Trained for 2000 steps on NVIDIA RTX 4090 (24GB VRAM), Learning Rate 2e-4
You should use Birkenstock to trigger the image generation
Birkenstock
Base model