### 1. Imports and class names setup ###
import gradio as gr
import os
import torch

from timeit import default_timer as timer
from typing import Tuple, Dict
import torchvision

from torch import nn
from torchvision.models import densenet121

def create_densenet121_model(num_classes: int = 2, seed: int = 42):
    """Creates a DenseNet121 model and transforms."""
    
    # Create DenseNet121 model
    model = densenet121(weights=None)  # Set to None since we will be loading our own weights

    # Freeze all layers in base model
    for param in model.parameters():
        param.requires_grad = False

    # Change classifier head with random seed for reproducibility
    torch.manual_seed(seed)
    model.classifier = nn.Linear(model.classifier.in_features, num_classes)

    transforms = torchvision.transforms.Compose([
        torchvision.transforms.Resize(224),
        torchvision.transforms.ToTensor(),
        torchvision.transforms.Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225])
    ])

    return model, transforms

# Create densenet121 model
densenet, densenet_transforms = create_densenet121_model()

# Load saved weights
state_dict = torch.load("model/FL_global_model_4be885f7-8d33-4498-a5ef-85aa301706bd.pt", map_location=torch.device("cpu"))
model_weights = state_dict["model"]
densenet.load_state_dict(model_weights,strict=False)  # Set strict to True since we now expect it to match

def predict(img) -> Tuple[Dict, float]:
    """Transforms and performs a prediction on img and returns prediction and time taken."""
    # Start the timer
    start_time = timer()

    # Transform the target image and add a batch dimension
    img = densenet_transforms(img).unsqueeze(0)

    # Put model into evaluation mode and turn on inference mode
    densenet.eval()
    with torch.inference_mode():
        pred_probs = torch.softmax(densenet(img), dim=1).squeeze()

    pred_labels_and_probs = {
        'Nodules': pred_probs[0].item(), 
        'Normal': pred_probs[1].item()
    }

    # Calculate the prediction time
    pred_time = round(timer() - start_time, 5)

    # Return the prediction dictionary and prediction time
    return pred_labels_and_probs, pred_time

example_list = [[f"examples/example{i+1}.jpg"] for i in range(3)]

title = "ChestXray Classification"
description = "A Densenet121 computer vision model to classify images of Xray Chest images as Normal or Nodules."
article = "model train by hamsteryang0"
# article = "Created at (https://github.com/azizche/chest_xray_Classification)."
demo = gr.Interface(
    fn=predict,
    inputs=gr.Image(type="pil"),
    outputs=[gr.Label(num_top_classes=2, label="Predictions"), gr.Number(label="Prediction time (s)")],
    examples=example_list,
    title=title,
    description=description,
    article=article
)

# Launch the demo!
demo.launch()