leon-se

/

gemma-3-27b-it-qat-W4A16-G128

like 9

just for my understanding.
i.e. you changed the model to a new data format, but kept the weights as they were already 4bit, so there are no performance changes, right?