Open
Conversation
rayg1234
commented
Feb 20, 2026
rayg1234
commented
Feb 20, 2026
github-merge-queue bot
pushed a commit
that referenced
this pull request
Feb 21, 2026
NVIDIA has released a suite of tools, of which includes a neighborlist. Initial benchmarks look promising for both CPU and GPU performance. Currently our external cpu implementation uses pymatgen. - [x] cpu support - [x] gpu support <img width="4766" height="2060" alt="image" src="https://github.com/user-attachments/assets/5d514209-c876-4ac5-a59e-60e4f337b922" /> Note - we should bump minimum python version to `3.11` for the CI. # TODO * GP is broken for V3. A separate PR will abstract it out to support the other graph gen methods, including v3. see #1791 * `external_graph_gen_method` is unused, to use nvidia graph gen with this, we need a followup PR to add this option to the FairchemCalculator (and deprecate `external_graph_gen` from inference_settings) --------- Co-authored-by: Ray Gao <rgao@meta.com> Co-authored-by: Ray Gao <7001989+rayg1234@users.noreply.github.com>
mshuaibii
reviewed
Feb 23, 2026
misko
reviewed
Feb 23, 2026
misko
reviewed
Feb 23, 2026
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Branch off #1737
Previously, we passed the
node_partitioninfo via data_dict to radius_graph_pbc_v2, this allowed v2 to generate a partial graph efficiently where the target atoms are restricted to only atoms in the current graph partition. With v3 (nvidia graph gen), we can't do this so instead we feed the partition intogenerate_graph, which is cleaner anyways, and do the partitioning that way. This would make v2 slightly slower but we should default to v3 anywaysget_pbc_distancesbecause we don't want to do the O(edge) operations on the entire graph. In particular, the edge distance bmm is very inefficient when number of edges is large (see Early Partition and Parallel Graph Gen #1630)MD speed with v2 before and after the change
