-
Notifications
You must be signed in to change notification settings - Fork 4
Add benchmark scripts #173
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.
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
base: main
Are you sure you want to change the base?
Changes from all commits
File filter
Filter by extension
Conversations
Jump to
Diff view
Diff view
There are no files selected for viewing
Original file line number | Diff line number | Diff line change |
---|---|---|
@@ -0,0 +1,23 @@ | ||
``benchmark``: performance sanity checking | ||
========================================== | ||
|
||
This provides a basic benchmarking script intended to catch major performance regressions (or improvements!). It emulates a "training"-style workload of computing energy and forces for a number of similar systems with pre-computed neighborlists and settings. In particular, we run PME forward and backward calculations for supercells of cubic CsCl crystals, in different sizes. Results are stored as a time-stamped ``.yaml`` file, together with some basic system and version information (which is just the output of ``git`` for now). | ||
There was a problem hiding this comment. Choose a reason for hiding this commentThe reason will be displayed to describe this comment to others. Learn more. We run PME or P3M? I think the latter is preferred. You can take the version of torch-pme. Thanks to setuptools-scm the version will contain the version and the commit! |
||
|
||
|
||
Usage | ||
----- | ||
|
||
First, make sure to install ``torch-pme`` with the right dependencies in a fresh enviroment: | ||
|
||
.. code-block:: bash | ||
|
||
pip install .[metatensor,examples] | ||
|
||
Then, run | ||
|
||
.. code-block:: bash | ||
|
||
python run.py | ||
|
||
|
||
Please include the output ``.yaml`` file in any major pull request. We recommend using a H100 as benchmark GPU (internally, we use the ``kuma`` cluster). |
There was a problem hiding this comment. Choose a reason for hiding this commentThe reason will be displayed to describe this comment to others. Learn more. We could just for consistency in the project use a |
Original file line number | Diff line number | Diff line change |
---|---|---|
@@ -0,0 +1,11 @@ | ||
#=============================================================================== | ||
# Created using the Atomic Simulation Environment (ASE) | ||
|
||
# Fri Sep 20 14:44:08 2024 | ||
|
||
#======================================================= | ||
lattice_vector 1.0000000000000000 0.0000000000000000 0.0000000000000000 | ||
lattice_vector 0.0000000000000000 1.0000000000000000 0.0000000000000000 | ||
lattice_vector 0.0000000000000000 0.0000000000000000 1.0000000000000000 | ||
atom 0.0000000000000000 0.0000000000000000 0.0000000000000000 Cs | ||
atom 0.5000000000000000 0.5000000000000000 0.5000000000000000 Cl |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Maybe add a comment that we run only single precision (because we showed in the paper that it makes no difference!)