Help me understand the difference btw Spaces and Gradient Notebooks

[russfischer]

a Gradient notebook is a provisioned Docker image, with Jupyter and specific runtimes installed for ML, is that right?

How does Spaces differ from that? Is a Space another type of provisioned Docker image with specific SDKs for building UIs? If I create a Space with a Docker SDK, I get a container that I can provision running “inside” a HuggingFace container?

I’m trying to build a mental map of these components so I can understand what I should use in different scenarios. Maybe if there was a feature comparison chart for Spaces and Gradient notebooks the lightbulb would turn on for me.

Thank you.

[radames]

Hi @russfischer,

@russfischer

A Gradient notebook is a provisioned Docker image, with Jupyter and specific runtimes installed for ML, is that right?

I’m not familiar with Gradient notebook, but it seems to be a service designed for Jupyter notebooks as you mentioned.

@russfischer

How does Spaces differ from that? Is a Space another type of provisioned Docker image with specific SDKs for building UIs? If I create a Space with a Docker SDK, do I get a container that I can provision running “inside” a HuggingFace container?

Spaces provides a simple way for ML practitioners and developers to build, share, and deploy machine learning-powered demo applications. We offer a few ready-to-run SDKs for static pages, Gradio and Streamlit apps, which use a Docker image under the hood. We also offer support for a Docker SDK, giving users complete control over building an app with a custom Dockerfile. You can read more about it here:

Spaces

Please let us know if this is helpful.