bad_words_ids no longer slow on mps (#39556)

* fix: bad_words_ids no longer slow on mps

* fix: SequenceBiasLogitsProcessor slow `_prepare_bias_variables` method

* fix: re-adding a deleted comment

* fix: bug in no_bad_words_logits

* Apply style fixes

---------

Co-authored-by: github-actions[bot] <github-actions[bot]@users.noreply.github.com>

Author: DWarez

src/transformers/generation/logits_process.py

@@ -1222,10 +1222,16 @@ def _prepare_bias_variables(self, scores: torch.FloatTensor):
         # Precompute the bias tensors to be applied. Sequences of length 1 are kept separately, as they can be applied
         # with simpler logic.
         self.length_1_bias = torch.zeros((vocabulary_size,), dtype=torch.float, device=scores.device)
+        # Extract single-token sequences and their biases
+        single_token_ids = []
+        single_token_biases = []
         for sequence_ids, bias in self.sequence_bias.items():
             if len(sequence_ids) == 1:
-                self.length_1_bias[sequence_ids[-1]] = bias
+                single_token_ids.append(sequence_ids[0])
+                single_token_biases.append(bias)
 
+        if single_token_ids:  # Only if we have any single-token sequences
+            self.length_1_bias[single_token_ids] = torch.tensor(single_token_biases, device=scores.device)
         self.prepared_bias_variables = True
 
     def _validate_arguments(self):
@@ -1340,10 +1346,10 @@ def __init__(
                     eos_token_id = [eos_token_id]
                 eos_token_id = torch.tensor(eos_token_id)
 
+            eos_token_id_list = eos_token_id.tolist()  # convert to python list before
             bad_words_ids = list(
-                filter(lambda bad_token_seq: all(bad_token_seq != [i] for i in eos_token_id), bad_words_ids)
+                filter(lambda bad_token_seq: all(bad_token_seq != [i] for i in eos_token_id_list), bad_words_ids)
             )
-
         # Forbidding a sequence is equivalent to setting its bias to -inf
         sequence_bias = {tuple(sequence): float("-inf") for sequence in bad_words_ids}
         super().__init__(sequence_bias=sequence_bias)