Stable-baselines: Question about masks for predict method
Are masks (dones) only useful when using recurrent policies ?
question
araffin
>All comments
When using the predict method of an algorithm, yes it should only be used in recurrent policies. It allows the LSTMs to reset their internal values when the environment resets.
hill-a
on 18 Aug 2018
👍2
Was this page helpful?
0 / 5 - 0 ratings
Related issues
ktattan
路
3Comments
shwang
路
3Comments
maystroh
路
3Comments
RyanRizzo96
路
3Comments
saeid93
路
3Comments
Most helpful comment
When using the predict method of an algorithm, yes it should only be used in recurrent policies. It allows the LSTMs to reset their internal values when the environment resets.