one more try

app.py

@@ -121,10 +121,16 @@ def generate_midi(seed, use_chords, chord_progression, bpm):
 
 @spaces.GPU(duration=120)
 def generate_music(midi_audio, prompt_duration, musicgen_model, num_iterations, bpm):
-    wav_filename, sample_rate = midi_audio
+    if isinstance(midi_audio, tuple):
+        wav_filename, sample_rate = midi_audio
+        song, sr = torchaudio.load(wav_filename)
+    else:
+        # Assuming midi_audio is a numpy array
+        with tempfile.NamedTemporaryFile(suffix=".wav", delete=False) as temp_file:
+            temp_filename = temp_file.name
+            torchaudio.save(temp_filename, torch.from_numpy(midi_audio), sample_rate=44100)
+            song, sr = torchaudio.load(temp_filename)
 
-    # Load the generated audio
-    song, sr = torchaudio.load(wav_filename)
     song = song.to(device)
 
     # Use the user-provided BPM value for duration calculation
@@ -178,8 +184,10 @@ def generate_music(midi_audio, prompt_duration, musicgen_model, num_iterations,
     combined_audio_filename = f"combined_audio_{random.randint(1, 10000)}.mp3"
     combined_audio.export(combined_audio_filename, format="mp3")
 
+    
     # Clean up temporary files
-    os.remove(temp_filename)
+    if not isinstance(midi_audio, tuple):
+        os.remove(temp_filename)
     for filename in all_audio_files:
         os.remove(filename)