Release v0.5.1#215
Open
scal444 wants to merge 17 commits into
Open
Conversation
The pip-build path was globbing every .so in rdkit.libs/ into RDKit_LIBS
and every libboost_* into Boost_LIBRARIES, then linking each into every
nvmolkit Python module. That dragged libcairo and libquadmath onto each
module's NEEDED list. libcairo in turn NEEDEDs libXrender/libX11/libXext;
those are on the manylinux lib_whitelist so rdkit-pypi's auditwheel pass
legally left them external, but the nvidia/cuda runtime container we test
in doesn't ship them, so `import nvmolkit.fingerprints` failed at load.
Narrow RDKit_LIBS to the 16 components nvmolkit actually uses (mirroring
the conda path's explicit list) and Boost_LIBRARIES to the same 4 (boost
serialization / iostreams / python<ver> / numpy<ver>). Also patchelf with
--force-rpath so the entry-point modules get DT_RPATH instead of
DT_RUNPATH. The libs inside rdkit.libs/ have no rpath of their own and
rely on RPATH inheritance to resolve second-level deps; rdkit's own
python bindings do the same. Drop the unused ${RDKit_LIBS} link from
_arrayHelpers, which touches zero RDKit symbols.
Verified the rdkit==2026.3.1 py3.12 wheel built with this change loads
and passes the full pytest suite (390 passed, 10 long deselected) on an
H200 in the manylinux+CUDA container with no system X/font libs.
Autotune now steps in 64 element increments by default, cutting down on the search space. CPU space is now physical core limited by default. (cherry picked from commit d0bce61)
(cherry picked from commit 3537c4c)
…IDIA-BioNeMo#194) Giant clustering tasks were hitting signed int overflow, moved to a 2D grid to expand limits by 5 OOM (cherry picked from commit b19bf4f)
Zero out a buffer that needed to be zero'd out every thread iteration, not just at the beginning of the batch. Wrote a reliable reproducer. (cherry picked from commit 4f897f3)
…VIDIA-BioNeMo#183) rdkit/rdkit#9298 (merged into RDKit 2026.03) fixed a bug where a negative energy value caused max(energy * gradScale, 1.0) to clamp to 1, artificially tightening the gradient-tolerance convergence test mid-minimisation. Force fields with stabilising electrostatic or dispersion terms (MMFF94, UFF) can produce negative intermediate energies, so this affected real workloads. Apply the same fix in both nvMolKit BFGS paths, gated on the linked RDKit version so that behaviour is unchanged when built against older RDKit: - src/minimizer/bfgs_minimize.cu (updateDGradKernel, batched path) - src/minimizer/bfgs_minimize_permol_kernels.cu (updateDGrad, per-mol path) Mirrors the version-conditional pattern already used for kRdkitHasGradScaleFix in scaleGradKernel. Signed-off-by: Clay Moore <claytonwaynemoore@gmail.com> (cherry picked from commit 7ad16cf)
…VIDIA-BioNeMo#205) GH issue 202: a non-negligible fraction of molecules embedded to zero conformers. RDKIT_NEW_FLAG_API was a constexpr bool but consumed with #if, which the preprocessor evaluates as 0, and etkdg_stages and rdkit_dist_geom_flattened never linked nvmolkit_versions so versions.h was not even visible. The gate was therefore always off. findChiralSets never set d_structureFlags and loadChiralDataset never read it, so fused-small-ring tetrahedral centers used volScale 1.0 instead of the 0.25 RDKit applies and were deterministically rejected at the tetrahedral check. The same gate guards the constrained tight-bounds branch in addLongRangeDistanceConstraints, which assigns to l and u; those are now non-const so the branch compiles. (cherry picked from commit 525286a)
…cessing (NVIDIA-BioNeMo#204) buildQueryBatchParallel ran addQueryToBatch inside an OpenMP parallel region. The validation error for fragment (disconnected) queries escaped the region and called std::terminate whenever more than one preprocessing thread was used, so the error only surfaced as a RuntimeError on the single-threaded path. Capture the first exception with OpenMPExceptionRegistry and rethrow once the region joins. (cherry picked from commit 6f967ed)
…A-BioNeMo#212) The _mmffOptimization and _uffOptimization bindings declared BatchHardwareOptions default argument values. Boost.Python converts default values to Python when a function is registered, which needs a to-Python converter that only _embedMolecules registers, so importing either module before nvmolkit.types raised "No to_python converter found for nvMolKit::BatchHardwareOptions". The Python wrappers already supply every argument, so the native defaults are dropped, and a regression test imports each module first in a fresh interpreter. (cherry picked from commit 7dc0a38)
* Normalize array inputs for clustering APIs * Simplify CUDA tensor input normalization (cherry picked from commit b6f3f42)
This was an ABI issue - a transitive flag is important for the structure of RDKit data that wasn't being included. It's an RDKit issue but we can fix it by manually including the flag we need to avoid the ABI mismatch. Was not present in conda-forge builds by luck. (cherry picked from commit ad4124e)
Backport the TFD correctness subset of ea09e2d without the unrelated performance changes.
evasnow1992
approved these changes
Jun 24, 2026
evasnow1992
left a comment
evasnow1992
left a comment
Collaborator
There was a problem hiding this comment.
Verified that the set of commits is complete and self-contained, and confirmed that the release notes are clear and comprehensive. The changes look good to me. Thank you for packaging this into a release.
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.
Cherry-picked a bunch from main, and then additional formatting and version change commits