SantaCoder: A new 1.1B code model for generation and infilling

Re:SantaCoder: A new 1.1B code model for generation and infi... by morgante on Thursday December 22, 2022 @05:25PM (#34099707)

We're building tools like this at Grit: https://www.grit.io/

These kinds of models are particularly good at repetitive, boring work like refactoring legacy code and completing framework migrations. Unlike Copilot, we've specialized specifically in these areas and completing them end-to-end (instead of just sitting in the IDE, we open already-verified PRs).

[ Parent | on HN ]

Re:SantaCoder: A new 1.1B code model for generation and infi... by ilaksh on Thursday December 22, 2022 @05:37PM (#34099840)

May I ask what model you are using?

[ Parent | on HN ]

Re:SantaCoder: A new 1.1B code model for generation and infi... by morgante on Thursday December 22, 2022 @05:54PM (#34100030)

We use a few depending on the task (Codex, fine-tuned T5, Bert models, etc.). Constantly experimenting with different variations. Since we focus on solving narrower problems in more depth, it leaves more room for optimizing accuracy.

[ Parent | on HN ]

Re:SantaCoder: A new 1.1B code model for generation and infi... by ilaksh on Thursday December 22, 2022 @06:24PM (#34100346)

Have you been able to get your rate limit increased for code-davinci-002? It defaults to a very small amount.

[ Parent | on HN ]

Re:SantaCoder: A new 1.1B code model for generation and infi... by morgante on Thursday December 22, 2022 @08:11PM (#34101162)

I recommend trying davinci-003 as well.

[ Parent | on HN ]

Re:SantaCoder: A new 1.1B code model for generation and infi... by ilaksh on Thursday December 22, 2022 @10:40PM (#34102119)

Yeah, been messing with that a lot.

[ Parent | on HN ]