5 ms·
AI Learning How to Play Pong in RealTime – PyTorch
- diegoalejogm 9y agoPyTorch code: https://github.com/diegoalejogm/deep-q-learning https://github.com/diegoalejogm/deep-q-learning I'll add final results and performance videos soon!
- paulgb 9y agoNeat! Of note: if I'm reading correctly, this is learning to play the video game pong with no prior knowledge using the screen pixels alone. Pretty cool.
- diegoalejogm 9y agoYeah! I couldn't have described it better than you just did! Feel free to take a look at the code and star the code if you like it. I'll keep upgrading it and adding more Reinforcement Learning variations
- diegoalejogm 9y agoAlso, you can check the FloydExecution in: https://www.floydhub.com/diegoalejogm/projects/atari/29 https://www.floydhub.com/diegoalejogm/projects/atari/29
- diegoalejogm 9y agoAlso, it will be training for several hours more (I approximate 20), so make sure to keep updated. I'll post videos on the current execution performance.
- ghop02 9y agoThe help link doesn't work for me, did the github repo change?
- diegoalejogm 9y agoNo, it hasn't. Here: http://github.com/diegoalejogm/deep-q-learning http://github.com/diegoalejogm/deep-q-learning Try again and let me know if this time it works :)
- ghop02 9y agoAh, I'm new to tensorboard. The help link in tensorboard was broken. But that link works :) very cool!
- diegoalejogm 9y agoNo worries! Here's also a link to the current progress https://www.floydhub.com/diegoalejogm/projects/atari/29 https://www.floydhub.com/diegoalejogm/projects/atari/29 in case you wanna check the machine usage too
- houqp 9y agoJust so you know, machine stats feature is not ramped to all the users yet, so most of them won't be able to see the fancy UI you were seeing. We will push it out to all users late next week ;)
- diegoalejogm 9y agoGood to know :) Thanks for the info, I'll have that in mind for the following posts!
- diegoalejogm 9y agoForgot to mention that it is learning from raw pixels!! No other info is passed to the agent, except the screen frames :)