v12.3.0

版本发布时间: 2023-12-07 15:57:50

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

This is the release note of v12.3.0. See here for the complete list of solved issues and merged PRs.

This is the last planned release for the CuPy v12 series. Please start testing your workload with the v13 release candidate to get ready for the final v13 release. To install: pip install -U --pre cupy-cuda11x -f https://pip.cupy.dev/pre. See the Upgrade Guide for the list of possible breaking changes in v13.

💬 Join the Matrix chat to talk with developers and users and ask quick questions!

🙌 Help us sustain the project by sponsoring CuPy!

✨ Highlights

Support for Python 3.12

Binary packages are now available for Python 3.12.

📝 Changes

Enhancements

Add a property to get access to the nccl handle. (#7824)
Add Linux CI for Python 3.12 (#7949)
Bump supported NumPy & SciPy versions (#8001)
ENH: Remove NINF, PINF, Inf,... usages (#7805)
Avoid using numpy.find_common_type (#7810)
Remove conversions of array with ndim > 0 to a scalar (#7895)
scipy.linalg.{tri/tril/triu} are deprecated in SciPy 1.11.0 (#7902)
Fix signal.medfilt complex error type for SciPy>=1.11 (#7909)
Fix return type of division of csr_matrix and dense array for SciPy 1.11 (#7912)
Skip TestSpmatrix on SciPy 1.11 or later (#7918)
Fix test of product, cumproduct, alltrue and sometrue for deprecation (#7936)
Skip fusion round_ tests (#7937)

Bug Fixes

Fix csr2dense to avoid race conditions (#7808)
Fix cuTENSOR contraction descriptor cache (#7817)
Provide .stop() method for cupyx.distributed._Backend (#7960)

Code Fixes

MNT: suppress compiler warning from cupyx.cusolver (#7819)
Fix coding style (#7846)
Remove deprecated ast expr in CuPy JIT (#7944)
Remove unnecessary CUB files from CuPy distribution (#7975)

Documentation

Add -U to pre-release installation command (#7806)
Fix CUDA 12.2 for Windows notice (#7926)

Installation

Fix ROCm Dockerfile not working (#7799)
Avoid warning when uploading packages (#7807)

Tests

Add timeout to Windows CI (#7859)
CI: Remove path length limitation on Windows CI image (#7860)
Fix Windows CI failures (#7865)
Fix Windows + CUDA 12.2 CI (#7910)
Skip test_pos_boolarray if numpy>=1.25 (#7913)
Skip some LOBPCG tests failing with SciPy 1.11 (#7931)
Add NumPy 1.25/1.26 & SciPy 1.11 to CI (#7932)
Skip logspace test in NumPy 1.25 & 1.26 (#7946) (#7951)
Support Python 3.12, add Windows CI (#7958)
Fix Windows test scripts (#7961)
Skip test_parameterize_pytest_impl test for pytest 7.4.3 (#7968)
Filter DeprecationWarning for distutils.dep_util used in Cython (#7999)
Fix TestLOBPCG.test_maxit_None CUDA 12.2 CI failure (#8007)

Others

Fix backport workflow (#7833)
Bump version to v12.3.0 (#8016)

👥 Contributors

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

@asi1024 @emcastillo @kmaehashi @leofang @mtsokol @mvnvidia

相关地址：原始地址下载(tar) 下载(zip)

1、 cupy-12.3.0.tar.gz 1.72MB

2、 cupy_cuda102-12.3.0-cp310-cp310-manylinux2014_aarch64.whl 34.43MB

3、 cupy_cuda102-12.3.0-cp310-cp310-manylinux2014_x86_64.whl 60.63MB

4、 cupy_cuda102-12.3.0-cp310-cp310-win_amd64.whl 42.44MB

5、 cupy_cuda102-12.3.0-cp311-cp311-manylinux2014_aarch64.whl 34.79MB

6、 cupy_cuda102-12.3.0-cp311-cp311-manylinux2014_x86_64.whl 61.13MB

7、 cupy_cuda102-12.3.0-cp311-cp311-win_amd64.whl 42.41MB

8、 cupy_cuda102-12.3.0-cp312-cp312-manylinux2014_aarch64.whl 34.4MB

9、 cupy_cuda102-12.3.0-cp312-cp312-manylinux2014_x86_64.whl 60.96MB

10、 cupy_cuda102-12.3.0-cp312-cp312-win_amd64.whl 42.35MB

11、 cupy_cuda102-12.3.0-cp38-cp38-manylinux2014_aarch64.whl 36.66MB

12、 cupy_cuda102-12.3.0-cp38-cp38-manylinux2014_x86_64.whl 63.12MB

13、 cupy_cuda102-12.3.0-cp38-cp38-win_amd64.whl 42.58MB

14、 cupy_cuda102-12.3.0-cp39-cp39-manylinux2014_aarch64.whl 35.06MB

15、 cupy_cuda102-12.3.0-cp39-cp39-manylinux2014_x86_64.whl 61.3MB

16、 cupy_cuda102-12.3.0-cp39-cp39-win_amd64.whl 42.58MB

17、 cupy_cuda110-12.3.0-cp310-cp310-manylinux2014_x86_64.whl 75.85MB

18、 cupy_cuda110-12.3.0-cp310-cp310-win_amd64.whl 57.61MB

19、 cupy_cuda110-12.3.0-cp311-cp311-manylinux2014_x86_64.whl 76.36MB

20、 cupy_cuda110-12.3.0-cp311-cp311-win_amd64.whl 57.58MB

21、 cupy_cuda110-12.3.0-cp312-cp312-manylinux2014_x86_64.whl 76.18MB

22、 cupy_cuda110-12.3.0-cp312-cp312-win_amd64.whl 57.52MB

23、 cupy_cuda110-12.3.0-cp38-cp38-manylinux2014_x86_64.whl 78.34MB

24、 cupy_cuda110-12.3.0-cp38-cp38-win_amd64.whl 57.75MB

25、 cupy_cuda110-12.3.0-cp39-cp39-manylinux2014_x86_64.whl 76.52MB

26、 cupy_cuda110-12.3.0-cp39-cp39-win_amd64.whl 57.75MB

27、 cupy_cuda111-12.3.0-cp310-cp310-manylinux2014_x86_64.whl 96.39MB

28、 cupy_cuda111-12.3.0-cp310-cp310-win_amd64.whl 79.13MB

29、 cupy_cuda111-12.3.0-cp311-cp311-manylinux2014_x86_64.whl 96.9MB

30、 cupy_cuda111-12.3.0-cp311-cp311-win_amd64.whl 79.1MB

31、 cupy_cuda111-12.3.0-cp312-cp312-manylinux2014_x86_64.whl 96.72MB

32、 cupy_cuda111-12.3.0-cp312-cp312-win_amd64.whl 79.04MB

33、 cupy_cuda111-12.3.0-cp38-cp38-manylinux2014_x86_64.whl 98.88MB

34、 cupy_cuda111-12.3.0-cp38-cp38-win_amd64.whl 79.27MB

35、 cupy_cuda111-12.3.0-cp39-cp39-manylinux2014_x86_64.whl 97.06MB

36、 cupy_cuda111-12.3.0-cp39-cp39-win_amd64.whl 79.27MB

37、 cupy_cuda11x-12.3.0-cp310-cp310-manylinux2014_aarch64.whl 96.47MB

38、 cupy_cuda11x-12.3.0-cp310-cp310-manylinux2014_x86_64.whl 85.21MB

39、 cupy_cuda11x-12.3.0-cp310-cp310-win_amd64.whl 66.8MB

40、 cupy_cuda11x-12.3.0-cp311-cp311-manylinux2014_aarch64.whl 97.67MB

41、 cupy_cuda11x-12.3.0-cp311-cp311-manylinux2014_x86_64.whl 85.72MB

42、 cupy_cuda11x-12.3.0-cp311-cp311-win_amd64.whl 66.77MB

43、 cupy_cuda11x-12.3.0-cp312-cp312-manylinux2014_aarch64.whl 96.72MB

44、 cupy_cuda11x-12.3.0-cp312-cp312-manylinux2014_x86_64.whl 85.55MB

45、 cupy_cuda11x-12.3.0-cp312-cp312-win_amd64.whl 66.71MB

46、 cupy_cuda11x-12.3.0-cp38-cp38-manylinux2014_aarch64.whl 98.94MB

47、 cupy_cuda11x-12.3.0-cp38-cp38-manylinux2014_x86_64.whl 87.7MB

48、 cupy_cuda11x-12.3.0-cp38-cp38-win_amd64.whl 66.94MB

49、 cupy_cuda11x-12.3.0-cp39-cp39-manylinux2014_aarch64.whl 97.21MB

50、 cupy_cuda11x-12.3.0-cp39-cp39-manylinux2014_x86_64.whl 85.88MB

51、 cupy_cuda11x-12.3.0-cp39-cp39-win_amd64.whl 66.94MB

52、 cupy_cuda12x-12.3.0-cp310-cp310-manylinux2014_aarch64.whl 89.28MB

53、 cupy_cuda12x-12.3.0-cp310-cp310-manylinux2014_x86_64.whl 78.02MB

54、 cupy_cuda12x-12.3.0-cp310-cp310-win_amd64.whl 59.61MB

55、 cupy_cuda12x-12.3.0-cp311-cp311-manylinux2014_aarch64.whl 90.49MB

56、 cupy_cuda12x-12.3.0-cp311-cp311-manylinux2014_x86_64.whl 78.52MB

57、 cupy_cuda12x-12.3.0-cp311-cp311-win_amd64.whl 59.58MB

58、 cupy_cuda12x-12.3.0-cp312-cp312-manylinux2014_aarch64.whl 89.53MB

59、 cupy_cuda12x-12.3.0-cp312-cp312-manylinux2014_x86_64.whl 78.34MB

60、 cupy_cuda12x-12.3.0-cp312-cp312-win_amd64.whl 59.52MB

61、 cupy_cuda12x-12.3.0-cp38-cp38-manylinux2014_aarch64.whl 91.75MB

62、 cupy_cuda12x-12.3.0-cp38-cp38-manylinux2014_x86_64.whl 80.51MB

63、 cupy_cuda12x-12.3.0-cp38-cp38-win_amd64.whl 59.75MB

64、 cupy_cuda12x-12.3.0-cp39-cp39-manylinux2014_aarch64.whl 90.02MB

65、 cupy_cuda12x-12.3.0-cp39-cp39-manylinux2014_x86_64.whl 78.69MB

66、 cupy_cuda12x-12.3.0-cp39-cp39-win_amd64.whl 59.75MB

67、 cupy_rocm_4_3-12.3.0-cp310-cp310-manylinux2014_x86_64.whl 36.55MB

68、 cupy_rocm_4_3-12.3.0-cp311-cp311-manylinux2014_x86_64.whl 37.02MB

69、 cupy_rocm_4_3-12.3.0-cp312-cp312-manylinux2014_x86_64.whl 36.79MB

70、 cupy_rocm_4_3-12.3.0-cp38-cp38-manylinux2014_x86_64.whl 38.77MB

71、 cupy_rocm_4_3-12.3.0-cp39-cp39-manylinux2014_x86_64.whl 37.14MB

72、 cupy_rocm_5_0-12.3.0-cp310-cp310-manylinux2014_x86_64.whl 54.45MB

73、 cupy_rocm_5_0-12.3.0-cp311-cp311-manylinux2014_x86_64.whl 54.93MB

74、 cupy_rocm_5_0-12.3.0-cp312-cp312-manylinux2014_x86_64.whl 54.7MB

75、 cupy_rocm_5_0-12.3.0-cp38-cp38-manylinux2014_x86_64.whl 56.67MB

76、 cupy_rocm_5_0-12.3.0-cp39-cp39-manylinux2014_x86_64.whl 55.04MB

查看：2023-12-07发行的版本