fix: deps

.gitignore

@@ -0,0 +1,2 @@
+.venv
+FluidR3_GM.sf2

app.py

@@ -47,11 +47,12 @@ def score_to_audio(score, sample_rate: int = 44100) -> tuple[int, np.ndarray]:
     # Convert to 16-bit integer PCM
     # Scale to full int16 range (-32768 to 32767)
     int_audio = (float_audio * 32767).astype(np.int16)
-    int_audio = np.trim_zeros(int_audio, "b")
+    int_audio = np.trim_zeros(int_audio, "fb")
     return sample_rate, int_audio
 
 
 device = "cuda" if torch.cuda.is_available() else "cpu"
+device = "cpu"
 
 tokenizer = MusicTokenizer.from_pretrained("shikhr/music_maker")
 
@@ -62,7 +63,7 @@ model.to(device)
 def generate_music():
     # Generate some music
     out = model.generate(
-        torch.tensor([[1]]).to(device), max_new_tokens=400, temperature=1.0, top_k=None
+        torch.tensor([[1]]).to(device), max_new_tokens=200, temperature=1.0, top_k=100
     )
 
     # Save the generated MIDI

main.py

@@ -0,0 +1,19 @@
+from transformers import AutoModel
+from miditok import MusicTokenizer
+import torch
+
+# device = "cuda" if torch.cuda.is_available() else "cpu"
+device = "cpu"
+
+tokenizer = MusicTokenizer.from_pretrained("shikhr/music_maker")
+
+model = AutoModel.from_pretrained("shikhr/music_maker", trust_remote_code=True)
+model.to(device)
+
+# Generate some music
+out = model.generate(
+    torch.tensor([[1]]).to(device), max_new_tokens=200, temperature=1.0, top_k=100
+)
+
+# Save the generated MIDI
+tokenizer(out[0].tolist()).dump_midi(f"generated.mid")

requirements.txt

@@ -1,4 +1,5 @@
-transformers
+transformers==4.48
 miditok
 torch
-pretty_midi
+pretty_midi
+pyfluidsynth