward-type latent --random-noise-frame --epsilon 5.0 --latent-control-intrinsic-reward-type delta_uG__NONE__relu__sum__clip_G__hold_uG --latent-control-discount 0.99 --num-grid 7 --G-skip 1 --aux 23 --vis --vis-interval 1 --log-interval 1 --eval-interval 200 --save-interval 500The code log multiple curves to help analysis the training process, run:
conda activate Mega-Agent-2 && tensorboard --logdir ../results/ --port 6010
and visit http://localhost:6010 for visualization with tensorboard.
If your port is blocked, use natapp to forward a port.
You should see the reward of Mega increases, indicating that it has learned to play the game without access to the extrinsic reward.