Skip to content
NumPy-like API accelerated with CUDA
Python C++ C Cuda PowerShell Batchfile
Branch: master
Clone or download
emcastillo Merge pull request #2465 from niboshi/ndarray-private-methods
Remove unncecessary `ndarray` private methods
Latest commit 36ecb44 Sep 20, 2019
Type Name Latest commit message Commit time
Failed to load latest commit information.
.github add issue template Nov 16, 2018
.pfnci update URL Aug 28, 2019
appveyor fix AppVeyor configuration to test CuPy Feb 28, 2018
cupy Merge pull request #2465 from niboshi/ndarray-private-methods Sep 20, 2019
cupyx Fix devInfo dtype in inv and unify specification Sep 10, 2019
docker fix numpy version in Dockerfile Aug 26, 2019
docs Merge pull request #2455 from leofang/correct_int_type Sep 19, 2019
examples ValueError: operands could not be broadcast together with shapes... f… Sep 6, 2019
install Merge pull request #2210 from anaruse/support_cuTensor Jul 16, 2019
tests Merge pull request #2472 from asi1024/avoid-race-adv-indexing-test Sep 19, 2019
.coveragerc Fix .coveragerc Apr 27, 2018
.flake8.cython Ignore W503 and W504 Jan 14, 2019
.gitignore Add setup option to copy include files in wheel May 24, 2019
.travis.yml Drop Python 2 Travis CI configuration Aug 26, 2019 rename CoC Dec 19, 2018
LICENSE Split LICENSE file Jun 5, 2018 bundle CUDA fp16 headers Nov 27, 2018 add cupy-cuda101 (#2196) May 17, 2019
appveyor.yml refactor PlanNd so cimport of numpy is not required. Dec 18, 2018
codecov.yml add codecov configuration Mar 1, 2018 API signature revised; exposed cupy.cuda.texture Aug 26, 2019
readthedocs.yml Add requirements Oct 18, 2016
setup.cfg update to correct regexp Oct 29, 2018 Add cupy.misc to Sep 9, 2019

CuPy : NumPy-like API accelerated with CUDA

pypi GitHub license travis coveralls Read the Docs

Website | Docs | Install Guide | Tutorial | Examples (Official) | Forum (en, ja)

CuPy is an implementation of NumPy-compatible multi-dimensional array on CUDA. CuPy consists of the core multi-dimensional array class, cupy.ndarray, and many functions on it. It supports a subset of numpy.ndarray interface.


For detailed instructions on installing CuPy, see the installation guide.

You can install CuPy using pip:

(Binary Package for CUDA 8.0)
$ pip install cupy-cuda80

(Binary Package for CUDA 9.0)
$ pip install cupy-cuda90

(Binary Package for CUDA 9.1)
$ pip install cupy-cuda91

(Binary Package for CUDA 9.2)
$ pip install cupy-cuda92

(Binary Package for CUDA 10.0)
$ pip install cupy-cuda100

(Binary Package for CUDA 10.1)
$ pip install cupy-cuda101

(Source Package)
$ pip install cupy

The latest version of cuDNN and NCCL libraries are included in binary packages (wheels). For the source package, you will need to install cuDNN/NCCL before installing CuPy, if you want to use it.

Run with Docker

We provide the official Docker image. Use nvidia-docker command to run CuPy image with GPU. You can login to the environment with bash, and run the Python interpreter.

$ nvidia-docker run -it cupy/cupy /bin/bash


Please see the contribution guide.

More information


MIT License (see LICENSE file).

CuPy is designed based on NumPy's API and SciPy's API (see docs/LICENSE_THIRD_PARTY file).


Ryosuke Okuta, Yuya Unno, Daisuke Nishino, Shohei Hido and Crissman Loomis. CuPy: A NumPy-Compatible Library for NVIDIA GPU Calculations. Proceedings of Workshop on Machine Learning Systems (LearningSys) in The Thirty-first Annual Conference on Neural Information Processing Systems (NIPS), (2017). URL

  author       = "Okuta, Ryosuke and Unno, Yuya and Nishino, Daisuke and Hido, Shohei and Loomis, Crissman",
  title        = "CuPy: A NumPy-Compatible Library for NVIDIA GPU Calculations",
  booktitle    = "Proceedings of Workshop on Machine Learning Systems (LearningSys) in The Thirty-first Annual Conference on Neural Information Processing Systems (NIPS)",
  year         = "2017",
  url          = ""
You can’t perform that action at this time.