2 minute recipe: How to rollback your conda environment

I was working on a personal Machine learning project for which I was using Anaconda environment with python 2.7(yes I know) because of some dependencies.

I was trying to upgrade the sklearn library and I accidentally updated Python to 3.6. I was about to delete my whole environment and recreate it again from YAML file(thank god I had that), but then I referred conda documentation and there it was, the magic command to save me from all this trouble.

List the history of each change to the current environment     conda list --revisions
Restore environment to a previous revision                     conda install --revision 2

The best way to explain is by a quick example. If you run conda list –revisions, you’ll get an output like this:

2017-04-18 23:29:36  (rev 1)
     requests  {2.12.3 -> 2.13.0}

2018-05-30 19:41:47  (rev 2)
     mkl  {11.3.3 -> 2018.0.2}
     numpy  {1.11.2 -> 1.14.3}
     pip  {9.0.1 -> 10.0.1}
     python  {2.7.12 -> 3.6.5}
     scikit-learn  {0.17.1 -> 0.19.1}
     scipy  {0.18.1 -> 1.1.0}
     setuptools  {27.2.0 -> 39.1.0}
     wheel  {0.29.0 -> 0.31.1}
    +blas-1.0
    +certifi-2018.4.16
    +icc_rt-2017.0.4
    +intel-openmp-2018.0.0
    +mkl_fft-1.0.1
    +mkl_random-1.0.1
    +numpy-base-1.14.3
    +vc-14
    +vs2015_runtime-14.0.25123
    +wincertstore-0.2

As you can see in the output that it list each revision along with updated packages(old version -> new version) and newly added packages(the one with + symbol). So now you know what changes were make in each revision, you can safely rollback to the previous versions of your environment by using conda install –revision revision number.

In my case I reverted to rev 1, so after rollback when I run conda list –revisions again, I can see the changes done in rollback.

2018-05-30 20:08:46  (rev 3)
     mkl  {2018.0.2 -> 11.3.3}
     numpy  {1.14.3 -> 1.11.2}
     pip  {10.0.1 -> 9.0.1}
     python  {3.6.5 -> 2.7.12}
     scikit-learn  {0.19.1 -> 0.17.1}
     scipy  {1.1.0 -> 0.18.1}
     setuptools  {39.1.0 -> 27.2.0}
     wheel  {0.31.1 -> 0.29.0}
    -blas-1.0
    -certifi-2018.4.16
    -icc_rt-2017.0.4
    -intel-openmp-2018.0.0
    -mkl_fft-1.0.1
    -mkl_random-1.0.1
    -numpy-base-1.14.3
    -vc-14
    -vs2015_runtime-14.0.25123
    -wincertstore-0.2

You can see that the changes for revision 3 are just the inverse of revision 2.

This comes handy if you accidentally screw up your environment.

More info:

Conda User Guide: https://conda.io/docs/user-guide/tasks/manage-environments.html

Conda Cheat Sheet: https://conda.io/docs/_downloads/conda-cheatsheet.pdf

If you have any suggestions let me know on twitter: @Sriramjaju