kitops icon indicating copy to clipboard operation
kitops copied to clipboard

Add `readme.md` into the Kitfile / ModelKit

Open bmicklea opened this issue 1 year ago • 4 comments

Describe the problem your feature would solve It can sometimes be difficult to understand what a model is meant to be used for and what datasets are / why they were included.

Describe the solution you'd like A readme included in the ModelKit would make it easy for the Kitfile producer and consumer to understand the context for the ModelKit and the assets it is packaging.

Describe alternatives you've considered Descriptions in the Kitfile are okay, but are either very short and keep the Kitfile readable, or are very long and wreck the readability.

Additional context [none]

bmicklea avatar Mar 08 '24 17:03 bmicklea

I suggest we turn this into a convention rather than implementation any code section with path README.md that points to the root of the ModelKit should be considered the README. We should document this convention but as far as the implementation we do not need to do any additional work.

gorkem avatar May 13 '24 13:05 gorkem

We might want to consider codifying it more directly. I'm thinking we might want to treat them differently (e.g. store them in a more parseable format) and also ensure they come exported with every model (i.e. not lost when using kit unpack --model)

Maybe something like a modelpart with a specific type (readme)?

amisevsk avatar May 13 '24 17:05 amisevsk

Yes, that's what I was thinking too. Something so that a platform or tool could choose to "highlight" the readme separately - similar to what GitHub, Hugging Face, and others do today with readme.

bmicklea avatar May 13 '24 20:05 bmicklea

GitHub does not treat READMEs differently in means of how it stores them. The special meaning is just a convention.

Let's separate how a README is represented on Kitfile and how it is stored in a ModelKit.

IMHO we do not need a special construct on Kitfile to indicate something is a README. It is in the name already.

When storing as part of the ModelKit. I agree that we can make it easier for clients to render it. Browsers today already support Gzip encoding and the compression streams API can decompress gzip. I think we should do a PoC to see how easy it is to use them today before we make a change.

gorkem avatar May 13 '24 20:05 gorkem