Diffusion64 (SDXL): v1.0: diffusion64-merge-v1.safetensors

A merge of two separate training runs, one ~23k steps with 4 gradient accumulation steps, one 51k steps at batch size 1 with different lr and scheduler.

A Stable Diffusion XL LoRA that generates Ocarina of Time / Majora's Mask style images. Training set is 1,500+ hand-tagged screenshots.

Uses sd_xl_base_1.0_0.9vae as base model.