Modelpublic

Jayce-Ping/Pickscore-Teacher

sourceHugging Faceupdated 4mo agoView on Hugging Face

0likes15downloads

Model Card

Pickscore-Teacher

DPPO-trained LoRA teacher for PickScore prompts (aesthetics), trained from SD3.5 Medium via Flow-Factory.

Field	Value
Base model	`stabilityai/stable-diffusion-3.5-medium`
LoRA rank / alpha	32 / 64
Source checkpoint	`dppo_pickscore_teacher` step 1500
Training	Flow-DPPO, no-cfg (guidance_scale=1.0)