v0.4.2

Added

• Precompiled binaries for Linux musl (#31)

Fixed

• Partially fixed building for ROCm, see notes (#30)

v0.4.1

Added

• Precompiled binaries for CUDA 11.4+ (cuDNN 8.2+) and CUDA 11.8+ (cuDNN 8.6+) (#27)

Changed

• Precompiled binaries to assume glibc 31+ (#27)

v0.4.0

Changed

• Bumped XLA (Tensorflow) version to 2.11.0 (#25)

v0.3.0

Changed

• Bumped XLA (Tensorflow) version to 2.8.0 (#15)

Removed

• Dropped support for CUDA 10.2 and 11.0, now 11.1+ is required (#17)

v0.2.0

Added

• Added support for Apple Silicon (#9)

Changed

• Bumped XLA (Tensorflow) version to 2.6.0 (#9)

v0.1.1

Changed

• Build for older glibc versions (#3)

v0.1.0

Initial release.