Set use_cache to true for better generation performance

config.json

@@ -26,7 +26,7 @@
   "tie_word_embeddings": false,
   "torch_dtype": "bfloat16",
   "transformers_version": "4.25.1",
-  "use_cache": false,
+  "use_cache": true,
   "use_parallel_residual": true,
   "vocab_size": 50280
 }