Upload OPTForMusicGeneration

model.py

@@ -4,16 +4,14 @@ import torch
 from miditok import TokSequence
 
 
-# class OPTForMusicGenerationConfig(OPTConfig):
-
-
 class OPTForMusicGeneration(OPTForCausalLM):
     
-    def generate_music(self, **kwargs):
+    def generate_music(self, tokenizer, **kwargs):
         input = torch.tensor([[self.config.bos_token_id]], device=self.device)
         midi = self.generate(input, **kwargs)
         generated_ts = TokSequence(ids=midi.tolist()[0], ids_bpe_encoded=True)
-        return generated_ts
+        generated_score = tokenizer(generated_ts)
+        return generated_score
 
 
 OPTForMusicGeneration.register_for_auto_class("AutoModel")