4 ms·
As another commenter said, it’s quite obvious about it. For such a key difference, I’d think they’d have the model run record a log for later inspection. Or the
by jsjohnst 2y ago
As another commenter said, it’s quite obvious about it. For such a key difference, I’d think they’d have the model run record a log for later inspection. Or they’d be watching the game play. Or something similar.
- WJW 2y agoIf the model was trained over (say) a week during there was no full moon, how would the model know what this unexpected message means? It would probably just ignore it and continue playing as normal. Nethack is full of messages that can be safely ignored, so just another one would not be unusual. I don't agree that the creators would be watching the game play either. Usually during such training phases you'd run as many copies of the game as the available hardware can manage. I wouldn't be surprised if they had at least hundreds of runs going in parallel and the researcher is definitely not going to be watching them all. If anything, they are going to bed and train the model overnight as much as possible.
- jsjohnst 2y ago> how would the model know what this unexpected message means? It would probably just ignore it and continue playing as normal. Then that’s a poor model then. Significant anomalies should be flagged for manual review, otherwise you corrupt datasets unintentionally.