Build Image

Enterprise Applicable to Enterprise tier only

This quickstart shows how to build a custom image by the feature of Image Builder. Here we are going to build an image with fastai v1 library for PyTorch provided by Fast.ai based on the Jupyter base-notebook. Fast.ai is a well-known for free online A.I. courses, please check its site if any interest.

Steps

1. Log in as an administrator and switch to Admin Portal.

2. Enter Image Builder management and click + Add for adding a custom image spec.

3. Fill Name with fastai-v1 (or other name you like).

4. Fill Base Image with jupyter/base-notebook.

5. Don't check off Use Image PullSecret, a pull-secret is unnecessary since it is a public repository.

6. Fill Conda field with -c pytorch -c fastai fastai [REF]

   (Due to the full command is conda install -c pytorch -c fastai fastai).

7. (Optional) Fill APT field with vim (or other tools).

   (Due to base image is Ubuntu base, APT is its package management tool.)

8. Click Confirm to save the custom image spec.

   

9. A job is created with the pending status until the image-build job is triggered.

10. Once the status of job turns into Succeeded, we can copy the url of image for adding the image later on Images management so that users are able to use this image with built-in fastai v1 library. (The url varies according to your circumstance.)

    

Alright, we have built our image successfully and gain the url of it. Go to [quickstart] add image to see how we add the image to be available for users on PrimeHub.

Next

When we add images or assign base images, pulling these images may require secrets to have the permission. Next, let's try to add these secrets on Secret Management.