app.py

import spaces
import gradio as gr
import torch
from diffusers import StableDiffusion3Pipeline
import os
from huggingface_hub import snapshot_download

# Retrieve the API token from the environment variable
huggingface_token = os.getenv("HUGGINGFACE_TOKEN")
if huggingface_token is None:
    raise ValueError("HUGGINGFACE_TOKEN environment variable is not set.")

# Check if CUDA is available
device = "cuda" if torch.cuda.is_available() else "cpu"

# Ensure GPU is available
if device == "cuda":
    print("CUDA is available. Using GPU.")
else:
    print("CUDA is not available. Using CPU.")

model_path = snapshot_download(
    repo_id="stabilityai/stable-diffusion-3-medium", 
    revision="refs/pr/26",
    repo_type="model", 
    ignore_patterns=["*.md", "*..gitattributes"],
    local_dir="stable-diffusion-3-medium",
    token=huggingface_token, 
    )
# Load the Stable Diffusion model
repo = "stabilityai/stable-diffusion-3-medium-diffusers"
#image_gen = StableDiffusion3Pipeline.from_pretrained(repo, text_encoder_3=None, tokenizer_3=None, use_auth_token=huggingface_token)
image_gen = StableDiffusion3Pipeline.from_pretrained(model_path, text_encoder_3=None, tokenizer_3=None,torch_dtype=torch.float16)

#pipe = StableDiffusion3Pipeline.from_pretrained(model_path, torch_dtype=torch.float16)

image_gen = image_gen.to(device)

@spaces.GPU(enable_queue=True)
def generate_image(prompt, num_inference_steps=50, guidance_scale=7.5):
    # Generate the image
    result = image_gen(
        prompt=prompt,
        num_inference_steps=num_inference_steps,
        guidance_scale=guidance_scale,
        negative_prompt="blurred, ugly, watermark, low resolution, blurry",
        height=512,
        width=512
    )
    # Get the generated image
    image = result.images[0]
    return image

# Create the Gradio interface
iface = gr.Interface(
    fn=generate_image,
    inputs=[
        gr.Textbox(label="Enter a prompt"),
        #gr.Slider(label="Number of inference steps", minimum=1, maximum=100, value=50, step=1, precision=0),
        gr.Slider(label="Number of inference steps", minimum=1, maximum=100, value=50, step=1),

        gr.Slider(label="Guidance scale", minimum=1.0, maximum=20.0, value=7.5)
    ],
    outputs=gr.Image(label="Generated Image"),
    title="Stable Diffusion Image Generator",
    description="Enter a prompt to generate an image using the Stable Diffusion model."
)

# Launch the Gradio app
iface.launch()