multiple robots trying here

app.py

@@ -117,27 +117,19 @@ def generate_midi(seed, use_chords, chord_progression, bpm):
     # Clean up temporary MIDI file
     os.remove(midi_filename)
 
-    # Load the generated audio
-    song, sr = torchaudio.load(wav_filename)
-
-    # Clean up temporary MIDI file
-    os.remove(midi_filename)
-    os.remove(wav_filename)
-
-    return song.numpy(), sr
+    return wav_filename
 
 @spaces.GPU(duration=120)
-def generate_music(midi_data, prompt_duration, musicgen_model, num_iterations, bpm):
-    audio_data, sample_rate = midi_data
-
-    # Convert the audio data to a PyTorch tensor
-    song = torch.from_numpy(audio_data).to(device)
+def generate_music(midi_audio_path, prompt_duration, musicgen_model, num_iterations, bpm):
+    # Load the generated audio
+    song, sr = torchaudio.load(midi_audio_path)
+    song = song.to(device)
 
     # Use the user-provided BPM value for duration calculation
     duration = calculate_duration(bpm)
 
     # Create slices from the song using the user-provided BPM value
-    slices = create_slices(song, sample_rate, 35, bpm, num_slices=5)
+    slices = create_slices(song, sr, 35, bpm, num_slices=5)
 
     # Load the model
     model_name = musicgen_model.split(" ")[0]
@@ -160,10 +152,10 @@ def generate_music(midi_data, prompt_duration, musicgen_model, num_iterations, b
         
         print(f"Running iteration {i + 1} using slice {slice_idx}...")
         
-        prompt_waveform = slices[slice_idx][..., :int(prompt_duration * sample_rate)]
+        prompt_waveform = slices[slice_idx][..., :int(prompt_duration * sr)]
         prompt_waveform = preprocess_audio(prompt_waveform)
         
-        output = model_continue.generate_continuation(prompt_waveform, prompt_sample_rate=sample_rate, progress=True)
+        output = model_continue.generate_continuation(prompt_waveform, prompt_sample_rate=sr, progress=True)
         output = output.cpu()  # Move the output tensor back to CPU
         
         # Make sure the output tensor has at most 2 dimensions
@@ -190,8 +182,6 @@ def generate_music(midi_data, prompt_duration, musicgen_model, num_iterations, b
 
     return combined_audio_filename
 
-
-
 # Define the expandable sections
 musiclang_blurb = """
 ## musiclang