search

hungtuchen / pytorch-dqn

Deep Q-Learning Network in pytorch (not actively maintained)
MIT License
386 stars 108 forks source link

train time #1

Closed ShaniGam closed 7 years ago

ShaniGam commented 7 years ago

How much time did the training take you on games such as pong and breakout?

hungtuchen commented 7 years ago

Sorry, I forgot the details. I remembered it takes almost 1m timesteps to see a considerable improvement on cumulative rewards. And it takes 1x hours to see the results close to paper. (But not as good as it, maybe we can see if there is anything to improve)

hungtuchen commented 7 years ago

If there is no further question, this issue would be closed.