Change inference model

app.py

@@ -52,9 +52,8 @@ def populate_metadata(link):
     audio = get_audio(link)
     return yt.thumbnail_url, yt.title, audio
 
-def inference(inputs):
-    print(inputs)
-    with open(inputs[0], "rb") as fd:
+def inference(yt_audio):
+    with open(yt_audio, "rb") as fd:
         contents = fd.read()
 
     audio = upload_audio(contents,sample_rate=16000)
@@ -68,7 +67,7 @@ def inference(inputs):
     int16_data = note_seq.audio_io.float_samples_to_int16(array_of_floats)
     # piano_roll = create_image_from_note_sequence(note_sequence)
     
-    return "./transcribed.mid"
+    return "./transcribed.mid", (SAMPLE_RATE, int16_data)
   
 title = "Transcribe music from YouTube videos using Transformers."
 description = """
@@ -110,13 +109,12 @@ with demo:
         with gr.Row():
             btn = gr.Button("Transcribe music")
         
-        audio_file = gr.File()
+        with gr.Row():
+            midi_file = gr.File()
+            midi_audio = gr.Audio()
         
-        btn.click(inference,
-                  inputs = [
-                      yt_audio
-                  ],
-                  outputs=audio_file)
+        btn.click(inference(yt_audio),
+                  outputs=[midi_file, midi_audio])
         
         gr.Markdown(article)