Fixed vocab mismatch and resized embeddings

tokenizer_config.json

@@ -2,7 +2,13 @@
   "backend": "tokenizers",
   "bos_token": "<|endoftext|>",
   "eos_token": "<|endoftext|>",
+  "is_local": false,
+  "local_files_only": false,
+  "max_length": 16,
   "model_max_length": 1000000000000000019884624838656,
   "pad_token": "<|pad|>",
-  "tokenizer_class": "TokenizersBackend"
+  "stride": 0,
+  "tokenizer_class": "TokenizersBackend",
+  "truncation_side": "right",
+  "truncation_strategy": "longest_first"
 }