Skip to content

Fix L-BFGS options #2080

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

Open
vl-dud wants to merge 1 commit into
base: master
Choose a base branch
from
Open

Fix L-BFGS options #2080

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
21 changes: 10 additions & 11 deletions deepxde/optimizers/config.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -60,6 +60,16 @@ def set_LBFGS_options(
LBFGS_options["maxfun"] = maxfun if maxfun is not None else int(maxiter * 1.25)
LBFGS_options["maxls"] = maxls

# Backend-dependent options
Copy link
Copy Markdown
Contributor

@echen5503 echen5503 Mar 31, 2026

Choose a reason for hiding this comment

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

Please verify my understanding: set_hvd_opt_options sets the LFBGS options along with doing the backend-specific operations, but in reality set_LFBGS_options should handle this itself, causing the bug that when you call set_LGBGS_options by itself, it wont update iter_per_step and fun_per_step.

Copy link
Copy Markdown
Contributor Author

@vl-dud vl-dud Mar 31, 2026

Choose a reason for hiding this comment

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

set_LBFGS_options is expected to update all L-BFGS-related parameters, including iter_per_step and fun_per_step used in the Paddle and PyTorch backends. However, the current implementation does not update them.

Copy link
Copy Markdown
Contributor

@echen5503 echen5503 Apr 1, 2026

Choose a reason for hiding this comment

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

Ok. This looks good.

if backend_name in ["pytorch", "paddle"]:
# number of iterations per optimization call
LBFGS_options["iter_per_step"] = min(1000, LBFGS_options["maxiter"])
LBFGS_options["fun_per_step"] = (
LBFGS_options["maxfun"]
* LBFGS_options["iter_per_step"]
// LBFGS_options["maxiter"]
)


def set_NNCG_options(
lr=1,
Expand Down Expand Up @@ -149,14 +159,3 @@ def set_hvd_opt_options(
set_NNCG_options()
if hvd is not None:
set_hvd_opt_options()


# Backend-dependent options
if backend_name in ["pytorch", "paddle"]:
# number of iterations per optimization call
LBFGS_options["iter_per_step"] = min(1000, LBFGS_options["maxiter"])
LBFGS_options["fun_per_step"] = (
LBFGS_options["maxfun"]
* LBFGS_options["iter_per_step"]
// LBFGS_options["maxiter"]
)
Loading