sd-simpsons-model

Simpsons diffusion

*Stable Diffusion fine tuned on images related to "The Simpsons"

If you want more details on how to generate your own blip cpationed dataset see this colab

Training was done using a slightly modified version of Hugging-Face's text to image training example script

About

Put in a text prompt and generate cartoony/simpsony images

A beautiful hungry demon girl, John Philip Falter, Very detailed painting, Mark Ryden

Gal Gadot, cartoon

More examples

The examples folder contains a few images generated by this model's ckpt file using stable-diffusion-webui which means their EXIF info contain the parameter used to generate them

Sample code

from diffusers import StableDiffusionPipeline, LMSDiscreteScheduler
import torch

# this will substitute the default PNDM scheduler for K-LMS  
lms = LMSDiscreteScheduler(
    beta_start=0.00085, 
    beta_end=0.012, 
    beta_schedule="scaled_linear"
)

guidance_scale=9
seed=7777
steps=100

model_id = "Norod78/sd-simpsons-model"
pipe = StableDiffusionPipeline.from_pretrained(model_id, scheduler=lms, torch_dtype=torch.float16)
pipe.to("cuda")

prompt = "a photo of an astronaut riding a horse on mars"
image = pipe(prompt).images[0]  
    
image.save("astronaut_rides_horse.png")

Dataset and Training

Finetuned for 10,000 iterations upon Runway ML's Stable-Diffusion v1.5 on BLIP captioned Simpsons images using 1xA5000 GPU on my home desktop computer

Trained by @Norod78