[cherry-pick] Ensure actual lenght of transformed seq by T5 Transform…

… cannot be longer than specified max_len (#2143)

* Ensure max seq length is not 1 > actual max seq len

Summary: Currently a "bug" wherein max length can actually be 1 greater than actual max_seq length specified b/c we add EOS token.

Reviewed By: yohann-benchetrit

Differential Revision: D44708339

fbshipit-source-id: dc24d268caa4eb783df8ac7e8a2463e33d3244d9

* Format new changes

---------

Co-authored-by: Joe Cummings <jrcummings@meta.com>

torchtext/models/t5/t5_transform.py

@@ -45,7 +45,7 @@ def __init__(self, sp_model_path: str, max_seq_len: int, eos_idx: int, padding_i
         self.max_seq_len = max_seq_len
         self.eos_idx = eos_idx
         self.padding_idx = padding_idx
-        self.pipeline = T.Sequential(T.Truncate(self.max_seq_len), T.AddToken(token=self.eos_idx, begin=False))
+        self.pipeline = T.Sequential(T.Truncate(self.max_seq_len - 1), T.AddToken(token=self.eos_idx, begin=False))
 
     def forward(self, input: Union[str, List[str]]) -> torch.Tensor:
         """