This is a great project. I'm curious to see how these phase transitions differ between quantization and model type.
I'd imagine that different models would have different 'temperatures' for their phase states. I imagine this is compounded by the noise introduced via quantization and even more so when discussing different transformer types.
