openai / gym

A toolkit for developing and comparing reinforcement learning algorithms.
https://www.gymlibrary.dev
Other
34.73k stars 8.61k forks source link

Montezuma's revenge - suggestion on increasing speed #699

Closed richielo closed 5 years ago

richielo commented 7 years ago

Hi All,

Not sure if this is the right place to ask such question. I am trying to train an agent to play Montezuma's revenge making use of hierarchical reinforcement learning. However, the process is extremely slow, even for one episode, as I am only running it on a built-in gpu of my laptop. Any general suggestions on speeding things up so I can see the results quicker? Like any cloud platform that would be compatible with work like this? I am really new to the area so I am not familiar with the common practices.

Thanks a lot in advance!

babak-badnava commented 7 years ago

Hi, I think you could use amazon aws. amazon have made some of their platforms free for students. https://aws.amazon.com/education/awseducate/

christopherhesse commented 5 years ago

AWS, GCP, and Azure are popular places to rent GPUs from.