Fine Tuning Endpoints are Part of Gateway Implementation

Hey @elentaure. fine-tuning API on the Gateway currently is unified across OpenAI, AWS Bedrock, and Fireworks AI right now: https://portkey.ai/docs/api-reference/inference-api/fine-tuning/create-fine-tuning-job

It may not be a part of your Gateway deployment though.

Also check out this doc for more details on the data service that powers the finetuning API: https://github.com/Portkey-AI/helm-chart/blob/main/helm/enterprise/DataService.md

Hi @Vrushank | Portkey what I mean is, if we wanted to extend it to support vertex as well, is this something I can contriubute to directly with a PR or do I have to request it as a feature request?

I’ll loop in @Witherwings & @b4s36t4 for this. We have plans to merge this back into oss but they can share more on the timeline

cool, thanks

hey @elentaure. we are working on this. The plan is to make universal finetuning api, following api signature with what openai does, with minimal changes to other providers.
Here is how we will be implementing this

• The file should be openai finetuning format and on the fly we convert it to provider specific finetuning format.
• The finetuning jobs api will closely follow openai format and will support the following providers at the start (openai, azure, bedrock). with future support for fireworks

We would love if you can create a PR for Vertex, but I would suggest you to wait for some time, till we make the basic structure GA.

Sounds good, any ETA?

unfortunately not able to provide any ETA, but this is one of our highest priority atm. thats all I can say

@elentaure. in the meanwhile, we can take this up a feature request. shall we discuss more over Slack?

Sure, was just as an initial check on the status. Ill follow up in the slack channel