Fix the tokenizer issue reported by @BramVanroy (#3)

config.json

@@ -1,5 +1,5 @@
 {
-  "_name_or_path": "./roberta-large--nl--oscar.3/checkpoint-26000",
+  "_name_or_path": "FremyCompany/roberta-large-nl-oscar23",
   "additional_special_tokens_ids": [],
   "architectures": [
     "RobertaForMaskedLM"
@@ -25,7 +25,7 @@
   "sep_token_id": 3,
   "tokenizer_class": "RobertaTokenizerFast",
   "torch_dtype": "float32",
-  "transformers_version": "4.29.2",
+  "transformers_version": "4.30.2",
   "type_vocab_size": 1,
   "unk_token_id": 2,
   "use_cache": true,

pytorch_model.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:7d3084a09d550ebd97e0a4f3ba0054bd2f41596d607964092430503ec92f53f9
-size 1420708626
+oid sha256:db4118175f1ec2d7bab2d2da17ecf3b08f48dd7426a0843efa72b9aa42b9b9e3
+size 1420693675

tokenizer.json

@@ -72,7 +72,7 @@
     "type": "RobertaProcessing",
     "sep": [
       "</s>",
-      2
+      3
     ],
     "cls": [
       "<s>",