more updates

app.py

@@ -20,7 +20,9 @@ import utils
 
 dtype = torch.float16
 device = torch.device("cuda")
-examples = [[True, True, "SD T2I", 4], [False, True, "Würstchen (T2I)", 4]]
+
+# pipeline_to_benchmark, batch_size, use_channels_last, do_torch_compile
+examples = [["SD T2I", 4, True, True], ["Würstchen (T2I)", 4, False, True]]
 
 pipeline_mapping = {
     "SD T2I": (DiffusionPipeline, "runwayml/stable-diffusion-v1-5"),
@@ -202,6 +204,8 @@ def generate(
         print(f"For {num_inference_steps} steps", end_time - start_time)
         print("Avg per step", (end_time - start_time) / num_inference_steps)
 
+    return f"Avg per step: {(end_time - start_time) / num_inference_steps} seconds."
+
 
 with gr.Blocks(css="style.css") as demo:
     do_torch_compile = gr.Checkbox(label="Enable torch.compile()?")
@@ -225,11 +229,12 @@ with gr.Blocks(css="style.css") as demo:
         rounded=(False, True, True, False),
         full_width=False,
     )
+    result = gr.Text(label="Result")
 
     gr.Examples(
         examples=examples,
         inputs=[pipeline_to_benchmark, batch_size, use_channels_last, do_torch_compile],
-        outputs="text",
+        outputs=result,
         fn=generate,
         cache_examples=True,
     )
@@ -237,7 +242,7 @@ with gr.Blocks(css="style.css") as demo:
     btn.click(
         fn=generate,
         inputs=[pipeline_to_benchmark, batch_size, use_channels_last, do_torch_compile],
-        outputs="text"
+        outputs=result,
     )
 
 demo.launch(show_error=True)