We build our Atari implementation on top of minGPT and benchmark our results on the DQN-replay dataset.
Dependencies can be installed with the following command:
conda env create -f conda_env.yml
Create a directory for the dataset and load the dataset using gsutil. Replace [DIRECTORY_NAME] and [GAME_NAME] accordingly (e.g., ./dqn_replay for [DIRECTORY_NAME] and Breakout for [GAME_NAME])
mkdir [DIRECTORY_NAME]
gsutil -m cp -R gs://atari-replay-datasets/dqn/[GAME_NAME] [DIRECTORY_NAME]
Scripts to reproduce our Decision Transformer results can be found in run.sh.
python run_dt_atari.py --seed 123 --block_size 90 --epochs 5 --model_type 'reward_conditioned' --num_steps 500000 --num_buffers 50 --game 'Breakout' --batch_size 128 --data_dir_prefix [DIRECTORY_NAME]