:tada: init

app.py

@@ -188,8 +188,9 @@ unet.set_attn_processor(
 
 
 @torch.inference_mode()
-def inference(image: Image.Image, prompt: str, progress=gr.Progress(track_tqdm=True)):
+def inference(image_path: str, prompt: str, progress=gr.Progress(track_tqdm=False)):
     progress(0, "Initializing...")
+    image = Image.open(image_path)
     image = image.convert("RGB").resize((512, 512))
     image = to_tensor(image).unsqueeze(0).to(device)
 
@@ -331,14 +332,14 @@ if __name__ == '__main__':
         description="This is a visualizer for the attention maps of the Stable Diffusion model. ",
         fn=inference,
         inputs=[
-            gr.Image(type="pil", label="Input", width=512, height=512),
+            gr.Image(type="filepath", label="Input", width=512, height=512),
             gr.Textbox(label="Prompt", placeholder="e.g.) A photo of dog...")
         ],
         outputs=ca_output,
         cache_examples=True,
         examples=[
-            [Image.open("assets/aeroplane.png"), "plane background"],
-            [Image.open("assets/dogcat.png"), "a photo of dog and cat"],
+            ["assets/aeroplane.png", "plane background"],
+            ["assets/dogcat.png", "a photo of dog and cat"],
         ]
     )