Skip to content

Commit

Permalink
checkpointing in InterpolatingAdjoint seems to be the issue
Browse files Browse the repository at this point in the history
  • Loading branch information
frankschae committed Jan 10, 2024
1 parent b2da78e commit f078d27
Showing 1 changed file with 1 addition and 2 deletions.
3 changes: 1 addition & 2 deletions src/callback_tracking.jl
Original file line number Diff line number Diff line change
Expand Up @@ -266,8 +266,7 @@ function _setup_reverse_callbacks(cb::Union{ContinuousCallback, DiscreteCallback
# if save_positions[2] = false, then the right limit is not saved. Thus, for
# the QuadratureAdjoint we would need to lift y from the left to the right limit.
# However, one also needs to update dgrad later on.
if (sensealg isa QuadratureAdjoint && !cb.save_positions[2]) # || (sensealg isa InterpolatingAdjoint && ischeckpointing(sensealg))
# lifting for InterpolatingAdjoint is not needed anymore. Callback is already applied.
if (sensealg isa QuadratureAdjoint && !cb.save_positions[2]) || (sensealg isa InterpolatingAdjoint && ischeckpointing(sensealg))

This comment has been minimized.

Copy link
@ChrisRackauckas

ChrisRackauckas Jan 10, 2024

Member

GaussAdjoint also does checkpointing

This comment has been minimized.

Copy link
@frankschae

frankschae Jan 10, 2024

Author Member

yeah very likely, we'll have to add it here as well .. GaussAdjoint is not included in the callback tests yet though. Not sure if @acoh64 ran into other problems.

w(y, y, integrator.p, integrator.t)
end

Expand Down

0 comments on commit f078d27

Please sign in to comment.