OSError - Errno 13 Permission denied

Another option is to ensure the file is not open anywhere else on your machine.


Probably you are facing problem when a download request is made by the maybe_download function call in base.py file.

There is a conflict in the permissions of the temporary files and I myself couldn't work out a way to change the permissions, but was able to work around the problem.

Do the following...

  • Download the four .gz files of the MNIST data set from the link ( http://yann.lecun.com/exdb/mnist/ )
  • Then make a folder names MNIST_data (or your choice in your working directory/ site packages folder in the tensorflow\examples folder).
  • Directly copy paste the files into the folder.
  • Copy the address of the folder (it probably will be ( C:\Python\Python35\Lib\site-packages\tensorflow\examples\tutorials\mnist\MNIST_data ))
  • Change the "\" to "/" as "\" is used for escape characters, to access the folder locations.
  • Lastly, if you are following the tutorials, your call function would be ( mnist = input_data.read_data_sets("MNIST_data/", one_hot=True) ) ; change the "MNIST_data/" parameter to your folder location. As in my case would be ( mnist = input_data.read_data_sets("C:/Python/Python35/Lib/site-packages/tensorflow/examples/tutorials/mnist/MNIST_data", one_hot=True) )

Then it's all done. Hope it works for you.


You need to change the directory permission so that web server process can change the directory.

  • To change ownership of the directory, use chown:

    chown -R user-id:group-id /path/to/the/directory
    
  • To see which user own the web server process (change httpd accordingly):

    ps aux | grep httpd | grep -v grep
    

    OR

    ps -efl | grep httpd | grep -v grep
    

This may also happen if you have a slash before the folder name:

path = '/folder1/folder2'

OSError: [Errno 13] Permission denied: '/folder1'

comes up with an error but this one works fine:

path = 'folder1/folder2'