Cupy

This is the release note of v6.7.0. See [here](https://github.com/cupy/cupy/milestone/64?closed=1) for the complete list of solved issues and merged PRs.

As [announced previously](https://chainer.org/announcement/2019/08/21/python2.html), this is the final release of v6 series, which is the last version supporting Python 2.


Enhancements

- Fix slow import of cupy (2765, thanks cgohlke!)
- Support 64 bit extent `randint` (2855)
- Fix `testing.numpy_cupy_` decorators for skips (2892)
- Show warning for cuFFT bug in `irfftn` (2959)

Bug Fixes

- Fix axes handling in `_fftn` (2752)
- Detect interpreter shutdown for proper `__del__` behavior (2812)
- Fix `true_divide` with dtype argument (2835)
- Fix split and array_split with indices overrun (2837)
- Fix split and array_split with unordered indices supplied (2856)
- Avoid looking up null pointers' attributes (2868, thanks leofang!)
- Fix `testing.shaped_random` for shape `()` (2895)

Documentation

- Fix the docstring format of `cupy.asarray` (2826, thanks leofang!)


Tests

- Use more stable input to test `linalg.matrix_power` (2791)
- Fix decorator usage in tests (2850)
- Test `ifloordiv` with numpy 1.18 (2879)
- Fix `test_helper.py` for NumPy 1.18 (2912)
- Avoid 0s in the diagonal of `TestSolveTriangular` inputs (2930)

Others

- Add Mergify configuration file (2934)

This is the release note of v6.6.0. See [here](https://github.com/cupy/cupy/milestone/61?closed=1) for the complete list of solved issues and merged PRs.

Highlights
Python 3.8 is now officially supported and we are providing wheels for this version.

Enhancements
- Support complex dtypes in `cupy.where` (2615, thanks AntoineDujardin!)
- Update `__cuda_array_interface__` to protocol version 2 (2669, thanks leofang!)
- Remove `cupy.cupyx` (2724)
- Support Python 3.8 on Windows (2751)

Bug Fixes
- Make exceptions picklable (2567)
- Support 0-sized arrays for `linalg.qr` (2602, thanks IvanYashchuk)
- Do not allow `reshape` on empty arrays (2652)
- Cache `ElementwiseKernel` kernel globally instead of per instance (2659)
- Handle tuple with value 0 and return empty array (2706, thanks quasiben!)

Code Fixes
- Remove insignificant assertion (2719)

Documentation
- Document more environment variables (2612, thanks leofang!)
- Update `CODE_OF_CONDUCT` typo (2628)
- Fix document format of `as_strided` (2713)
- Update requirements (2755)

Tests
- Add missing FlexCI configurations (2591)
- Add FlexCI configurations (2715)
- Skip tests that segfault when using SciPy 1.3.x (2721, thanks grlee77!)
- Fix broken version specification in FlexCI dockerfiles (2736)

Installation
- Package tests in sdist (2572, thanks jakirkham!)
- Fix url to use the home page address (2590)
- Import CUDA headers from CUDA 10.1 Update 2 (10.1.243) (2605)
- Fix invaild requirements (2640)
- Add software description to `setup.py` (2653)

Examples
- Show better performance improvement in `examples/stream/map_reduce.py` (2603, thanks leofang!)

Others
- Fix `AttributeError` of `stride_tricks` (2705)

This is the release note of v6.5.0. See [here](https://github.com/cupy/cupy/milestone/59?closed=1) for the complete list of solved issues and merged PRs.

Enhancements

- Ignore warning caused by `fastrlock` (2501)
- Fix `cupy.repeat` error message about `repeats` argument type (2506)

Bug Fixes

- Fix `coosort` (2487, thanks econtal!)

Documentation

- Fix some typo (2527, thanks garanews!)
- Fix dead links in NumPy docs in random functions (2554)

Tests

- Fix memory pool disabled during tests (2502)
- Move CI requirements to CuPy repository (2543)
- Fix for NumPy 1.14.x compatibility (2546)

This is the release note of v6.4.0. See [here](https://github.com/cupy/cupy/milestone/58?closed=1) for the complete list of solved issues and merged PRs.

Enhancements

- Support for `shape` argument in `*_like` functions (2418, thanks pentschev!)

Bug Fixes

- Normalize hidden layer strides in cuDNN RNN (2448)

Documentation

- Add appropriate pointer types in Cython to the contribute guide (2480, thanks leofang!)

Installation

- Fix NumPy version in Dockerfile (2447)

Tests

- Add NumPy 1.17.1 to skip list of 0-length `ifft` test (2445)
- Fix `TestDLTensorMemory.test_delete` (2458)
- Fix data race in advanced indexing test (2482)

This is the release note of v6.3.0. See [here](https://github.com/cupy/cupy/milestone/56?closed=1) for the complete list of solved issues and merged PRs.

Highlights

- NumPy 1.17 is now officially supported.

New Features

- Add `diagonal` method for `cupyx.scipy.sparse.dia_matrix` (2412, thanks grlee77!)
- Support `allow_pickle` in `cupy.load` and `cupy.save` (2291)

Bug Fixes

- Make `cupy.tensordot` use Tensor Core also in case of compute-capability > 70 (2335)

Code Fixes

- Fix unused imports and import orders (2336)

Documentation

- Add new functions to API reference (2317)
- Fix for `linalg.svd` documentation (2322, thanks IvanYashchuk!)
- Fix markup in `linalg.svd` docs (2327)
- Fix doctest failure related to `__array_function__` support (2353)
- Add 1.17 to supported NumPy versions (2413)

Tests

- Skip some ndarray-elementwise-op tests as temporary fix (2378)
- Remove `random.power` test with forbidden value (2380)
- Skip some FFT tests on NumPy 1.17.0 due to a NumPy bug (2381, thanks grlee77!)
- Avoid using `numpy.nonzero` in `_make_decorator` (2395)
- Skip `nonzero` test for 0d arrays in NumPy 1.17 (2401)
- Increase numbers of retries for K-S tests (2402)

This is the release note of v6.2.0. See [here](https://github.com/cupy/cupy/milestone/53?closed=1) for the complete list of solved issues and merged PRs.

New Features

- Allow copying in the format `cupy_array[:] = numpy_array` (2219, thanks pentschev!)

Enhancements

- Fix `_preprocess_args` to avoid calling hasattr (2263)
- Simplify `_prepare_multiple_array_indexing` (2264)
- Add `constant` modification to `vector_equal` (2265)
- Remove unused code (2275)
- Remove warning about nvcc absence (2302)

Bug Fixes

- Make `__cuda_array_interface__()['strides']` be `tuple` (2273, thanks leofang!)
- Avoid `CUDNN_STATUS_BAD_PARAMS` (2285, thanks himkt!)
- Add wrappers for `can_cast`, `common_type` and `result_type` functions (2304, thanks pentschev!)

Documentation

- Document interoperability with `mpi4py` (2288, thanks leofang!)
- Update installation guide (2297)

Tests

- Enhance tests for type routines (2311)