Installing PyTorch under conda fails with permissions error and Rolling back transaction

10,906

Solution 1

Try the following steps in Windows:

  1. Create a virtual environment using the command :

    conda create -n py_env python=3.5

  2. source activate py_env

  3. conda install pytorch-cpu -c pytorch

  4. pip install torchvision

Note: You can use any name instead of py_env

Thanks

Solution 2

What is your platform?

For your first installation method, the error message says that you don't have the permission. I encountered that error before on a Linux system. The reason was that Anaconda was installed by another user. I configured the path to point python to that installation so that I could run python without installing my own copy of Anaconda. However, it didn't permit me installing new packages and I got the same error message.

Solution: I installed my own copy of Anaconda and everything worked.

Solution 3

just run:

pip install torch torchvision
Share:
10,906
RussAbbott
Author by

RussAbbott

Updated on July 26, 2022

Comments

  • RussAbbott
    RussAbbott almost 2 years

    I'd like to use PyTorch in a Python program. The instructions for installing it require conda. After installing Conda I ran:

    >conda install -c pytorch pytorch (as instructed on the PyTorch [page][1])
    

    It looked promising -- until the end.

    Solving environment: done
    
    ## Package Plan ##
    
      environment location: C:\ProgramData\Miniconda3
    
      added / updated specs:
        - pytorch
    
    
    The following packages will be downloaded:
    
        package                    |            build
        ---------------------------|-----------------
        icc_rt-2017.0.4            |       h97af966_0         8.0 MB
        vs2015_runtime-15.5.2      |                3         2.2 MB
        pytorch-0.4.0              |py36_cuda80_cudnn7he774522_1       529.2 MB      pytorch
        mkl-2018.0.3               |                1       178.1 MB
        numpy-1.14.5               |   py36h9fa60d3_4          35 KB
        intel-openmp-2018.0.3      |                0         1.7 MB
        numpy-base-1.14.5          |   py36h5c71026_4         3.8 MB
        vc-14.1                    |       h0510ff6_3           5 KB
        blas-1.0                   |              mkl           6 KB
        conda-4.5.8                |           py36_0         1.0 MB
        mkl_fft-1.0.2              |   py36hb217b18_0         113 KB
        mkl_random-1.0.1           |   py36h77b88f5_1         268 KB
        ------------------------------------------------------------
                                               Total:       724.4 MB
    
    The following NEW packages will be INSTALLED:
    
        blas:           1.0-mkl
        icc_rt:         2017.0.4-h97af966_0
        intel-openmp:   2018.0.3-0
        mkl:            2018.0.3-1
        mkl_fft:        1.0.2-py36hb217b18_0
        mkl_random:     1.0.1-py36h77b88f5_1
        numpy:          1.14.5-py36h9fa60d3_4
        numpy-base:     1.14.5-py36h5c71026_4
        pytorch:        0.4.0-py36_cuda80_cudnn7he774522_1 pytorch
    
    The following packages will be UPDATED:
    
        conda:          4.5.4-py36_0                               --> 4.5.8-py36_0
        vc:             14-h0510ff6_3                              --> 14.1-h0510ff6_3
        vs2015_runtime: 14.0.25123-3                               --> 15.5.2-3
    
    Proceed ([y]/n)? y
    
    
    Downloading and Extracting Packages
    icc_rt-2017.0.4      |  8.0 MB | ############################################################################## | 100%
    vs2015_runtime-15.5. |  2.2 MB | ############################################################################## | 100%
    pytorch-0.4.0        | 529.2 MB | ############################################################################# | 100%
    mkl-2018.0.3         | 178.1 MB | ############################################################################# | 100%
    numpy-1.14.5         |   35 KB | ############################################################################## | 100%
    intel-openmp-2018.0. |  1.7 MB | ############################################################################## | 100%
    numpy-base-1.14.5    |  3.8 MB | ############################################################################## | 100%
    vc-14.1              |    5 KB | ############################################################################## | 100%
    blas-1.0             |    6 KB | ############################################################################## | 100%
    conda-4.5.8          |  1.0 MB | ############################################################################## | 100%
    mkl_fft-1.0.2        |  113 KB | ############################################################################## | 100%
    mkl_random-1.0.1     |  268 KB | ############################################################################## | 100%
    Preparing transaction: done
    Verifying transaction: done
    

    But then this.

    Executing transaction: failed
    ERROR conda.core.link:_execute(502): An error occurred while uninstalling package 'defaults::conda-4.5.4-py36_0'.
    PermissionError(13, 'Access is denied')
    Attempting to roll back.
    
    Rolling back transaction: done
    
    PermissionError(13, 'Access is denied')
    

    Apparently it was at least partly installed because PyCharm was able to see it. But when I asked PyCharm to install it in an environment, I got this error.

    RuntimeError: PyTorch does not currently provide packages for PyPI (see status at https://github.com/pytorch/pytorch/issues/566).
    
    Please follow the instructions at http://pytorch.org/ to install with miniconda instead.
    

    It suggests an alternative way to install PyTorch. So I tried that.

    >conda install pytorch torchvision -c pytorch
    Solving environment: failed
    
    PackagesNotFoundError: The following packages are not available from current channels:
    
      - torchvision
    
    Current channels:
    
      - https://conda.anaconda.org/pytorch/win-64
      - https://conda.anaconda.org/pytorch/noarch
      - https://repo.anaconda.com/pkgs/main/win-64
      - https://repo.anaconda.com/pkgs/main/noarch
      - https://repo.anaconda.com/pkgs/free/win-64
      - https://repo.anaconda.com/pkgs/free/noarch
      - https://repo.anaconda.com/pkgs/r/win-64
      - https://repo.anaconda.com/pkgs/r/noarch
      - https://repo.anaconda.com/pkgs/pro/win-64
      - https://repo.anaconda.com/pkgs/pro/noarch
      - https://repo.anaconda.com/pkgs/msys2/win-64
      - https://repo.anaconda.com/pkgs/msys2/noarch
    
    To search for alternate channels that may provide the conda package you're
    looking for, navigate to
    
        https://anaconda.org
    
    and use the search bar at the top of the page.
    

    But when I do that and search for PyTorch, I eventually find myself back at the original instructions.

    When I search for Torchvision, no Windows versions are listed.

  • OnePunchMan
    OnePunchMan about 5 years
    could you please tell how to install the pytorch-cpu and torchvision-cpu without using conda. I have created a virtual environment, I want to install it there. (CPU machine, Ubuntu 16.04, NO CUDA support). Thanks.