CuPy v11 Release Plan

Planned Features

• #6244 📌 v11.0.0a2
• #4897 📌 v11.0.0b1
• Explore possibility of #3972 📌 v11.0.0b2
• Wheels using CUDA Enhanced Compatibility https://github.com/cupy/cupy/pull/6730 📌 v11.0.0rc1
• Make downstream packages easier to use CuPy 📌 v11.0.0b2
  • Provide an universal wheel package for downstream libraries: #6012, #6688
  • Document guideline for downstream libraries (c.f., NumPy: For downstream package authors)
• #6180 📌 v11.0.0b3
• Deprecation 📌 v11.0.0b3
  • CUDA: Keep CUDA 10.2 for Jetson. ~Drop some of the 11.x variants.~ Keep supporting all 11.x variants by unifying the wheel into single cupy-cuda11x package.
  • ROCm: Drop 4.2 or earlier. Keep supporting ROCm 4.3 as it is the last version that runs on gfx80x.
  • Python: No deprecation (support 3.7+).
  • NumPy: Drop NumPy 1.19.
  • SciPy: Drop SciPy 1.4.
• Increase NumPy/SciPy API Coverage

Schedule

• v11.0.0a1 (v10.1.0) (2022-01-20)
• v11.0.0a2 (v10.2.0) (2022-02-25)
• v11.0.0b1 (v10.3.0) (2022-03-31)
• v11.0.0b2 (v10.4.0) (2022-04-27)
• v11.0.0b3 (v10.5.0) (2022-05-26)
• v11.0.0rc1 (v10.6.0) (2022-06-30)
  • Revisit #6460
• v11.0.0 (2022-07-28)

Notes:

• The schedule and listed features are tentative and may be subject to change depending on the progress.
• The code will be frozen around 15:00 on 2 days before the release date for QA. Dates are in JST.
• v10.x.x releases will only contain backported pull-requests (mainly bug-fixes) from the v11 (master) branch.

Past Release Plans

• v9: https://github.com/cupy/cupy/issues/3891
• v10: https://github.com/cupy/cupy/issues/5049

[asi1024]

Released v11.0.0! https://github.com/cupy/cupy/releases/tag/v11.0.0

[jiapei100]

@kmaehashi

I got some issue when using XGBoost, please refer to https://stackoverflow.com/questions/71378334/attributeerror-module-cupy-has-no-attribute-ndarray.