This is the release note of v7.8.0. See here for the complete list of solved issues and merged PRs.

Highlights

This release adds support for CUDA 11, NumPy 1.19, and SciPy 1.5.
We expect this version to be the final release for v7.x series. Please start testing your workloads with the latest v8.x pre-release.

Notes on CUDA 11.0 support

Update on 2020-09-23: cupy-cuda110 package is now available on PyPI! cupy-cuda110 wheel packages are currently available only for Windows. We are going to publish Linux wheels once we got approval from the PyPI team. (update on 2020-08-21: Meanwhile, Linux wheels can be downloaded from the Assets section below (or pip install cupy-cuda110 -f https://github.com/cupy/cupy/releases/tag/v7.8.0). Those wheels will be removed once we published the package on PyPI.)
cupy-cuda110 packages are built with cuDNN support but without bundled cuDNN shared libraries (see #3724 for the discussion). To use cuDNN features, you need to install cuDNN v8.0.x via the system package manager (e.g, apt install libcudnn8 or yum install libcudnn8) or manually install it and set LD_LIBRARY_PATH (Linux) or PATH (Windows) environment variables.
When building CuPy from source with CUDA 11.0, g++-6 or later is required. See the installation guide for the detailed instructions.

Fix broken message format (#3698)
Support order argument in cupy.ones, cupy.full and cupy.eye (#3699, thanks @grlee77!)
Fix sparse matrix related test failures on CUDA11 (#3761)
Allow MatDescriptor to be pickle-able (#3771)
Skip FFT input checks for some CUDA >= 10.1 cases (#3792)
Add temporary fixes for cuDNN v8 (#3794)
Fix error message broken (#3800)
Fix cuSparse build failure on Windows (#3809)

Fix wrong test file name (#3754)
Relax tolerance in TestArrayElementwiseOp::test_doubly_broadcasted_pow (#3762)
Skip tests failing due to exception type changes in NumPy 1.19 (#3787)
Avoid testing exception type match on NumPy 1.19 (#3797)
Skip TestDiaMatrixScipyComparison failing with scipy>=1.5.0 (#3805)