feat: two run buttons - Run (30s) for normal tasks, Run Long (60s) for counting/detection

app.py

@@ -101,14 +101,8 @@ def draw_bboxes(image, bboxes):
 # ---------------------------------------------------------------------------
 # Inference (supports both single-image and paired-image modes)
 # ---------------------------------------------------------------------------
-@spaces.GPU(duration=30)
-def predict(
-    reference_image: Image.Image,
-    test_image: Image.Image,
-    prompt: str,
-    max_new_tokens: float,
-):
-    # Determine mode: single image or paired comparison
+def _run_inference(reference_image, test_image, prompt, max_new_tokens):
+    """Core inference logic shared by both predict functions."""
     has_ref = reference_image is not None
     has_test = test_image is not None
 
@@ -119,7 +113,6 @@ def predict(
         t_start = time.time()
         max_new_tokens = int(max_new_tokens)
 
-        # Build message content based on available images
         content = []
 
         if has_ref and has_test:
@@ -174,13 +167,11 @@ def predict(
             clean_up_tokenization_spaces=False,
         )[0]
 
-        # Try to visualize bboxes if present
         bboxes = parse_bboxes(output)
         vis_image = None
         if bboxes:
             vis_image = draw_bboxes(vis_source, bboxes)
 
-        # Append timing info
         prep_time = t_preprocess - t_start
         gen_time = t_generate - t_preprocess
         output += f"\n\n---\nPreprocessing: {prep_time:.1f}s | Inference: {gen_time:.1f}s"
@@ -192,6 +183,16 @@ def predict(
         return f"Error:\n{tb}", None
 
 
+@spaces.GPU(duration=30)
+def predict(reference_image, test_image, prompt, max_new_tokens):
+    return _run_inference(reference_image, test_image, prompt, max_new_tokens)
+
+
+@spaces.GPU(duration=60)
+def predict_long(reference_image, test_image, prompt, max_new_tokens):
+    return _run_inference(reference_image, test_image, prompt, max_new_tokens)
+
+
 # ---------------------------------------------------------------------------
 # Gradio UI
 # ---------------------------------------------------------------------------
@@ -291,7 +292,8 @@ with gr.Blocks(theme=gr.themes.Soft(), title=TITLE) as demo:
             step=16,
             label="Max New Tokens",
         )
-        run_btn = gr.Button("Run", variant="primary", scale=2)
+        run_btn = gr.Button("Run (30s)", variant="primary", scale=2)
+        run_long_btn = gr.Button("Run Long (60s)", variant="secondary", scale=1)
 
     output = gr.Textbox(label="Model Output", lines=4)
     vis_output = gr.Image(label="Detection Visualization")
@@ -301,6 +303,11 @@ with gr.Blocks(theme=gr.themes.Soft(), title=TITLE) as demo:
         inputs=[ref_img, test_img, prompt, max_tokens],
         outputs=[output, vis_output],
     )
+    run_long_btn.click(
+        fn=predict_long,
+        inputs=[ref_img, test_img, prompt, max_tokens],
+        outputs=[output, vis_output],
+    )
 
     gr.Examples(
         examples=EXAMPLES,