search

dotnet / machinelearning

ML.NET is an open source and cross-platform machine learning framework for .NET.
https://dot.net/ml
MIT License
8.92k stars 1.86k forks source link

Add support/examples for Google Gemma lightweight LLM models #7026

Open abbottdev opened 4 months ago

abbottdev commented 4 months ago

Is your feature request related to a problem? Support complex LLM reasoning tasks with ML .NET. My specific use case here is structured data extraction into a known format.

Being able to take this models and fine tune them using the ML .NET training/pipeline APIs would be a huge benefit.

Describe the solution you'd like Google have opened sourced their base model Gemma which is a lightlight deriviative of the Gemini LLM. There is a keras implementation of it, a TensorFlow implementation. There are other options available but they havent been instruction tuned so the model would need more work for say, pytorch?

Additional context https://blog.google/technology/developers/gemma-open-models/ https://www.kaggle.com/models/google/gemma

TeddyAlbina commented 4 months ago

Basically AI is unusable outside of python. Models are almost never exported to Onnx because the AI guys live in a bubble and never make applications used by real people.