fix: CPU OOM issue during LoRA training (#41)

Author: SongwuJob

mova/engine/trainer/accelerate/accelerate_trainer.py

@@ -276,7 +276,7 @@ def _setup(self):
         if self.is_main_process:
             os.makedirs(self.save_path, exist_ok=True)
 
-        torch.cuda.memory._record_memory_history(enabled='all')
+        # torch.cuda.memory._record_memory_history(enabled='all')
 
     def _setup_lora(self):
         """Configure LoRA"""