app.py

import gradio as gr
import torch
import torch.nn as nn
import numpy as np
from PIL import Image
import os

# Set device
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")

# Model parameters
nz = 100
ngf = 64
num_classes = 10

# Generator class
class Generator(nn.Module):
    def __init__(self):
        super(Generator, self).__init__()
        
        self.label_embedding = nn.Embedding(num_classes, num_classes)
        
        self.main = nn.Sequential(
            nn.ConvTranspose2d(nz + num_classes, ngf * 8, 4, 1, 0, bias=False),
            nn.BatchNorm2d(ngf * 8),
            nn.ReLU(True),
            
            nn.ConvTranspose2d(ngf * 8, ngf * 4, 4, 2, 1, bias=False),
            nn.BatchNorm2d(ngf * 4),
            nn.ReLU(True),
            
            nn.ConvTranspose2d(ngf * 4, ngf * 2, 4, 2, 1, bias=False),
            nn.BatchNorm2d(ngf * 2),
            nn.ReLU(True),
            
            nn.ConvTranspose2d(ngf * 2, 1, 4, 2, 1, bias=False),
            nn.Tanh()
        )
        
        self.resize = nn.AdaptiveAvgPool2d((28, 28))
    
    def forward(self, noise, labels):
        label_embedding = self.label_embedding(labels)
        label_embedding = label_embedding.view(label_embedding.size(0), num_classes, 1, 1)
        input_tensor = torch.cat([noise, label_embedding], dim=1)
        output = self.main(input_tensor)
        output = self.resize(output)
        return output

# Load model function (NO @st.cache_resource decorator!)
def load_model():
    generator = Generator().to(device)
    
    try:
        checkpoint = torch.load('mnist_gan_model.pth', map_location=device)
        generator.load_state_dict(checkpoint['generator_state_dict'])
        generator.eval()
        print("✅ Model loaded successfully!")
        return generator
    except Exception as e:
        print(f"❌ Error loading model: {e}")
        return None

# Initialize generator
generator = load_model()

# Generation function
def generate_digit_images(digit):
    """Generate 5 images of the specified digit"""
    
    if generator is None:
        return [Image.new('L', (112, 112), 128)] * 5
    
    digit = int(digit)
    num_images = 5
    
    with torch.no_grad():
        noise = torch.randn(num_images, nz, 1, 1).to(device)
        labels = torch.full((num_images,), digit, dtype=torch.long).to(device)
        
        generated_images = generator(noise, labels)
        
        images = generated_images.cpu().numpy()
        images = (images + 1) / 2.0
        images = np.squeeze(images)
        
        pil_images = []
        for img in images:
            img_uint8 = (img * 255).astype(np.uint8)
            pil_img = Image.fromarray(img_uint8, mode='L')
            pil_img = pil_img.resize((112, 112), Image.NEAREST)
            pil_images.append(pil_img)
        
        return pil_images

# Gradio interface
def create_app():
    with gr.Blocks(title="Handwritten Digit Generator", theme=gr.themes.Soft()) as app:
        
        gr.Markdown("# 🔢 Handwritten Digit Generator")
        gr.Markdown("Generate synthetic MNIST-like digit images using a trained GAN model.")
        
        with gr.Row():
            with gr.Column(scale=1):
                digit_input = gr.Dropdown(
                    choices=list(range(10)), 
                    value=8, 
                    label="Choose a digit (0-9)"
                )
                generate_btn = gr.Button("🎨 Generate Images", variant="primary")
            
            with gr.Column(scale=2):
                gr.Markdown("### Generated Images")
                image_gallery = gr.Gallery(
                    label="Generated Digit Images", 
                    show_label=False,
                    columns=5,
                    rows=1,
                    height=200
                )
        
        generate_btn.click(
            fn=generate_digit_images,
            inputs=[digit_input],
            outputs=[image_gallery]
        )
        
        # Auto-generate on load
        app.load(
            fn=lambda: generate_digit_images(8),
            outputs=[image_gallery]
        )
        
        gr.Markdown("---")
        gr.Markdown("**🤖 Model**: Conditional GAN | **⚡ Framework**: PyTorch + Gradio")
    
    return app

# Launch
if __name__ == "__main__":
    app = create_app()
    app.launch()