[AutoTokenizer] Allow creation of tokenizers by tokenizer type (#13668)

* up

* up

tests/fixtures/merges.txt

@@ -0,0 +1,5 @@
+#version: 0.2
+Ġ l
+Ġl o
+Ġlo w
+e r

tests/fixtures/vocab.json

@@ -0,0 +1 @@
+{"l": 0, "o": 1, "w": 2, "e": 3, "r": 4, "s": 5, "t": 6, "i": 7, "d": 8, "n": 9, "Ġ": 10, "Ġl": 11, "Ġn": 12, "Ġlo": 13, "Ġlow": 14, "er": 15, "Ġlowest": 16, "Ġnewer": 17, "Ġwider": 18, "<unk>": 19, "<|endoftext|>": 20}

tests/fixtures/vocab.txt

@@ -0,0 +1,10 @@
+[PAD]
+[SEP]
+[MASK]
+[CLS]
+[unused3]
+[unused4]
+[unused5]
+[unused6]
+[unused7]
+[unused8]