TabbyML
Models per language
It'd be cool to be able to specify different models per language (or other heuristics).
For example, https://huggingface.co/Tesslate/Tessa-Rust-T1-7B-Q8_0-GGUF is trained specifically for Rust, but in my project I use Rust + TypeScript; currently I'd lose performance if I switch to a specialized model for this project. With this feature implemented, however, Tabby would switch back and forth between the Rust model and, say, another general-purpose, "fallback" model.
Hello @vladfaust, thank you for the suggestion. However, it appears that there are only a few models trained for specific languages. Are you aware of any additional models that are trained for particular languages?
As it stands, most models are capable of working with multiple languages.
Could you share your experiences regarding the performance of the Tessellate Rust model?
Actually I have the same query. Currently I'm using DeepSeek Coder for my Flutter projects which is well and good but it can perform better if I can use Flutter Specific model like https://huggingface.co/pavansn2000/ollama3.2_flutter
