Google’s Datalab is a convenient tool to experiment with data and to build out machine learning models. As long as you don’t need to access any data that lives on the Google Cloud, such as Cloud Storage or BigQuery, you can run Datalab entirely local and without needing your Google Credentials on your Linux workstation or VM.
First, make sure your Linux distribution has a GUI installed because you’ll need to run Datalab in a web browser.
Second, Datalab runs in a Docker container, so you’ll have to install Docker. If you don’t have it already, follow the instructions for installing docker-ce on your flavour of Linux. It usually only takes a few commands to install.
Finally, download and install Datalab. Since docker commands should be run as root user, it’s probably easier to do all this from the root account.
$ sudo su - $ cd $ pwd /root
Finally, download Datalab from github:
git clone https://github.com/GoogleCloudPlatform/training-data-analyst.git
Then cd to training-data-analyst/datalab/local.
You should now be in the following directory: /root/training-data-analyst/datalab/local
Then start Datalab by running the start_datalab.sh script:
$ ./start_datalab.sh Defaulting to /root to store your notebooks latest: Pulling from cloud-datalab/datalab Digest: sha256:2a25fffa983e52e48c9f655182b2d695f76ed7952d3c37a26c8020d038cba10e Status: Image is up to date for gcr.io/cloud-datalab/datalab:latest Verifying that the /tmp directory is writable The /tmp directory is writable / / From https://github.com/googledatalab/notebooks * branch master -> FETCH_HEAD HEAD is now at f3130a8 Updating comments redux (#160) / Already on 'master' Your branch is up-to-date with 'origin/master'. Open your browser to http://localhost:8081/ to connect to Datalab.
You can now open your browser to http://localhost:8081/ and start creating your notebooks. By default, Datalab will create the notebooks in the following directory (if you’re running as root):
/root/datalab/notebooks