add whisper tiny

app.py

@@ -119,18 +119,6 @@ def transcribe(inputs):
     distil_text = distil_pipe(inputs.copy(), batch_size=BATCH_SIZE)["text"]
     yield distil_text, distil_runtime, None, None, None, None
 
-    def _forward_time(*args, **kwargs):
-        global runtime
-        start_time = time.time()
-        result = pipe_forward(*args, **kwargs)
-        runtime = time.time() - start_time
-        runtime = round(runtime, 2)
-        return result
-
-    pipe._forward = _forward_time
-    text = pipe(inputs.copy(), batch_size=BATCH_SIZE)["text"]
-    yield distil_text, distil_runtime, text, runtime, None, None
-
     def _forward_tiny_time(*args, **kwargs):
         global tiny_runtime
         start_time = time.time()
@@ -141,8 +129,19 @@ def transcribe(inputs):
 
     tiny_pipe._forward = _forward_tiny_time
     tiny_text = tiny_pipe(inputs.copy(), batch_size=BATCH_SIZE)["text"]
+    yield distil_text, distil_runtime, tiny_text, tiny_runtime, None, None
 
-    yield distil_text, distil_runtime, text, runtime, tiny_text, tiny_runtime
+    def _forward_time(*args, **kwargs):
+        global runtime
+        start_time = time.time()
+        result = pipe_forward(*args, **kwargs)
+        runtime = time.time() - start_time
+        runtime = round(runtime, 2)
+        return result
+
+    pipe._forward = _forward_time
+    text = pipe(inputs.copy(), batch_size=BATCH_SIZE)["text"]
+    yield distil_text, distil_runtime, tiny_text, tiny_runtime, text, runtime
 
 if __name__ == "__main__":
     with gr.Blocks() as demo:
@@ -179,23 +178,23 @@ if __name__ == "__main__":
         button = gr.Button("Transcribe")
         with gr.Row():
             distil_runtime = gr.components.Textbox(label="distil-large-v3 Transcription Time (s)")
-            runtime = gr.components.Textbox(label="whisper-largel-v3  Transcription Time (s)")
             tiny_runtime = gr.components.Textbox(label="whisper-tiny Transcription Time (s)")
+            runtime = gr.components.Textbox(label="whisper-largel-v3  Transcription Time (s)")
 
         with gr.Row():
             distil_transcription = gr.components.Textbox(label="distil-large-v3 Transcription", show_copy_button=True)
-            transcription = gr.components.Textbox(label="whisper-largel-v3 Transcription", show_copy_button=True)
             tiny_transcription = gr.components.Textbox(label="whisper-tiny Transcription", show_copy_button=True)
+            transcription = gr.components.Textbox(label="whisper-largel-v3 Transcription", show_copy_button=True)
         button.click(
             fn=transcribe,
             inputs=audio,
-            outputs=[distil_transcription, distil_runtime, transcription, runtime, tiny_transcription, tiny_runtime],
+            outputs=[distil_transcription, distil_runtime, tiny_transcription, tiny_runtime, transcription, runtime],
         )
         gr.Markdown("## Examples")
         gr.Examples(
             [["./assets/example_1.wav"], ["./assets/example_2.wav"]],
             audio,
-            outputs=[distil_transcription, distil_runtime, transcription, runtime, tiny_transcription, tiny_runtime],
+            outputs=[distil_transcription, distil_runtime, tiny_transcription, tiny_runtime, transcription, runtime],
             fn=transcribe,
             cache_examples=False,
         )