Create app.py

app.py

@@ -0,0 +1,32 @@
+import torch
+import gradio as gr
+from torchvision.utils import save_image
+from torchvision.transforms import ToPILImage
+from model import Generator
+
+DEVICE = "cuda" if torch.cuda.is_available() else "cpu"
+NOISE_DIM = 256
+
+G = Generator().to(DEVICE)
+G.load_state_dict(torch.load("generator.pth", map_location=DEVICE))
+G.eval()
+
+to_pil = ToPILImage()
+
+def generate_image():
+    noise = torch.randn(1, NOISE_DIM).to(DEVICE)
+
+    with torch.no_grad():
+        image = G(noise)
+
+    image = (image + 1) / 2
+    return to_pil(image.squeeze(0))
+
+demo = gr.Interface(
+    fn=generate_image,
+    inputs=None,
+    outputs=gr.Image(),
+    title="GAN Image Generator API"
+)
+
+demo.launch()