app.py

import gradio as gr
import torch
from huggingface_hub import hf_hub_download
from safetensors.torch import load_file
import spaces
from PIL import Image
from diffusers import StableDiffusionPipeline, UniPCMultistepScheduler

device = "cuda" if torch.cuda.is_available() else "cpu"
device = "cpu"
print(device)
# Set seed for reproducibility
#seed = 42
#generator = torch.Generator(device=device)
#generator.manual_seed(seed)
#torch.manual_seed(42)

negative_prompt = "deformed face, extra limbs, extra fingers, out of frame"

# Constants
base = 'stabilityai/stable-diffusion-2'
checkpoints = {
    "Mambo Museum" : "adapters/mambo",
    "Street Art Museum" : "adapters/street",
    "Colonial Museum" : "adapters/colonial",
    "Oro Museum" : "adapters/oro",
}
loaded = None

# Ensure model and scheduler are initialized in GPU-enabled function
if torch.cuda.is_available():
    pipe = StableDiffusionPipeline.from_pretrained(base, torch_dtype=torch.float16, variant="fp16").to("cuda")
    print('initialized cuda model')
else:
    pipe = StableDiffusionPipeline.from_pretrained(base, torch_dtype=torch.float16, variant="fp16")
    
# Function 
@spaces.GPU(enable_queue=True)
def generate_image(prompt, ckpt):
    global loaded
    print(prompt, ckpt)

    checkpoint = checkpoints[ckpt]
    num_inference_steps = 50

    if loaded != num_inference_steps:
        #pipe.scheduler = UniPCMultistepScheduler.from_config(pipe.scheduler.config, timestep_spacing="trailing", prediction_type="sample" if num_inference_steps==1 else "epsilon")
        pipe.unet.load_attn_procs(checkpoint)
        loaded = num_inference_steps
        
    results = pipe(prompt, num_inference_steps=num_inference_steps, guidance_scale=7.5, negative_prompt=negative_prompt)#, generator=generator)

    return results.images[0]



# Gradio Interface

with gr.Blocks(css="style.css") as demo:
    gr.HTML("<h1><center>Mambo Museum Workshop ⚡</center></h1>")
    gr.HTML("<p><center>LoRAs to generate different aesthetics</center></p>")
    with gr.Group():
        with gr.Row():
            prompt = gr.Textbox(label='Enter your prompt (English)', scale=8)
            ckpt = gr.Dropdown(label='Select LoRA model',choices=['Mambo Museum', 'Street Art Museum', 'Colonial Museum', 'Oro Museum'], value='Mambo Museum', interactive=True)
            submit = gr.Button(scale=1, variant='primary')
    img = gr.Image(label='SGenerated Image')

    prompt.submit(fn=generate_image,
                 inputs=[prompt, ckpt],
                 outputs=img,
                 )
    submit.click(fn=generate_image,
                 inputs=[prompt, ckpt],
                 outputs=img,
                 )
    
demo.queue().launch()