handle when from_pretrained_id is a list

tests/test_tokenization_common.py

@@ -1580,17 +1580,15 @@ class TokenizerTesterMixin:
                     self.assertEqual(overflowing_tokens, seq1_tokens[-(2 + stride) :])
 
     # TODO: FIXME @ArthurZucker
-    @unittest.skip(
-        reason="start to fail after # 29473. See https://github.com/huggingface/transformers/pull/29473#pullrequestreview-1945687810"
-    )
+
     @slow
     @require_read_token
     def test_encode_decode_fast_slow_all_tokens(self):
         if self.rust_tokenizer_class is not None:
-            pretrained_name = self.from_pretrained_id
-
-            slow_tokenizer = self.tokenizer_class.from_pretrained(pretrained_name, legacy=False)
+            pretrained_names = self.from_pretrained_id if isinstance(self.from_pretrained_id, list) else [self.from_pretrained_id]
+            for pretrained_name in pretrained_names:
                 with self.subTest(f"{pretrained_name}"):
+                    slow_tokenizer = self.tokenizer_class.from_pretrained(pretrained_name, legacy=False)
                     rust_tokenizer = self.rust_tokenizer_class.from_pretrained(
                         pretrained_name, from_slow=True, legacy=False
                     )