Thanks a lot for the great series 🔥! For anyone watching this rn, after a little bit of trial and error I found (epsilon_init: 0.1, decay: 0.99995,min: 0.00001) works great. Since this game runs at high fps , even e = 0.1 can lead to a click 3 times per sec.
when will we know it finish training?Cause my training stay till at 115.9 rewards and when i make a new terminal to run the game,bird can't get even 1 points.
This is a great series. Thank you! Are you planning to post any videos around the Double DQN, Dueling DQN using Pytorch?
I want to summarize the Samsum dataset using deep q-learning. Also train the pegasus model with deep Q learning. Is that possible? What method should I implement to get more accuracy? How to train a pegasus model with deep q-learning.Could you please help me?
@johnnycode