add- torch compile

src/server/main.py

@@ -386,6 +386,9 @@ class TranslateManager:
             torch_dtype=torch.float16,
             attn_implementation="flash_attention_2"
         ).to(self.device_type)
+
+        model = torch.compile(model, mode="reduce-overhead")
+        print("Model compiled with torch.compile")
         return tokenizer, model
 
 class ModelManager: