Unable to install CuPy 12.2.0 on Google Colab

Question

I want to use CuPy 12.2.0 on Colab, but the default CuPy version installed on Colab is 11.0.0. I tried several ways to install version 12.2.0, all of which failed.

1. When I tried to install cupy with the command, pip install cupy-cuda12x, and import cupy as import cupy, I received the following message.

/usr/local/lib/python3.10/dist-packages/cupy/_environment.py:447: UserWarning: 
--------------------------------------------------------------------------------

  CuPy may not function correctly because multiple CuPy packages are installed
  in your environment:

    cupy-cuda11x, cupy-cuda12x

  Follow these steps to resolve this issue:

    1. For all packages listed above, run the following command to remove all
       existing CuPy installations:

         $ pip uninstall <package_name>

      If you previously installed CuPy via conda, also run the following:

         $ conda uninstall cupy

    2. Install the appropriate CuPy package.
       Refer to the Installation Guide for detailed instructions.

         https://docs.cupy.dev/en/stable/install.html

--------------------------------------------------------------------------------

  warnings.warn(f'''
---------------------------------------------------------------------------
ImportError                               Traceback (most recent call last)
/usr/local/lib/python3.10/dist-packages/cupy/__init__.py in <module>
     16 try:
---> 17     from cupy import _core  # NOQA
     18 except ImportError as exc:

/usr/local/lib/python3.10/dist-packages/cupy/_core/__init__.py in <module>
      2 
----> 3 from cupy._core import core  # NOQA
      4 from cupy._core import fusion  # NOQA

ImportError: libcudart.so.12: cannot open shared object file: No such file or directory

The above exception was the direct cause of the following exception:

ImportError                               Traceback (most recent call last)
<ipython-input-2-329ec5cf1bc8> in <cell line: 1>()
----> 1 import cupy

/usr/local/lib/python3.10/dist-packages/cupy/__init__.py in <module>
     17     from cupy import _core  # NOQA
     18 except ImportError as exc:
---> 19     raise ImportError(f'''
     20 ================================================================
     21 {_environment._diagnose_import_error()}

ImportError: 
================================================================
Failed to import CuPy.

If you installed CuPy via wheels (cupy-cudaXXX or cupy-rocm-X-X), make sure that the package matches with the version of CUDA or ROCm installed.

On Linux, you may need to set LD_LIBRARY_PATH environment variable depending on how you installed CUDA/ROCm.
On Windows, try setting CUDA_PATH environment variable.

Check the Installation Guide for details:
  https://docs.cupy.dev/en/latest/install.html

Original error:
  ImportError: libcudart.so.12: cannot open shared object file: No such file or directory
================================================================


---------------------------------------------------------------------------
NOTE: If your import is failing due to a missing package, you can
manually install dependencies using either !pip or !apt.

To view examples of installing some common dependencies, click the
"Open Examples" button below.
---------------------------------------------------------------------------

It seemed I need to uninstall cupy-cuda11x, so I uninstalled it as pip uninstall cupy-cuda11x. Then I tried to import cupy again as

import cupy
cupy.arange(3)

I received a different error message,

AttributeError                            Traceback (most recent call last)
<ipython-input-8-0a93c279befd> in <cell line: 2>()
      1 import cupy
----> 2 cupy.arange(3)

AttributeError: module 'cupy' has no attribute 'arange'

which is like, the cupy module is empty.

As for the CUDA compatibility, the CUDA version of Colab is 11.8, which is compatible with CuPy 12.2.

>>nvcc --version
nvcc: NVIDIA (R) Cuda compiler driver
Copyright (c) 2005-2022 NVIDIA Corporation
Built on Wed_Sep_21_10:33:58_PDT_2022
Cuda compilation tools, release 11.8, V11.8.89
Build cuda_11.8.r11.8/compiler.31833905_0

---

2. When I tried to install cupy with the command, pip install cupy, I received the following error message.

Building wheels for collected packages: cupy
  error: subprocess-exited-with-error
  
  × python setup.py bdist_wheel did not run successfully.
  │ exit code: 1
  ╰─> See above for output.
  
  note: This error originates from a subprocess, and is likely not a problem with pip.
  Building wheel for cupy (setup.py) ... error
  ERROR: Failed building wheel for cupy
  Running setup.py clean for cupy
Failed to build cupy
ERROR: Could not build wheels for cupy, which is required to install pyproject.toml-based projects

At this point, I had no idea how I could fix it.

---

3. When I tried to install cupy with the following as recommended,

pip install chainer
curl https://colab.chainer.org/install | sh -

I just received the following error message and I gave up. The runtime type was selected as T4 GPU.

  % Total    % Received % Xferd  Average Speed   Time    Time     Time  Current
                                 Dload  Upload   Total   Spent    Left  Speed
100  1553  100  1553    0     0  12757      0 --:--:-- --:--:-- --:--:-- 12834
********************************************************************************
CUDA version could not be detected!
Try installing Chainer manually by: pip install chainer
********************************************************************************

Answer 1

CuPy is installed on Google Colab by default, so you can just import cupy. If you prefer the latest version of CuPy, try pip install -U cupy-cuda11x.