app.py

import gradio as gr
import torch
import os
from diffusers import StableDiffusion3Pipeline
from safetensors.torch import load_file
from spaces import GPU  # Remove if not in HF Space

# ... (HF_TOKEN, model_id - same as before)

pipeline = None  # Global pipeline variable

# Load Stable Diffusion and LoRA *immediately* (before Gradio)
try:
    pipeline = StableDiffusion3Pipeline.from_pretrained(
        model_id,
        use_auth_token=hf_token,
        torch_dtype=torch.float16,
        cache_dir="./model_cache"
    )
    pipeline.enable_model_cpu_offload()
    pipeline.enable_attention_slicing()

    lora_filename = "lora_trained_model.safetensors"
    lora_path = os.path.join("./", lora_filename)
    print(f"Loading LoRA from: {lora_path}")

    if os.path.exists(lora_path):
        lora_weights = load_file(lora_path)
        text_encoder = pipeline.text_encoder
        text_encoder.load_state_dict(lora_weights, strict=False)
        print("LoRA loaded successfully!") # Confirmation message
    else:
        print(f"Error: LoRA file not found at {lora_path}")
        exit() # Exit if LoRA is not found


    print("Stable Diffusion model loaded successfully!")

except Exception as e:
    print(f"Error loading Stable Diffusion or LoRA: {e}")
    exit()  # Exit if there's an error


# Function for image generation (now much simpler)
@GPU(duration=65)  # Use GPU decorator (ONLY if in HF Space)
def generate_image(prompt):
    global pipeline
    if pipeline is None:  # This should never happen now
        return "Error: Stable Diffusion model not loaded!"

    try:
        image = pipeline(prompt).images[0]
        return image
    except Exception as e:
        return f"Error generating image: {e}"


# Create the Gradio interface (no "Load Model" button needed)
with gr.Blocks() as demo:
    prompt_input = gr.Textbox(label="Prompt")
    image_output = gr.Image(label="Generated Image")
    generate_button = gr.Button("Generate")

    generate_button.click(
        fn=generate_image,
        inputs=prompt_input,
        outputs=image_output,
    )

demo.launch()