Green-Sky/CyberRealisticPony-GGUF

This quant has https://huggingface.co/madebyollin/sdxl-vae-fp16-fix merged in.

The vae_f16 variant was quantized with f16 override.

--tensor-type-rules "^first_stage_model\.=f16"

Which makes the file a little smaller and looks almost identical.

use with sd.cpp like so:

$ bin/sd
-m CyberRealisticPony_V12.7-vae_f16-q8_0.gguf
--diffusion-fa
--clip-skip 2
--schedule karras
--sampling-method dpm++2m
-W 832 -H 1216
--cfg-scale 5
--steps 36
-n "score_6, score_5, score_4, (worst quality:1.2), (low quality:1.2), (normal quality:1.2), lowres, bad anatomy, bad hands, signature, watermarks, ugly, imperfect eyes, skewed eyes, unnatural face, unnatural body, error, extra limb, missing limbs"
-p "score_9, score_8_up, score_7_up, a lovely cat"