models/web_interface.py

import gradio as gr
import torch
from diffusers import StableDiffusionPipeline
from peft import PeftModel
import os
from PIL import Image
import random

class LoRAWebInterface:
    def __init__(self, base_model="runwayml/stable-diffusion-v1-5", lora_path="models/lora_model"):
        self.device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
        self.lora_path = lora_path
        
        print("Loading models...")
        # Load base pipeline
        self.pipeline = StableDiffusionPipeline.from_pretrained(
            base_model,
            torch_dtype=torch.float16 if torch.cuda.is_available() else torch.float32,
            safety_checker=None,
            requires_safety_checker=False
        )
        
        # Load LoRA weights if they exist
        if os.path.exists(lora_path):
            print(f"Loading LoRA model from {lora_path}")
            try:
                self.pipeline.unet = PeftModel.from_pretrained(self.pipeline.unet, lora_path)
                self.lora_loaded = True
            except Exception as e:
                print(f"Error loading LoRA: {e}")
                self.lora_loaded = False
        else:
            print("No LoRA model found, using base model")
            self.lora_loaded = False
        
        self.pipeline.to(self.device)
        
        # Enable memory efficient attention
        try:
            self.pipeline.enable_xformers_memory_efficient_attention()
        except:
            pass
        
        print("Model loaded successfully!")
    
    def generate_image(self, prompt, negative_prompt, num_steps, guidance_scale, 
                      width, height, seed, use_random_seed):
        """Generate image with given parameters"""
        
        if use_random_seed:
            seed = random.randint(0, 999999)
        
        if seed is not None and seed >= 0:
            torch.manual_seed(int(seed))
        
        try:
            with torch.autocast(self.device.type):
                image = self.pipeline(
                    prompt=prompt,
                    negative_prompt=negative_prompt,
                    num_inference_steps=int(num_steps),
                    guidance_scale=guidance_scale,
                    width=int(width),
                    height=int(height)
                ).images[0]
            
            return image, f"✅ Generated successfully! Seed: {seed}"
            
        except Exception as e:
            error_msg = f"❌ Error generating image: {str(e)}"
            print(error_msg)
            # Return a blank image on error
            blank_image = Image.new('RGB', (512, 512), color='white')
            return blank_image, error_msg
    
    def create_interface(self):
        """Create Gradio interface"""
        
        with gr.Blocks(title="LoRA Image Generator", theme=gr.themes.Soft()) as interface:
            
            gr.Markdown("# 🎨 LoRA Image Generator")
            gr.Markdown(f"**Model Status:** {'✅ LoRA model loaded' if self.lora_loaded else '⚠️ Using base model only'}")
            
            with gr.Row():
                with gr.Column(scale=1):
                    # Input controls
                    prompt = gr.Textbox(
                        label="Prompt",
                        placeholder="Describe the image you want to generate...",
                        value="a beautiful artistic composition",
                        lines=3
                    )
                    
                    negative_prompt = gr.Textbox(
                        label="Negative Prompt (Optional)",
                        placeholder="Things you don't want in the image...",
                        value="blurry, low quality, distorted",
                        lines=2
                    )
                    
                    with gr.Row():
                        num_steps = gr.Slider(
                            minimum=10,
                            maximum=100,
                            value=50,
                            step=5,
                            label="Inference Steps"
                        )
                        
                        guidance_scale = gr.Slider(
                            minimum=1.0,
                            maximum=20.0,
                            value=7.5,
                            step=0.5,
                            label="Guidance Scale"
                        )
                    
                    with gr.Row():
                        width = gr.Slider(
                            minimum=256,
                            maximum=1024,
                            value=512,
                            step=64,
                            label="Width"
                        )
                        
                        height = gr.Slider(
                            minimum=256,
                            maximum=1024,
                            value=512,
                            step=64,
                            label="Height"
                        )
                    
                    with gr.Row():
                        seed = gr.Number(
                            label="Seed (-1 for random)",
                            value=-1,
                            precision=0
                        )
                        
                        use_random_seed = gr.Checkbox(
                            label="Use Random Seed",
                            value=True
                        )
                    
                    generate_btn = gr.Button("🎨 Generate Image", variant="primary")
                
                with gr.Column(scale=1):
                    # Output
                    output_image = gr.Image(
                        label="Generated Image",
                        type="pil",
                        height=512
                    )
                    
                    status_text = gr.Textbox(
                        label="Status",
                        interactive=False,
                        lines=2
                    )
            
            # Example prompts
            gr.Markdown("## 💡 Example Prompts")
            example_prompts = [
                "a serene landscape in artistic style",
                "abstract flowing patterns with vibrant colors",
                "geometric composition with soft lighting",
                "organic forms inspired by nature",
                "minimalist design with elegant curves"
            ]
            
            examples = gr.Examples(
                examples=[[prompt] for prompt in example_prompts],
                inputs=[prompt],
                label="Click an example to try:"
            )
            
            # Event handlers
            generate_btn.click(
                fn=self.generate_image,
                inputs=[prompt, negative_prompt, num_steps, guidance_scale, 
                       width, height, seed, use_random_seed],
                outputs=[output_image, status_text]
            )
            
            # Auto-disable seed input when random is selected
            use_random_seed.change(
                fn=lambda x: gr.update(interactive=not x),
                inputs=[use_random_seed],
                outputs=[seed]
            )
        
        return interface
    
    def launch(self, share=False, server_port=7860):
        """Launch the interface"""
        interface = self.create_interface()
        interface.launch(
            share=share,
            server_port=server_port,
            inbrowser=True
        )

def main():
    import argparse
    parser = argparse.ArgumentParser()
    parser.add_argument("--lora_path", default="models/lora_model", help="Path to LoRA model")
    parser.add_argument("--share", action="store_true", help="Create public link")
    parser.add_argument("--port", type=int, default=7860, help="Server port")
    
    args = parser.parse_args()
    
    # Create and launch interface
    interface = LoRAWebInterface(lora_path=args.lora_path)
    interface.launch(share=args.share, server_port=args.port)

if __name__ == "__main__":
    main()