|
--- |
|
license: creativeml-openrail-m |
|
tags: |
|
- coreml |
|
- stable-diffusion |
|
- text-to-image |
|
--- |
|
## These are several flavors of the piece of the SDXL Refiner model that Mochi Diffusion needs to turn a SDXL Base model into a combined Base + Refiner model. |
|
|
|
Put the correct version of the UnetRefiner.mlmodelc in with the rest of the pieces of your Base model. Mochi will automatically recognize the 2 Unets as a combineed model. |
|
|
|
It will use the Base model Unet for the first 80% of the steps and then it will automatically switch to the Refiner model Unet for the final 20%. The transition point is hardcoded. |
|
|
|
Take care when unzipping after downloading as they all unzip to the same UnetRefiner.mlmodelc name. |
|
|
|
|
|
**Keep in mind that the Refiner is a generic type model and using it with a Base model that has a specific "personality" might ruin rather than improve images.** |