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 looks like that restarting from a checkpoint is not bit-for-bit? I think the issue is that when restarting from a checkpoint the time step is not restored -- it is still using the initial time step defined in Simulation(), not the last_Δt from the saved Clock object in the checkpoint file. See the example below, which is the output of the attached test case.
In the pickup run I changed the onscreen output from every 10 iterations to every 1 iterations to see the time step. Rather than using the previous time step (5.973 s) from the checkpoint, the pickup run is using a time step of 10 s which is the value when defining simulation = Simulation(model, Δt=10, stop_iteration=220). I’m using v0.91.5.
It looks like that restarting from a checkpoint is not bit-for-bit? I think the issue is that when restarting from a checkpoint the time step is not restored -- it is still using the initial time step defined in
Simulation()
, not thelast_Δt
from the savedClock
object in the checkpoint file. See the example below, which is the output of the attached test case.In the pickup run I changed the onscreen output from every 10 iterations to every 1 iterations to see the time step. Rather than using the previous time step (5.973 s) from the checkpoint, the pickup run is using a time step of 10 s which is the value when defining
simulation = Simulation(model, Δt=10, stop_iteration=220)
. I’m using v0.91.5.Initial run
Pickup run
run.jl.zip
@glwagner
The text was updated successfully, but these errors were encountered: