ReleaseNVIDIANVIDIApublished Oct 13, 2025seen Jun 26

NVIDIA/cuEquivariance v0.7.0

NVIDIA/cuEquivariance

Open original ↗

Captured source

source ↗
GH
GitHub/github.com/NVIDIA/cuEquivariance
NVIDIA/cuEquivariance v0.7.0
Source ↗
published Oct 13, 2025seen Jun 26captured Aug 31http 200method plain

v0.7.0

Repository: NVIDIA/cuEquivariance

Tag: v0.7.0

Published: 2025-10-13T20:10:16Z

Prerelease: no

Release notes:

Added

  • Support for CUDA 13
  • Support for Python 3.13
  • [Torch] MACE example in the documentation
  • [JAX] MACE and NequIP examples in the folder cuequivariance_jax/examples
  • [Torch/JAX] The Segmented Polynomial operation's math_dtype argument now accepts method-specific string values, with each method supporting different options
  • [Torch] ONNX export and TensorRT runtime plugin support for triangle attention, triangle multiplication and attention with pairwise bias
  • [Torch] Support for caching the "bias tensor z" calculated from the proj_z linear layer in the Attention with pairwise bias kernel (as implemented in the Boltz code)

Bug fix

  • Correct the documentation CUEQ_TRITON_TUNING_MODE -> CUEQ_TRITON_TUNING
  • [JAX] Make triangle_multiplicative_update use the same tuning cache files as its PyTorch counterpart

What's Changed

  • [JAX] expand math_dtype argument by @mariogeiger in https://github.com/NVIDIA/cuEquivariance/pull/153
  • [JAX] add (partial) jax.vmap support for triangle attention and triangle multiplicative update by @mariogeiger in https://github.com/NVIDIA/cuEquivariance/pull/155
  • [JAX] Use nondiff_argnums instead of nondiff_argnames by @mariogeiger in https://github.com/NVIDIA/cuEquivariance/pull/158
  • [JAX] trimul batch dims and int64 support by @mariogeiger in https://github.com/NVIDIA/cuEquivariance/pull/157
  • Merge release to main by @mariogeiger in https://github.com/NVIDIA/cuEquivariance/pull/169
  • apb w/ caching api changes by @hsadasivan in https://github.com/NVIDIA/cuEquivariance/pull/174
  • [JAX] fix benchmarking function by @mariogeiger in https://github.com/NVIDIA/cuEquivariance/pull/161
  • Enabling richer string math_dtype for SegmentedPolynomial by @phiandark in https://github.com/NVIDIA/cuEquivariance/pull/172
  • [JAX] make trimul cache name/keys similar to their torch counterpart by @mariogeiger in https://github.com/NVIDIA/cuEquivariance/pull/173
  • api defaults change for apb with caching by @hsadasivan in https://github.com/NVIDIA/cuEquivariance/pull/179
  • setup copy-pr-bot by @mariogeiger in https://github.com/NVIDIA/cuEquivariance/pull/180
  • update ruff by @mariogeiger in https://github.com/NVIDIA/cuEquivariance/pull/181
  • GHA tests by @mariogeiger in https://github.com/NVIDIA/cuEquivariance/pull/182
  • Update tolerance parameters in spherical harmonics test by @mariogeiger in https://github.com/NVIDIA/cuEquivariance/pull/185
  • update aot flag name in docstring by @hsadasivan in https://github.com/NVIDIA/cuEquivariance/pull/186
  • [JAX] add sharding notes in docstrings by @mariogeiger in https://github.com/NVIDIA/cuEquivariance/pull/197
  • update changelog and version by @mariogeiger in https://github.com/NVIDIA/cuEquivariance/pull/196

Full Changelog: https://github.com/NVIDIA/cuEquivariance/compare/v0.6.0...v0.7.0