Icon 
██████╗ ██╗   ██╗██████╗ ██╗  ██╗███████╗   ███████╗██╗   ██╗
██╔══██╗██║   ██║██╔══██╗██║ ██╔╝██╔════╝   ██╔════╝██║   ██║
██████╔╝██║   ██║██████╔╝█████╔╝ █████╗     ███████╗██║   ██║
██╔══██╗██║   ██║██╔══██╗██╔═██╗ ██╔══╝     ╚════██║██║   ██║
██████╔╝╚██████╔╝██║  ██║██║  ██╗███████╗██╗███████║╚██████╔╝
╚═════╝  ╚═════╝ ╚═╝  ╚═╝╚═╝  ╚═╝╚══════╝╚═╝╚══════╝ ╚═════╝

Share

A collection of neat things I've found or created.

  • devstral-small-2-jinja-fix.txt

    A fix for the default Jinja template in the lmstudio-community Devstral Small 2 repo

    I personally haven't used Jinja before - this was generated by Gemini to fix a template issue that prevented the LLM from properly being used by Roo code.
    I really like this model! I'll probably write more about it somewhere on this site when I have more experience, but if you want a good programming assistant that can run entirely in 24GB of VRAM with no CPU offload, this fits the bill. As a plus, it also supports image input, which I imagine is huge for really finetuning UIs. (frontend development is probably where LLMs shine the most as of now anyways)