app.py

import gradio as gr
from PIL import Image
import torch
from inference.model_loader import load_model_and_tokenizer
import json

# Check device
device = "cuda" if torch.cuda.is_available() else "cpu"
print(f"🖥️  Using device: {device}")

# Load model
print("📥 Loading model from Hugging Face...")
model, processor = load_model_and_tokenizer(use_lora=True)
print("✅ Model loaded successfully!")

def extract_pokemon_card_info(image, max_tokens=256, temperature=0.7, top_p=0.9):
    """
    Extract card information from Pokémon card image
    
    Args:
        image: PIL Image
        max_tokens: Maximum number of tokens to generate
        temperature: Sampling temperature
        top_p: Nucleus sampling parameter
        
    Returns:
        raw_output: Raw text output from model
        json_output: Parsed JSON output (if available)
    """
    try:
        if image is None:
            return "⚠️ Please upload an image first!", ""
        
        # Prepare instruction
        instruction = "You are an OCR expert specialized in Pokémon cards. Extract the card name and card number in JSON format."
        
        # Prepare conversation
        conversation = [
            {
                "role": "user",
                "content": [
                    {"type": "text", "text": instruction},
                    {"type": "image"},
                ],
            },
        ]
        
        # Apply chat template
        prompt = processor.apply_chat_template(conversation, add_generation_prompt=True)
        
        # Process inputs
        inputs = processor(images=image, text=prompt, return_tensors="pt").to(device)
        
        # Generate output
        with torch.no_grad():
            outputs = model.generate(
                **inputs,
                max_new_tokens=max_tokens,
                temperature=temperature,
                top_p=top_p,
                do_sample=True if temperature > 0 else False,
            )
        
        # Decode output
        decoded_output = processor.decode(outputs[0], skip_special_tokens=True)
        
        # Try to parse as JSON for better display
        json_output = ""
        try:
            # Extract JSON from output
            json_start = decoded_output.find('{')
            json_end = decoded_output.rfind('}') + 1
            if json_start >= 0 and json_end > json_start:
                json_str = decoded_output[json_start:json_end]
                result_json = json.loads(json_str)
                json_output = json.dumps(result_json, indent=2, ensure_ascii=False)
        except Exception as json_error:
            json_output = "Could not parse output as JSON"
        
        return decoded_output, json_output
        
    except Exception as e:
        import traceback
        error_msg = f"❌ Error during inference:\n{str(e)}\n\n{traceback.format_exc()}"
        return error_msg, ""

# Create Gradio interface
with gr.Blocks(title="Pokémon Card OCR Demo", theme=gr.themes.Soft()) as demo:
    gr.Markdown("# 🎴 Pokémon Card OCR Demo")
    gr.Markdown("Extract card name and number from Pokémon card images using AI")
    gr.Markdown("**Models:** `unsloth/Llama-3.2-11B-Vision-Instruct-bnb-4bit` + `netprtony/Llama-3.2-11B-Vision-PokemonCard-OCR-LoRA`")
    
    # Device info
    if device == "cpu":
        gr.Markdown("⚠️ **Warning:** Running on CPU - Processing will be VERY slow. GPU strongly recommended for production use.")
    else:
        gr.Markdown(f"✅ **Using GPU:** {torch.cuda.get_device_name(0)}")
    
    with gr.Row():
        with gr.Column(scale=1):
            # Input section
            gr.Markdown("### 📥 Input")
            image_input = gr.Image(type="pil", label="Upload Pokémon Card Image")
            
            # Settings
            gr.Markdown("### ⚙️ Settings")
            max_tokens = gr.Slider(minimum=64, maximum=512, value=256, step=1, label="Max Tokens")
            temperature = gr.Slider(minimum=0.1, maximum=1.0, value=0.7, step=0.1, label="Temperature")
            top_p = gr.Slider(minimum=0.1, maximum=1.0, value=0.9, step=0.1, label="Top P")
            
            # Extract button
            extract_btn = gr.Button("🔍 Extract Information", variant="primary", size="lg")
        
        with gr.Column(scale=1):
            # Output section
            gr.Markdown("### 🎯 Results")
            raw_output = gr.Textbox(label="Raw Output", lines=10, max_lines=20)
            json_output = gr.Code(label="Parsed JSON", language="json", lines=10)
    
    # Example images
    gr.Markdown("### 📋 Examples")
    gr.Examples(
        examples=[
            ["https://images.pokemontcg.io/base1/4.png"],
            ["https://images.pokemontcg.io/base1/16.png"],
            ["https://images.pokemontcg.io/xy1/1.png"],
        ],
        inputs=[image_input],
        label="Click to load example images"
    )
    
    # Event handlers
    extract_btn.click(
        fn=extract_pokemon_card_info,
        inputs=[image_input, max_tokens, temperature, top_p],
        outputs=[raw_output, json_output]
    )
    
    # Footer
    gr.Markdown("---")
    gr.Markdown("🎴 Pokémon Card OCR Demo | Powered by Llama-3.2-11B-Vision with LoRA fine-tuning")
    gr.Markdown("Base Model: [unsloth/Llama-3.2-11B-Vision-Instruct-bnb-4bit](https://huggingface.co/unsloth/Llama-3.2-11B-Vision-Instruct-bnb-4bit)")
    gr.Markdown("LoRA Adapter: [netprtony/Llama-3.2-11B-Vision-PokemonCard-OCR-LoRA](https://huggingface.co/netprtony/Llama-3.2-11B-Vision-PokemonCard-OCR-LoRA)")

# Launch the app
if __name__ == "__main__":
    demo.launch(
        server_name="0.0.0.0",  # Allow external access
        server_port=7860,       # Default Gradio port
        share=False,            # Set to True to create a public link
        show_error=True,
    )