Totoro LoRA text2image fine-tuning

These are LoRA adaption weights for nitrosocke/Ghibli-Diffusion.

pipe = StableDiffusionPipeline.from_pretrained(
  model_id,
  torch_dtype=torch.float16,
  safety_checker=None,
)
pipe.enable_xformers_memory_efficient_attention()
pipe.load_lora_weights(./lora_weights)
pipe = pipe.to('cuda')

image = pipe(
  prompt='Totoro',
  num_inference_steps=66,
  guidance_scale=6.9,
  cross_attention_kwargs={'scale': 0.6 },
).images[0]