NVIDIA/cuEquivariance v0.8.0
NVIDIA/cuEquivariance
Captured source
source ↗published Dec 8, 2025seen 5dcaptured 10hhttp 200method plain
v0.8.0
Repository: NVIDIA/cuEquivariance
Tag: v0.8.0
Published: 2025-12-08T21:27:57Z
Prerelease: no
Release notes:
Added
- Support for CUDA 13 on ARM
- [Torch/JAX] Blackwell-optimized BF16/FP16 forward and backward kernels for
cuet.triangle_attention(runs on compute capabilities 10.0 and 10.3). These kernels provide superior performance especially for long sequences and higher head dimensions. This is only supported on cu13 builds as of this release
Bug fix
- [Torch/JAX] Fixed index overflow and out of bound issues leading to illegal memory access in
cuet.triangle_attention
Notes
- [Torch/JAX] Blackwell-optimized kernels require the sequence length N to be a multiple of 8 for the forward pass; pad the sequence if necessary
- [Torch/JAX] Blackwell-optimized kernels are currently supported only for CUDA 13 builds
What's Changed
- 16bit math datatype for uniform_1d by @mariogeiger in https://github.com/NVIDIA/cuEquivariance/pull/188
- [JAX] code examples by @mariogeiger in https://github.com/NVIDIA/cuEquivariance/pull/156
- [docs] add mace perf figure in docs by @mariogeiger in https://github.com/NVIDIA/cuEquivariance/pull/192
- Adding examples to the documentation by @phiandark in https://github.com/NVIDIA/cuEquivariance/pull/191
- Small SegmentedPolynomial improvements by @phiandark in https://github.com/NVIDIA/cuEquivariance/pull/210
- fix by @mariogeiger in https://github.com/NVIDIA/cuEquivariance/pull/212
- Added wrappers for ONNX/TRT initialization by @borisfom in https://github.com/NVIDIA/cuEquivariance/pull/214
- [JAX] triatt 16bit fix by @mariogeiger in https://github.com/NVIDIA/cuEquivariance/pull/213
- [Torch] Improve error messages for uniform 1d by @mariogeiger in https://github.com/NVIDIA/cuEquivariance/pull/217
- triatt jax bwd dtype issue by @mariogeiger in https://github.com/NVIDIA/cuEquivariance/pull/219
- doc-string updates+ rel-notes by @hsadasivan in https://github.com/NVIDIA/cuEquivariance/pull/220
Full Changelog: https://github.com/NVIDIA/cuEquivariance/compare/v0.7.0...v0.8.0
Notability
notability 6.0/10Notable library update from NVIDIA, equitable equivariance focus