In many projects there are many custom python requirements that are required for the different python files to be able to run successfully.
Once you know your requirements for your project, you don't need for the Notebook or Experiment to install those requirements everytime you want to spin one of them in cnvrg.
Users can now build docker images directly from requirements.txt files, and choose them as their image for running notebooks and experiments.

Build Your Custom Image

  • Go to your project
  • Click on Files  Tab
  • Click on your requirements.txt
  • Click on Build Image 
  • Set Image name, Image tag
  • Choose the base image to extend
  • Choose the compute to build the image on ( for gpu images choose compute based GPU)
  • Click On Create
  • Once the image is done building, you'll receive an email.

When you're image is ready you'll be able to run your notebooks and experiments with the newly created image!

Did this answer your question?