Cache examples

app.py

@@ -16,8 +16,12 @@ args.print_step = None
 device = "cuda" if torch.cuda.is_available() else "cpu"
 model, _, preprocess = open_clip.create_model_and_transforms(args.clip_model, pretrained=args.clip_pretrain, device=device)
 
+args.counter = 0
 
 def inference(target_image, prompt_len, iter):
+    args.counter += 1
+    print(args.counter)
+    
     if prompt_len is not None:
         args.prompt_len = int(prompt_len)
     else:
@@ -29,10 +33,13 @@ def inference(target_image, prompt_len, iter):
         args.iter = 1000
         
     learned_prompt = optimize_prompt(model, preprocess, args, device, target_images=[target_image])
-
+    
     return learned_prompt
-
+    
 def inference_text(target_prompt, prompt_len, iter):
+    args.counter += 1
+    print(args.counter)
+    
     if prompt_len is not None:
         args.prompt_len = min(int(prompt_len), 75)
     else:
@@ -85,7 +92,7 @@ with demo:
     image_button.click(inference, inputs=[input_image, prompt_len_field, num_step_field], outputs=output_prompt)
     prompt_button.click(inference_text, inputs=[input_prompt, prompt_len_field, num_step_field], outputs=output_prompt)
 
-    gr.Examples([["sample.jpeg", 8, 1000]], inputs=[input_image, prompt_len_field, num_step_field])
-    gr.Examples([["digital concept art of old wooden cabin in florida swamp, trending on artstation", 3, 1000]], inputs=[input_prompt, prompt_len_field, num_step_field])
+    gr.Examples([["sample.jpeg", 8, 1000]], inputs=[input_image, prompt_len_field, num_step_field], fn=inference, outputs=output_prompt, cache_examples=True)
+    gr.Examples([["digital concept art of old wooden cabin in florida swamp, trending on artstation", 3, 1000]], inputs=[input_prompt, prompt_len_field, num_step_field], fn=inference_text, outputs=output_prompt, cache_examples=True)
 
 demo.launch(enable_queue=True)