Add model and resource file downloads

app.py

@@ -17,6 +17,10 @@ from huggingface_hub import hf_hub_download
 repo_id = "amaai-lab/text2midi"
 # Download the model.bin file
 model_path = hf_hub_download(repo_id=repo_id, filename="pytorch_model.bin")
+# Download the vocab_remi.pkl file
+tokenizer_path = hf_hub_download(repo_id=repo_id, filename="vocab_remi.pkl")
+# Download the soundfont file
+soundfont_path = hf_hub_download(repo_id=repo_id, filename="soundfont.sf2")
 
 
 def save_wav(filepath):
@@ -29,8 +33,9 @@ def save_wav(filepath):
     wav_filepath = os.path.join(directory, f"{stem}.wav")
 
     # Run the fluidsynth command to convert MIDI to WAV
+    # f"fluidsynth -r 16000 soundfont.sf2 -g 1.0 --quiet --no-shell {midi_filepath} -T wav -F {wav_filepath} > /dev/null",
     process = subprocess.Popen(
-        f"fluidsynth -r 16000 soundfont.sf2 -g 1.0 --quiet --no-shell {midi_filepath} -T wav -F {wav_filepath} > /dev/null",
+        f"fluidsynth -r 16000 {soundfont_path} -g 1.0 --quiet --no-shell {midi_filepath} -T wav -F {wav_filepath} > /dev/null",
         shell=True
     )
     process.wait()
@@ -77,9 +82,9 @@ def generate_midi(caption, temperature=0.9, max_len=500):
     device = 'cuda' if torch.cuda.is_available() else 'cpu'
     artifact_folder = 'artifacts'
 
-    tokenizer_filepath = os.path.join(artifact_folder, "vocab_remi.pkl")
+    # tokenizer_filepath = os.path.join(artifact_folder, "vocab_remi.pkl")
     # Load the tokenizer dictionary
-    with open(tokenizer_filepath, "rb") as f:
+    with open(tokenizer_path, "rb") as f:
         r_tokenizer = pickle.load(f)
 
     # Get the vocab size