added tokenizer_config, so the tokenizer is cased by default

tokenizer_config.json

@@ -0,0 +1,4 @@
+{
+  "do_lower_case": false,
+  "max_len": 512
+}