Closed hy-kiera closed 5 months ago
Hi. @amacrutherford @Aidandos Could you please review my PR? Thanks.
Hey! Apologies for the delay on this, @Aidandos had a paper deadline but should get to it over the next few days :smile:
Hi @hy-kiera . Apologies for the delayed response and thanks for pointing out this issue.
Your fix does fix the inner_t
counting, but not the outer_t
counting. I edited your PR to fix both.
The counting was broken both in storm_2p.py
and storm_env.py
.
I added some print statements to the tutorials, showing that inner_t
and outer_t
are incrementing correctly. It also prints the done
logic.
To run the tutorials, simply run
python3 jaxmarl/tutorials/storm_2p_introduction.py
or
python3 jaxmarl/tutorials/storm_introduction.py
The
_step()
was returning a state where some fields were not updated, leading to inaccuracies in the calculation ofdone
. Specifically, whileagent_freezes
,agent_positions
, and other fields were updated,inner_t
remained unchanged in the returnedstate
.To fix this, I've replaced
state_nxt
that all fields are updated, ensuring accuratedone
calculations.Please review, and feel free to provide any additional feedback or suggestions for further refinement.