search

PWhiddy / PokemonRedExperiments

Playing Pokemon Red with Reinforcement Learning
MIT License
6.88k stars 629 forks source link

Feature Request: Useful Interrupt #139

Closed HughMungis closed 11 months ago

HughMungis commented 11 months ago

So I'd like to stop my run manually but I don't want to lose the data. Is it possible to add an interrupt (Ctrl - L or whatever) so that the run ends, but the data isn't lost?

PWhiddy commented 11 months ago

Hi! All run data is written to files as well as model checkpoints, so if a run terminates you can restart where you left off. Change the file_name variable in the training script to point to the checkpoint you want to start from.