Provide Lora Extraction Download Option - Save 83% Storage!

Please automatically process all pre-existing and newly uploaded checkpoint files on the website to perform lora extraction and offer this as a download option.

Why?

Because a lora extraction using rank 128 and fp16 on a SDXL model results in a 1GB Lora that can be used with the base SDXL model and will give literally almost exact results as if you just used the checkpoint. It even inherits features like LCM or Turbo. Reducing file size from 6GB to 1GB will massively save storage and bandwidth, incentivising users to switch over to SDXL.

Here are some comparisons:

Original Dreamshapers SDXL Turbo:

Lora Extraction 128-rank fp-16 (1GB):

The differences might aswell be seed variation and the fact that I didn't use an upscaler. For 83% bandwidth savings, easy model merging (as its infinitely easier and storage efficient to just prompt multiple loras than to merge checkpoints) I think this option should be offered for all SDXL checkpoints.