MuZero: Mastering Go, chess, shogi and Atari without rules

Re:MuZero: Mastering Go, chess, shogi and Atari without rules by nightcracker on Wednesday December 23, 2020 @07:01PM (#25523736)

Eh, I'd say it's fun if you have a couple thousand TPUs lying around.

If you're just messing around with 1 GPU and a desktop PC you should be happy to get Atari breakout to work.

[ Parent | on HN ]

Re:MuZero: Mastering Go, chess, shogi and Atari without rules by syntaxing on Wednesday December 23, 2020 @07:17PM (#25523848)

Definitely the worse part of RL is that it takes so long to train. But it works surprisingly well on Google Colab or equivalent.

[ Parent | on HN ]

Re:MuZero: Mastering Go, chess, shogi and Atari without rules by confuseshrink on Wednesday December 23, 2020 @09:14PM (#25524475)

The published hyperparameters are usually ridiculously conservative, for the simple games like breakout and pong you can usually converge in far fewer frames than in the papers.

[ Parent | on HN ]

Re:MuZero: Mastering Go, chess, shogi and Atari without rules by nightcracker on Thursday December 24, 2020 @09:21AM (#25527657)

I have tried reproducing the papers, with mixed success. I do not share your sentiment.

[ Parent | on HN ]