app.py

import gradio as gr
import torch
import numpy as np
import modin.pandas as pd
from PIL import Image
from diffusers import StableDiffusion3Pipeline #DiffusionPipeline #, StableDiffusion3Pipeline
from huggingface_hub import hf_hub_download
from diffusers import BitsAndBytesConfig, SD3Transformer2DModel

device = 'cuda' if torch.cuda.is_available() else 'cpu'
torch.cuda.max_memory_allocated(device=device)
torch.cuda.empty_cache() 

model_id = "stabilityai/stable-diffusion-3.5-large-turbo"

nf4_config = BitsAndBytesConfig(
    load_in_4bit=True,
    bnb_4bit_quant_type="nf4",
    bnb_4bit_compute_dtype=torch.bfloat16
)
model_nf4 = SD3Transformer2DModel.from_pretrained(
    model_id,
    subfolder="transformer",
    quantization_config=nf4_config,
    torch_dtype=torch.bfloat16
)

t5_nf4 = T5EncoderModel.from_pretrained("diffusers/t5-nf4", torch_dtype=torch.bfloat16)

pipeline = StableDiffusion3Pipeline.from_pretrained(
    model_id, 
    transformer=model_nf4,
    text_encoder_3=t5_nf4,
    torch_dtype=torch.bfloat16
)
pipeline.enable_model_cpu_offload()

def genie (Prompt, height, width, seed):
    generator = np.random.seed(0) if seed == 0 else torch.manual_seed(seed)
       
    image = pipeline(Prompt, num_inference_steps=4, height=height, width=width, guidance_scale=0.0,).images[0]

    return image
    
gr.Interface(fn=genie, inputs=[#gr.Radio(['PhotoReal', 'Animagine XL 4',], value='PhotoReal', label='Choose Model'),
                               gr.Textbox(label='What you want the AI to generate. 77 Token Limit.'), 
                               #gr.Textbox(label='What you Do Not want the AI to generate. 77 Token Limit'),
                               gr.Slider(512, 1024, 768, step=128, label='Height'),
                               gr.Slider(512, 1024, 768, step=128, label='Width'),
                               #gr.Slider(3, maximum=12, value=5, step=.25, label='Guidance Scale', info="5-7 for PhotoReal and 7-10 for Animagine"), 
                               #gr.Slider(25, maximum=50, value=25, step=25, label='Number of Iterations'), 
                               gr.Slider(minimum=0, step=1, maximum=9999999999999999, randomize=True, label='Seed: 0 is Random'), 
                               ],
             outputs=gr.Image(label='Generated Image'), 
             title="Manju Dream Booth V2.5 - GPU", 
             description="<br><br><b/>Warning: This Demo is capable of producing NSFW content.", 
             article = "If You Enjoyed this Demo and would like to Donate, you can send any amount to any of these Wallets. <br><br>SHIB (BEP20): 0xbE8f2f3B71DFEB84E5F7E3aae1909d60658aB891 <br>PayPal: https://www.paypal.me/ManjushriBodhisattva <br>ETH: 0xbE8f2f3B71DFEB84E5F7E3aae1909d60658aB891 <br>DOGE: DL5qRkGCzB2ENBKfEhHarvKm1qas3wyHx7<br><br>Code Monkey: <a href=\"https://huggingface.co/Manjushri\">Manjushri</a>").launch(debug=True, max_threads=80)