Adds CodeGemma for FIM Completion

twinnydotdev / twinny

The most no-nonsense, locally or API-hosted AI code completion plugin for Visual Studio Code - like GitHub Copilot but completely free and 100% private.

https://twinny.dev

MIT License

2.36k stars 130 forks source link

Adds CodeGemma for FIM Completion #205

Closed jeffistyping closed 3 months ago

jeffistyping commented 3 months ago

This PR adds FIM sentinels, stop words and other config to use CodeGemma 2B PT and Codegemma 7B PT for inline code completion. PR tested working with default server settings in LM Studio 0.2.13 (latest as of writing).

Download models:

CodeGemma 2B PT: https://huggingface.co/google/codegemma-2b
CodeGemma 7B PT: https://huggingface.co/google/codegemma-7b

Recommended Usage:

LM Studio (recommended), Ollama
repeat penalty = 1.0 is optimal for this model

rjmacarthy commented 3 months ago

Hey @jeffistyping thanks for this! I was not aware that there was a code version of Gemma model.

jeffistyping commented 3 months ago

we just released them today :)