Enter each folder and run based on its own following instruction.

original code of rapid

change model struture, run 'python main.py --env MiniGrid-KeyCorridorS3R2-v0 --sl_until 1200000 --model_type=mlp', model_type can be "mlp", "cnn1d", "cnn2d"

change sample method, run 'python main.py --env MiniGrid-KeyCorridorS3R2-v0 --sl_until 1200000 --choice_type=original' choice_type can be "original", "softmax" and "epsilon-greedy"

give larger weight to beneficial states (key, box and door), run 'python main.py --env MiniGrid-KeyCorridorS3R2-v0 --sl_until 1200000 --local_score_type=original' local_score_type can be "original" and "new"

anneal weight of local score and global store with different start point based on setting ratio, run 'python main.py --env MiniGrid-KeyCorridorS3R2-v0 --sl_until 1200000 --ratio=0.5' ratio can be any float number less than 1, we pick 0.25, 0.5 and 0.75

anneal weight of local score and global store when extrinsic reward is not 0, run just 'python main.py --env MiniGrid-KeyCorridorS3R2-v0 --sl_until 1200000'