Skip to content

Allow more AD backends in optimization based solvers #380

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom
Merged
merged 9 commits into from
Nov 3, 2025
Merged

Allow more AD backends in optimization based solvers #380

merged 9 commits into from
Nov 3, 2025

Conversation

@ErikQQY
Copy link
Member

@ErikQQY ErikQQY commented Oct 22, 2025 •
edited
Loading

Fix #379

The previous implementation can't exploit the sparsity pattern of the loss function due to the value updates in loss function, this PR allows sparse AD in optimmization based internal solvers.

Before this PR:

image

With this PR:

image

@github-actions
Copy link
Contributor

github-actions bot commented Oct 22, 2025 •
edited
Loading

Benchmark Results

Click to check benchmark results
master 1a80f69... master / 1a80f69...
Simple Pendulum/IIP/BoundaryValueDiffEqMIRK.MIRK2() 0.289 ± 0.018 s 0.29 ± 0.018 s 0.993 ± 0.086
Simple Pendulum/IIP/BoundaryValueDiffEqMIRK.MIRK3() 10.3 ± 0.96 ms 10.3 ± 1.5 ms 0.997 ± 0.17
Simple Pendulum/IIP/BoundaryValueDiffEqMIRK.MIRK4() 2.62 ± 0.082 ms 2.59 ± 0.081 ms 1.01 ± 0.045
Simple Pendulum/IIP/BoundaryValueDiffEqMIRK.MIRK5() 7.42 ± 1.3 ms 7.39 ± 1.3 ms 1 ± 0.26
Simple Pendulum/IIP/BoundaryValueDiffEqMIRK.MIRK6() 1.34 ± 0.13 ms 1.35 ± 0.21 ms 0.986 ± 0.18
Simple Pendulum/IIP/MultipleShooting(10, Tsit5; grid_coarsening = false) 1.73 ± 0.71 ms 1.73 ± 0.71 ms 1 ± 0.58
Simple Pendulum/IIP/MultipleShooting(10, Tsit5; grid_coarsening = true) 2.93 ± 0.67 ms 2.9 ± 0.67 ms 1.01 ± 0.33
Simple Pendulum/IIP/MultipleShooting(100, Tsit5; grid_coarsening = false) 0.0674 ± 0.0083 s 0.0682 ± 0.01 s 0.988 ± 0.19
Simple Pendulum/IIP/MultipleShooting(100, Tsit5; grid_coarsening = true) 0.0833 ± 0.014 s 0.0834 ± 0.022 s 0.999 ± 0.31
Simple Pendulum/IIP/Shooting(Tsit5()) 0.205 ± 0.04 ms 0.204 ± 0.039 ms 1 ± 0.28
Simple Pendulum/OOP/BoundaryValueDiffEqMIRK.MIRK2() 0.482 ± 0.02 s 0.488 ± 0.016 s 0.988 ± 0.053
Simple Pendulum/OOP/BoundaryValueDiffEqMIRK.MIRK3() 13.9 ± 8.6 ms 13.5 ± 8.4 ms 1.03 ± 0.9
Simple Pendulum/OOP/BoundaryValueDiffEqMIRK.MIRK4() 3.15 ± 0.26 ms 3.11 ± 0.25 ms 1.01 ± 0.12
Simple Pendulum/OOP/BoundaryValueDiffEqMIRK.MIRK5() 8.9 ± 0.74 ms 8.94 ± 0.73 ms 0.996 ± 0.12
Simple Pendulum/OOP/BoundaryValueDiffEqMIRK.MIRK6() 1.62 ± 0.15 ms 1.61 ± 0.15 ms 1.01 ± 0.13
Simple Pendulum/OOP/MultipleShooting(10, Tsit5; grid_coarsening = false) 3.52 ± 2.2 ms 3.53 ± 2.2 ms 0.997 ± 0.89
Simple Pendulum/OOP/MultipleShooting(10, Tsit5; grid_coarsening = true) 6.09 ± 4.3 ms 6.1 ± 4.3 ms 0.999 ± 0.99
Simple Pendulum/OOP/MultipleShooting(100, Tsit5; grid_coarsening = false) 0.128 ± 0.01 s 0.126 ± 0.0092 s 1.01 ± 0.11
Simple Pendulum/OOP/MultipleShooting(100, Tsit5; grid_coarsening = true) 0.158 ± 0.025 s 0.157 ± 0.036 s 1.01 ± 0.28
Simple Pendulum/OOP/Shooting(Tsit5()) 0.56 ± 0.22 ms 0.559 ± 0.22 ms 1 ± 0.56
time_to_load 4.7 ± 0.021 s 4.69 ± 0.013 s 1 ± 0.0052
### Benchmark Plots A plot of the benchmark results has been uploaded as an artifact to the workflow run for this PR. Go to "Actions"->"Benchmark a pull request"->[the most recent run]->"Artifacts" (at the bottom).

@ErikQQY
Copy link
Member Author

ErikQQY commented Nov 3, 2025

CI failing seems unrelated

@ErikQQY ErikQQY merged commit 78cceb2 into master Nov 3, 2025
22 of 31 checks passed
@ErikQQY ErikQQY deleted the qqy/sparse_optimization branch November 3, 2025 18:54
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Reviewers

No reviews

Assignees

No one assigned

Labels

None yet

Projects

None yet

Milestone

No milestone

Development

Successfully merging this pull request may close these issues.

Allow more autodiff backends when using internal optimization solvers

2 participants

@ErikQQY