Why is the random seed not valid in atari_ppo.py?

davide97l / rl-policies-attacks-defenses

Adversarial attacks on Deep Reinforcement Learning (RL)

MIT License

76 stars 12 forks source link

Why is the random seed not valid in atari_ppo.py? #8

Closed chenzhenhao876 closed 3 years ago

chenzhenhao876 commented 3 years ago

The results are different every time. Thank you.

davide97l commented 3 years ago

Hi @chenzhenhao876, this framework is currently being developed according to the progress made on the main framework Tianshou. Up to now, only DQN is working, but we are trying to make also A2C and PPO work. Once we will have fixed this problem we will inform you. Tianshou: https://github.com/thu-ml/tianshou

davide97l commented 3 years ago

Now PPO has been implemented and tested on some of the Atari games. You can use the --seed command to fix the seed