Interesting read. Thank you for providing examples.
If you can add example for text_encoder (NOT text_encoder_2) covering different like Qwen etc,, will be helpful.
e.g. Like this covers transformer
Bonus - saving and loading Diffusers models in Quanto
