import gradio as gr
import utils
from PIL import Image
import torch
import math
from torchvision import transforms


device = "cpu"
years = [str(y) for y in range(1880, 2020, 10)]


orig_models = {}

for year in years:
    G, w_avg = utils.load_stylegan2(f"pretrained_models/{year}.pkl", device)
    orig_models[year] = { "G": G.eval()}
    
transform = transforms.Compose([
        transforms.Resize((256, 256)),
        transforms.ToTensor(),
        transforms.Normalize([0.5, 0.5, 0.5], [0.5, 0.5, 0.5])])

# Download human-readable labels for ImageNet.
def predict(inp):
  #with torch.no_grad():     
  return inp


gr.Interface(fn=predict, 
             inputs=gr.Image(type="pil"),
             outputs=gr.Image(type="pil"),
             #examples=["lion.jpg", "cheetah.jpg"]
             ).launch()