fix a typo in flax T5 attention - attention_mask variable is misnamed (#26663)

* fix a typo in flax t5 attention

* fix the typo in flax longt5 attention

src/transformers/models/longt5/modeling_flax_longt5.py

@@ -545,7 +545,7 @@ class FlaxLongT5Attention(nn.Module):
         # During fast autoregressive decoding, we feed one position at a time,
         # and cache the keys and values step by step.
         if self.causal and (self.has_variable("cache", "cached_key") or init_cache):
-            key_states, value_states, attention_attention_mask = self._concatenate_to_cache(
+            key_states, value_states, attention_mask = self._concatenate_to_cache(
                 key_states, value_states, query_states, attention_mask
             )
 

src/transformers/models/t5/modeling_flax_t5.py

@@ -405,7 +405,7 @@ class FlaxT5Attention(nn.Module):
         # During fast autoregressive decoding, we feed one position at a time,
         # and cache the keys and values step by step.
         if self.causal and (self.has_variable("cache", "cached_key") or init_cache):
-            key_states, value_states, attention_attention_mask = self._concatenate_to_cache(
+            key_states, value_states, attention_mask = self._concatenate_to_cache(
                 key_states, value_states, query_states, attention_mask
             )