Fix Constrained beam search duplication and weird output issue (#17814)

* fix(ConstrainedBeamSearchScorer.step_sentence_constraint): avoid hypothesis duplication between topk and advance

* fix(GenerationMixin.constrained_beam_search): appropriately assign beam scores instead of token scores

src/transformers/generation_beam_search.py

@@ -655,7 +655,13 @@ class ConstrainedBeamSearchScorer(BeamScorer):
         full_hypotheses = torch.cat((input_ids[sent_beam_indices], sent_beam_tokens.unsqueeze(-1)), dim=-1)
 
         # need to make new hypothesis that advance the constraints
-        track_new = {"new_seqs": [], "new_states": [], "new_indices": [], "new_tokens": [], "new_scores": []}
+        track_new = {
+            "new_seqs": full_hypotheses.tolist(),
+            "new_states": [],
+            "new_indices": [],
+            "new_tokens": [],
+            "new_scores": [],
+        }
         for seq_idx, pre_seq in enumerate(this_batch_input_ids):
             # pre_seq = ith sequence generated before this step.
 

src/transformers/generation_utils.py

@@ -3220,10 +3220,10 @@ class GenerationMixin:
 
             next_token_scores_processed = logits_processor(input_ids, next_token_scores)
 
-            scores_for_all_vocab = next_token_scores_processed.clone()
-
             next_token_scores = next_token_scores_processed + beam_scores[:, None].expand_as(next_token_scores)
 
+            scores_for_all_vocab = next_token_scores.clone()
+
             # Store scores, attentions and hidden_states when required
             if return_dict_in_generate:
                 if output_scores: