Trained on an RTX 3090, purely on chat data, as a test to see how small you could make LLMs before they can't chat anymore.
Not very useful, but good if you need some cheap chat embeddings.
-
