Use `AlbertConverter` for FNet instead of using FNet's own converter (#14365)

* Add normalizer to FNetConverter

* Style

* Directly use AlbertConverter

src/transformers/convert_slow_tokenizer.py

@@ -526,18 +526,6 @@ class AlbertConverter(SpmConverter):
         )
 
 
-class FNetConverter(SpmConverter):
-    def post_processor(self):
-        return processors.TemplateProcessing(
-            single="[CLS]:0 $A:0 [SEP]:0",
-            pair="[CLS]:0 $A:0 [SEP]:0 $B:1 [SEP]:1",
-            special_tokens=[
-                ("[CLS]", self.original_tokenizer.convert_tokens_to_ids("[CLS]")),
-                ("[SEP]", self.original_tokenizer.convert_tokens_to_ids("[SEP]")),
-            ],
-        )
-
-
 class BarthezConverter(SpmConverter):
     def unk_id(self, proto):
         unk_id = 3
@@ -938,7 +926,7 @@ SLOW_TO_FAST_CONVERTERS = {
     "DPRQuestionEncoderTokenizer": BertConverter,
     "DPRContextEncoderTokenizer": BertConverter,
     "ElectraTokenizer": BertConverter,
-    "FNetTokenizer": FNetConverter,
+    "FNetTokenizer": AlbertConverter,
     "FunnelTokenizer": FunnelConverter,
     "GPT2Tokenizer": GPT2Converter,
     "HerbertTokenizer": HerbertConverter,