Sonoma 1.2B
The 0901 build, plus a GGUF for running it locally. A 1.2B fine-tune that taught me more than it learned.
Hugging Face
They did not go well. They are on Hugging Face anyway, for science.
What it does
They did not turn out great, but the weights are public, so you can download them and see for yourself.
Start from a small open model that can run on normal hardware.
Train it further on my own data to give it a new style and skills.
Export the weights, plus a GGUF build of Sonoma for running it locally.
Upload to Hugging Face with an honest model card.