v12.2.0

版本发布时间: 2023-08-08 16:41:11

cupy/cupy最新发布版本:v13.3.0(2024-08-22 15:42:45)

This is the release note of v12.2.0. 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!

✨ Highlights

Support for CUDA 12.2

CuPy now supports CUDA 12.2. Note that there is a known issue on CUDA 12.2 for Windows. See #7776 for details.

Support us via GitHub Sponsors!

As a part of our effort to make CuPy sustainable, we have enrolled in GitHub Sponsors to accept donations. Help us to support CuPy’s development and contribute to ease the required infrastructure costs due to the need of GPU enabled CI platforms and resources to build binary packages.

As a NumFOCUS Sponsored Project, funds sponsored through the GitHub Sponsors are collected and disbursed via NumFOCUS, a 501(c)(3) public charity in the United States, which acts as the fiscal sponsor for the project.

🛠️ Changes without compatibility

Deprecation of `cupy-wheel` Package

Due to the recent specification change in Pip 23.1, it became difficult for cupy-wheel to ensure detecting the CUDA version installed correctly. As discussed in RFC #7628, we have decided to remove this package in CuPy v13. To allow existing projects using cupy-wheel to continue to work, the package remains available for v12 releases.

📝 Changes

Enhancements

Minor updates for cuQuantum/cuTensorNet support (#7730)
Bump mypy version to 1.4.1 (#7736)
Support CUDA 12.2 (#7752)

Performance Improvements

Fix random module performance regression (#7592)

Bug Fixes

Fix returned CUDA statuses not being checked (#7618)
Fix cuSPARSE error message (#7684)
Fix memory pool to try resolve fragmentation when limit is set (#7685)
Fix type/exception annotations in cuSPARSE binding (#7703)
Update pylibcugraph weakly connected components call (#7704)
Improve detection for package installation source on Windows (#7711)
Temporarily disable CUB histogram (#7716)
Fix aweights type not checked in cupy.cov (#7717)
Revert FP16 headers from CUDA 12.2.0 to CUDA 12.1.1 (#7773)

Code Fixes

Introduce cython-lint (#7612)

Documentation

Improve README and Installation Guide (#7599)
update badges (#7600)
Fix small typos in docstrings (#7657)
Fix docstring of asarray (#7695)
Add CUDA 12.2 to list of supported CUDA (#7756)
Remove incorrect cupyx.distributed.NCCLBackend.all_gather comment (#7765)
Fix Note highlight sections in README (#7770)
Add notes for CUDA 12.2 on Windows support (#7778)

Installation

Fix cupy-wheel package installation fails with pip 23.1+ (#7624)

Tests

Bump versions of static checkers (#7598)
Fix build-cuda test restore-keys not working (#7614)
[v12] Require numpy<1.25 for round_ tests (#7642)
Ignore pkg_resources deprecation warning on import (#7656)
Skip TestLOBPCG::test_maxit_None in CUDA 12.1.1 & cuSOVLER 11.4.5 (#7670)
Bump CUDA minor versions used in CI (#7683)
[v12] Allow specifying Docker repository for CI images (#7690)
Use "/test" tag configuration from pull-request base branch (#7706)
XFAIL known test failures in cuSPARSE module (#7725)
Fix test_fht not to feed cupy.ndarray to scipy.fft.fhtoffset (#7728)
CI: remove explicit Cython installation (#7731)
Fix test_sum_duplicates_incompatibility for SciPy 1.11 (#7768)

Others

Fix flake8-cython not working (#7606)
Add env var to disable RPATH (#7718)
Bump version to v12.2.0 (#7755)

👥 Contributors

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

@12rambau @asi1024 @emcastillo @jnke2016 @kmaehashi @leofang @pelmers @pri1311 @RandomY-2 @takagi

相关地址：原始地址下载(tar) 下载(zip)

1、 cupy-12.2.0.tar.gz 1.87MB

2、 cupy_cuda102-12.2.0-cp310-cp310-manylinux2014_aarch64.whl 34.65MB

3、 cupy_cuda102-12.2.0-cp310-cp310-manylinux2014_x86_64.whl 60.84MB

4、 cupy_cuda102-12.2.0-cp310-cp310-win_amd64.whl 42.66MB

5、 cupy_cuda102-12.2.0-cp311-cp311-manylinux2014_aarch64.whl 35.01MB

6、 cupy_cuda102-12.2.0-cp311-cp311-manylinux2014_x86_64.whl 61.35MB

7、 cupy_cuda102-12.2.0-cp311-cp311-win_amd64.whl 42.64MB

8、 cupy_cuda102-12.2.0-cp38-cp38-manylinux2014_aarch64.whl 36.88MB

9、 cupy_cuda102-12.2.0-cp38-cp38-manylinux2014_x86_64.whl 63.34MB

10、 cupy_cuda102-12.2.0-cp38-cp38-win_amd64.whl 42.81MB

11、 cupy_cuda102-12.2.0-cp39-cp39-manylinux2014_aarch64.whl 35.28MB

12、 cupy_cuda102-12.2.0-cp39-cp39-manylinux2014_x86_64.whl 61.52MB

13、 cupy_cuda102-12.2.0-cp39-cp39-win_amd64.whl 42.8MB

14、 cupy_cuda110-12.2.0-cp310-cp310-manylinux2014_x86_64.whl 76.07MB

15、 cupy_cuda110-12.2.0-cp310-cp310-win_amd64.whl 57.84MB

16、 cupy_cuda110-12.2.0-cp311-cp311-manylinux2014_x86_64.whl 76.57MB

17、 cupy_cuda110-12.2.0-cp311-cp311-win_amd64.whl 57.81MB

18、 cupy_cuda110-12.2.0-cp38-cp38-manylinux2014_x86_64.whl 78.56MB

19、 cupy_cuda110-12.2.0-cp38-cp38-win_amd64.whl 57.98MB

20、 cupy_cuda110-12.2.0-cp39-cp39-manylinux2014_x86_64.whl 76.74MB

21、 cupy_cuda110-12.2.0-cp39-cp39-win_amd64.whl 57.97MB

22、 cupy_cuda111-12.2.0-cp310-cp310-manylinux2014_x86_64.whl 96.61MB

23、 cupy_cuda111-12.2.0-cp310-cp310-win_amd64.whl 79.35MB

24、 cupy_cuda111-12.2.0-cp311-cp311-manylinux2014_x86_64.whl 97.11MB

25、 cupy_cuda111-12.2.0-cp311-cp311-win_amd64.whl 79.32MB

26、 cupy_cuda111-12.2.0-cp38-cp38-manylinux2014_x86_64.whl 99.1MB

27、 cupy_cuda111-12.2.0-cp38-cp38-win_amd64.whl 79.49MB

28、 cupy_cuda111-12.2.0-cp39-cp39-manylinux2014_x86_64.whl 97.28MB

29、 cupy_cuda111-12.2.0-cp39-cp39-win_amd64.whl 79.49MB

30、 cupy_cuda11x-12.2.0-cp310-cp310-manylinux2014_aarch64.whl 96.69MB

31、 cupy_cuda11x-12.2.0-cp310-cp310-manylinux2014_x86_64.whl 85.43MB

32、 cupy_cuda11x-12.2.0-cp310-cp310-win_amd64.whl 67.02MB

33、 cupy_cuda11x-12.2.0-cp311-cp311-manylinux2014_aarch64.whl 97.9MB

34、 cupy_cuda11x-12.2.0-cp311-cp311-manylinux2014_x86_64.whl 85.93MB

35、 cupy_cuda11x-12.2.0-cp311-cp311-win_amd64.whl 66.99MB

36、 cupy_cuda11x-12.2.0-cp38-cp38-manylinux2014_aarch64.whl 99.17MB

37、 cupy_cuda11x-12.2.0-cp38-cp38-manylinux2014_x86_64.whl 87.92MB

38、 cupy_cuda11x-12.2.0-cp38-cp38-win_amd64.whl 67.16MB

39、 cupy_cuda11x-12.2.0-cp39-cp39-manylinux2014_aarch64.whl 97.43MB

40、 cupy_cuda11x-12.2.0-cp39-cp39-manylinux2014_x86_64.whl 86.1MB

41、 cupy_cuda11x-12.2.0-cp39-cp39-win_amd64.whl 67.16MB

42、 cupy_cuda12x-12.2.0-cp310-cp310-manylinux2014_aarch64.whl 89.5MB

43、 cupy_cuda12x-12.2.0-cp310-cp310-manylinux2014_x86_64.whl 78.24MB

44、 cupy_cuda12x-12.2.0-cp310-cp310-win_amd64.whl 59.84MB

45、 cupy_cuda12x-12.2.0-cp311-cp311-manylinux2014_aarch64.whl 90.71MB

46、 cupy_cuda12x-12.2.0-cp311-cp311-manylinux2014_x86_64.whl 78.74MB

47、 cupy_cuda12x-12.2.0-cp311-cp311-win_amd64.whl 59.81MB

48、 cupy_cuda12x-12.2.0-cp38-cp38-manylinux2014_aarch64.whl 91.98MB

49、 cupy_cuda12x-12.2.0-cp38-cp38-manylinux2014_x86_64.whl 80.73MB

50、 cupy_cuda12x-12.2.0-cp38-cp38-win_amd64.whl 59.98MB

51、 cupy_cuda12x-12.2.0-cp39-cp39-manylinux2014_aarch64.whl 90.24MB

52、 cupy_cuda12x-12.2.0-cp39-cp39-manylinux2014_x86_64.whl 78.91MB

53、 cupy_cuda12x-12.2.0-cp39-cp39-win_amd64.whl 59.97MB

54、 cupy_rocm_4_3-12.2.0-cp310-cp310-manylinux2014_x86_64.whl 36.77MB

55、 cupy_rocm_4_3-12.2.0-cp311-cp311-manylinux2014_x86_64.whl 37.24MB

56、 cupy_rocm_4_3-12.2.0-cp38-cp38-manylinux2014_x86_64.whl 38.99MB

57、 cupy_rocm_4_3-12.2.0-cp39-cp39-manylinux2014_x86_64.whl 37.36MB

58、 cupy_rocm_5_0-12.2.0-cp310-cp310-manylinux2014_x86_64.whl 54.67MB

59、 cupy_rocm_5_0-12.2.0-cp311-cp311-manylinux2014_x86_64.whl 55.15MB

60、 cupy_rocm_5_0-12.2.0-cp38-cp38-manylinux2014_x86_64.whl 56.89MB

61、 cupy_rocm_5_0-12.2.0-cp39-cp39-manylinux2014_x86_64.whl 55.26MB

查看：2023-08-08发行的版本