app.py

import torch
from diffusers import StableDiffusion3Pipeline
from huggingface_hub import login
import os
import gradio as gr

# Retrieve Hugging Face token
token = os.getenv("HF_TOKEN")
if token:
    login(token=token)
else:
    raise ValueError("Hugging Face token not found. Please set it as a repository secret.")

# Load the Stable Diffusion 3.5 model
model_id = "stabilityai/stable-diffusion-3.5-large"
pipe = StableDiffusion3Pipeline.from_pretrained(
    model_id,
    torch_dtype=torch.float16,
    low_cpu_mem_usage=True,
    device_map="balanced"
)

# Enable attention slicing for reduced memory usage
pipe.enable_attention_slicing()

# Define the path to the LoRA model
lora_model_path = "./lora_model.pth"

# Load and apply the LoRA weights
def load_lora_model(pipe, lora_model_path):
    lora_weights = torch.load(lora_model_path, map_location="cuda")
    pipe.unet.load_attn_procs(lora_weights)
    return pipe

pipe = load_lora_model(pipe, lora_model_path)

# Generate image function
def generate_image(prompt, steps, scale):
    with torch.inference_mode():  # Avoid gradient computation for inference
        image = pipe(prompt, num_inference_steps=steps, guidance_scale=scale).images[0]
    return image

# Gradio interface
iface = gr.Interface(
    fn=generate_image,
    inputs=[
        gr.Textbox(label="Enter your prompt"),
        gr.Slider(10, 50, step=1, value=30, label="Number of Inference Steps"),
        gr.Slider(1.0, 20.0, step=0.5, value=7.5, label="Guidance Scale"),
    ],
    outputs="image",
    title="Optimized Stable Diffusion with LoRA",
    description="Generate images using Stable Diffusion 3.5 with optimized memory usage."
)

# Launch the Gradio interface
iface.launch()