update the get_vocab_size to compute actual length of the get_vocab function

tokenizers/src/tokenizer/mod.rs

@@ -668,11 +668,10 @@ where
     pub fn get_vocab_size(&self, with_added_tokens: bool) -> usize {
         // TODO ArthurZ THIS IS WRONG! We need to measure the length of the `set` because
         // now some tokens can be both in the added_tokens_encoder and in the vocab
-        self.model.get_vocab_size()
-            + if with_added_tokens {
-                self.added_vocabulary.len()
+        if with_added_tokens {
+            self.get_vocab(with_added_tokens).len()
         } else {
-                0
+            self.model.get_vocab_size()
         }
     }