Hi folks, I made a research tools that allows you to perform deterministic inference on any local large language model. This way you can test any variable changes and see for yourself the affects those changes have on the output of the LLM's response. It also allows you to perform automated reasoning benchmarking of a local language model of your choice, this way you can measure the perplexity drop of any quantized model or differences between reasoning capabilities of models or sampling parameters. It also has a fully automated way of converging on the best sampling parameters for a given model when it comes to reasoning capabilities. I made 2 videos for the project so you can see what its about at a glance the main guide is here
https://www.youtube.com/watch?v=EyE5BrUut2o, the instillation video is here
https://youtu.be/FJpmD3b2aps and the repo is here
https://github.com/manfrom83/Sample-Forge. If you have more questions id be glad to answer them here. Cheers.