Typically, it is a good idea to unset the PYTHONPATH variable before installing Miniconda as it can cause problems if there are packages installed and accessible via directories included in the PYTHONPATH that are not compatible with the version of Python included with Miniconda.

!echo $PYTHONPATHĪt the time of writing this command returns only /env/python (oddly this is a directory that doesn’t seem to exist within Google Colab filesystem). The most recent version of Miniconda that targets Python 3.6 is Miniconda 4.5.4 so this is the version that you should install.įinally, check to see if the PYTHONPATH variable has been set. More recent versions (i.e., 4.5.12+) of Miniconda target either Python 3.7 or Python 3.8 by default. This means that, in order to use all of the preinstalled Google Colab packages, you will need to install a version of Miniconda that is compatible with Python 3.6 by default. !python -versionĪt time of writing the above command returns Python 3.6.9. Now check the version number of this default Python. !which python # should return /usr/local/bin/python Running the following command returns the absolute path to the default Python executable. Preliminariesįirst, you need to confirm which Python is being used by default in Google Colab. In this article, I will walk you through the process that I go through when I need to use Conda to install packages when working in Google Colab. Unfortunately, Conda is not available by default on Google Colab and getting Conda installed and working properly within Google Colab’s default Python environment is a bit of a chore. Google Colab is a great teaching platform and is also perhaps the only free solution available for sharing GPU or TPU accelerated code with your peers. Google Colab is a free service providing interactive computing resources via a user interface that is very similar to Jupyter notebooks, runs on Google Cloud Platform (GCP), and provides free access to GPUs and TPUs. Conda also dramatically simplifies the process of installing popular deep learning tools like TensorFlow. SourceĬonda is the recommended environment and package management solution for a number of popular data science tools including Pandas, Scikit-Learn, PyTorch, NVIDIA Rapids and many others. Getting Conda to work on Google Colab is a bit tedious but necessary if you can’t get by with pip.

Latest conda version