You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
It also causes the the output of Nanosim to be not reproducible despite using the same value for the --seed arg. I think these two lines (L1505 and L1506) should be dropped. Or maybe they should be replaced with something like?
This would give each thread a new random seed that is deterministically defined by the --seed value which should ensure reproducibility across runs with the same --seed.
The text was updated successfully, but these errors were encountered:
Thanks Baraa!
I agree with you. I have suspicions that this is an issue when I was reading through simulator.py recently. @SaberHQ and I will look into it!
Hi @kmnip , is there a reason we would need to seed repetitively by the number of threads or can we remove these lines as suggested by @baraaorabi to enable reproducibility? Happy to MR if appropriate. Thanks for your consideration!
In line 1505 of the
simulator.py
script:NanoSim/src/simulator.py
Lines 1504 to 1506 in 23911b6
The seed is reset before each process/thread is started. This overrides the setting of the random seed in here:
NanoSim/src/simulator.py
Lines 2180 to 2181 in 23911b6
It also causes the the output of Nanosim to be not reproducible despite using the same value for the
--seed
arg. I think these two lines (L1505 and L1506) should be dropped. Or maybe they should be replaced with something like?This would give each thread a new random seed that is deterministically defined by the
--seed
value which should ensure reproducibility across runs with the same--seed
.The text was updated successfully, but these errors were encountered: