Skip to content
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

Integrate batch Hessian in ReducedSpaceEvaluator routine #179

Closed
wants to merge 8 commits into from
Closed

Integrate batch Hessian in ReducedSpaceEvaluator routine #179

wants to merge 8 commits into from

Conversation

frapac
Copy link
Member

@frapac frapac commented Jun 18, 2021

This PR allows to use the batch Hessian in any optimization routine.

Further:

  • we can now evaluate the Jacobian on the GPU, in batch
  • we use some metaprogramming to generate automatically the different batch routines in src/Evaluators/reduced_evaluator.jl
  • we remove all the previous code that we used before to compute the reduced Hessian (inefficient on the GPU)
  • add a BridgeDeviceEvaluator to wrap an Evaluator on the GPU and move the results back to the CPU

@frapac
evaluator: add HessianLagrangian in ReducedSpaceEvaluator
8b82f76 
* add support for batch Hessian
@frapac
evaluator: add batch_jacobian function
f3c41d6
@frapac
evaluator: add bridge in SlackEvaluator
83fdf2c
@frapac
evaluator: proper definition of batch Hessian
28b431d 
* generate Hessian code with metaprogramming
* change signature of hessian_lagrangian_penalty_prod!
* ReducedSpaceEvaluator now uses batch Hessian by default
@frapac
evaluator: add BridgeDeviceEvaluator
6298689 
Allow to deport the computation on the GPU, while using
a CPU compatible solver.
@frapac
evaluator: fix batch Hessian on the GPU
c284569 
* add CUSOLVERRF in the dependencies
@frapac frapac added the Q3 label Jun 18, 2021
@frapac frapac added this to the Q3 milestone Jun 18, 2021
@frapac frapac requested a review from michel2323 June 18, 2021 14:18
@frapac
Copy link
Member Author

frapac commented Jun 18, 2021

Side note: this PR target the PR#178, to ease the reviewing process.

This was referenced Jun 25, 2021
Closed
Merged
michel2323
michel2323 approved these changes Jun 30, 2021
View reviewed changes
.github/workflows/action.yml Show resolved Hide resolved
Project.toml
@@ -19,7 +20,7 @@ SparseDiffTools = "47a9eef4-7e08-11e9-0b38-333d64bd3804"
TimerOutputs = "a759f4b9-e2f1-59dc-863e-4aeb61b1ea8f"

[compat]
CUDA = "~2.3, 3.0"
CUDA = "~2.3, ~3.2"
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Sorry, this creates a conflict with my push to develop where I forced 2.6 instead of 2.3. 2.3 has issues on Summit.

Copy link
Member Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This is resolved

src/Evaluators/bridge_evaluator.jl Show resolved Hide resolved
src/Evaluators/common.jl Show resolved Hide resolved
test/Evaluators/api.jl Show resolved Hide resolved
src/Evaluators/reduced_evaluator.jl Show resolved Hide resolved
benchmark/batch_hessian.jl Show resolved Hide resolved
@frapac frapac mentioned this pull request Jul 2, 2021
Closed
@frapac
Copy link
Member Author

frapac commented Jul 2, 2021

Closing in favor of #185

@frapac frapac closed this Jul 2, 2021
@frapac frapac deleted the fp/review branch July 23, 2021 13:23
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Reviewers

@michel2323 michel2323 michel2323 approved these changes

Assignees
No one assigned
Labels
Q3
Projects
None yet
Milestone
Q3
Development

Successfully merging this pull request may close these issues.

None yet
2 participants
@frapac @michel2323