This is the release note of v9.6.0. See here for the complete list of solved issues and merged PRs.

We are running a Gitter chat for general discussions and quick questions. Feel free to join the channel to talk with developers and users!

Announcements

Final release for v9.x series

This is expected to be the last release of the CuPy v9 series. Please start trying your workflow with CuPy v10.0.0rc1 and let us know if you have any feedback!

CuPy now supports CUDA 11.5

Wheels for CUDA 11.5 (cupy-cuda115) are now available.

Removal of Alpha/Beta/RC Wheels from PyPI

• As per the discussion in #5671, we stopped uploading pre-release binary wheels to PyPI for the health of the ecosystem. Pre-release wheels can now be downloaded from the recently introduced custom index (e.g., pip install cupy-cudaXXX -f https://pip.cupy.dev/pre) . Note that the sdist package is available in PyPI for all versions.

• Outdated (v8.0.0rc1 or earlier) pre-release binaries have been removed from PyPI. See #5667 for details.

Changes

Enhancements

• Make show_config runnable without GPU (#5839)
• Merge fp16 headers for CUDA 11.2+ (#6004)
• Support cuTENSOR 1.3.3 (#6005)
• Support CUDA 11.5 for library installer (#6010)
• Display license terms when downloading libraries (#6041)
• Fix error type/message for duplicate value in axis (#5987)

Bug Fixes

• Do not use cuTENSOR unless available (#5885)
• Fix non-determinisitc behavior in cupy.random.shuffle (#5887)
• Fix ndarray.clip to match numpy (#5916)
• Fix __repr__ of mode and scalar in cuTENSOR (#5917)
• Fix max blocksize used in cupyx.optimizing.optimize for HIP (#5931)
• Fix ravel for strides 0 (#5998)
• Fix cuTENSOR installation on Windows (#6022)
• Allow generating cubins for the max known CC (#6024)

Documentation

• Update upgrade guide (#5834)
• Document ppc64le and aarch64 are supported on conda-forge (#5869)
• Improve the comparison table (#5911)
• Add footnotes for functions unimplemented in CuPy (#5954)
• Update the docstring for cholesky (#5960)
• Document CUPY_ACCELERATORS (#5975)
• Add favicon to docs (#5983)
• Support CUDA 11.5 on documents (#6006)
• Replace favicon with high resolution one (#6008)
• Fix typo in copyright line (#6035)

Tests

• Clean up plan cache in a FFT slow test (#5825)
• Copy source directory to support pip 21.3 (#5896)
• Simplify legacy ROCm test script for FlexCI (#5936)
• Relax sparse linalg testing tolerance (#5958)
• CI: Fix ROCm build test (FlexCI) failing (#5965)
• Improve handling of FlexCI test runs (#6002)
• Upload cache even when test failed in FlexCI (#6003)
• CI: Increase timeout for CUDA 11.4 / 11.5 tests (#6040)
• CI: Do not run full combination test even for branch tests for ROCm (#5974)

Others

• Avoid triggering docker workflow on release of forked repos (#5886)
• Bump version to v9.6.0 (#6043)

Contributors

The CuPy Team would like to thank all those who contributed to this release!

@asi1024 @drbeh @emcastillo @kmaehashi @leofang @takagi @toslunar