fix wrong tokenizer checkpoint name in flax marian (#15391)

Co-authored-by: ydshieh <ydshieh@users.noreply.github.com>

src/transformers/models/marian/modeling_flax_marian.py

@@ -986,7 +986,7 @@ class FlaxMarianPreTrainedModel(FlaxPreTrainedModel):
         ```python
         >>> from transformers import MarianTokenizer, FlaxMarianMTModel
 
-        >>> tokenizer = MarianTokenizer.from_pretrained("facebook/marian-large-cnn")
+        >>> tokenizer = MarianTokenizer.from_pretrained("Helsinki-NLP/opus-mt-en-de")
         >>> model = FlaxMarianMTModel.from_pretrained("Helsinki-NLP/opus-mt-en-de")
 
         >>> text = "My friends are cool but they eat too many carbs."
@@ -1054,7 +1054,7 @@ class FlaxMarianPreTrainedModel(FlaxPreTrainedModel):
         >>> import jax.numpy as jnp
         >>> from transformers import MarianTokenizer, FlaxMarianMTModel
 
-        >>> tokenizer = MarianTokenizer.from_pretrained("facebook/marian-large-cnn")
+        >>> tokenizer = MarianTokenizer.from_pretrained("Helsinki-NLP/opus-mt-en-de")
         >>> model = FlaxMarianMTModel.from_pretrained("Helsinki-NLP/opus-mt-en-de")
 
         >>> text = "My friends are cool but they eat too many carbs."