Request for a feature: Tracking environment for jupyter notebook using datmo about datmo HOT 2 CLOSED

Yes, environment reproducibility is one of the core features of datmo (even for Jupyter notebooks)!

In the case of VM's, here's how you'd share your jupyter notebook environment:

Initial Setup:

• Define the environment that you want to be duplicated in a Dockerfile. We have directions for this in our documentation here.
• Alternatively, we have an environment setup wizard for many common machine learning and AI environments available through the $ datmo environment setup command as well. If you'd like to use one of these, you can follow [this tutorial] for getting your Dockerfile through the wizard. I'd recommend this path, and modifying the existing Dockerfiles we provide with any additional libraries your environment needs.
• Check your Dockerfile into source control and push it to your project repository that others will be pulling from

To reproduce the environment:

• Spin up the VM as usual
• Follow the installation steps for datmo
  • Install datmo prereqs in the VM
  • Install and launch Docker
  • (If using GPU) Install CUDA 9.0
  • (If using GPU) Install nvidia-docker (Step 3)
  • Install the datmo python package inside the VM with $ pip install datmo
• In the directory for your repository inside the VM, run $ datmo init
• Run datmo notebook and it will open a Jupyter notebook inside the VM with the environment you outlined in the Dockerfile.
• Note: if you are using a Dockerfile that you wrote from scratch, you will need to add the code block found in this section of the docs under 5. Getting datmo workspaces to work with your custom environment to enable jupyter notebook to launch with the datmo notebook alias.

If you'd like more assistance in getting this setup, feel free to reach out, always happy to help!

Cheers

from datmo.

Thanks @nmwalsh

from datmo.