Update config.json

Using cached embeddings to decrease inference time.

config.json

@@ -25,6 +25,6 @@
   "tie_word_embeddings": false,
   "torch_dtype": "float16",
   "transformers_version": "4.31.0",
-  "use_cache": false,
+  "use_cache": true,
   "vocab_size": 32000
 }