Constrained Beam Search [without disjunctive decoding]

docs/source/internal/generation_utils.mdx

@@ -16,8 +16,9 @@ This page lists all the utility functions used by [`~generation_utils.Generation
 [`~generation_utils.GenerationMixin.greedy_search`],
 [`~generation_utils.GenerationMixin.sample`],
 [`~generation_utils.GenerationMixin.beam_search`],
-[`~generation_utils.GenerationMixin.beam_sample`], and
-[`~generation_utils.GenerationMixin.group_beam_search`].
+[`~generation_utils.GenerationMixin.beam_sample`],
+[`~generation_utils.GenerationMixin.group_beam_search`], and
+[`~generation_utils.GenerationMixin.constrained_beam_search`].
 
 Most of those are only useful if you are studying the code of the generate methods in the library.
 
@@ -190,6 +191,16 @@ A [`StoppingCriteria`] can be used to change when to stop generation (other than
 [[autodoc]] MaxTimeCriteria
     - __call__
 
+## Constraints
+
+A [`Constraint`] can be used to force the generation to include specific tokens or sequences in the output.
+
+[[autodoc]] Constraint
+
+[[autodoc]] PhrasalConstraint
+
+[[autodoc]] ConstraintListState
+
 ## BeamSearch
 
 [[autodoc]] BeamScorer
@@ -200,6 +211,10 @@ A [`StoppingCriteria`] can be used to change when to stop generation (other than
     - process
     - finalize
 
+[[autodoc]] ConstrainedBeamSearchScorer
+    - process
+    - finalize
+
 ## Utilities
 
 [[autodoc]] top_k_top_p_filtering

src/transformers/__init__.py

@@ -611,7 +611,12 @@
         "TextDatasetForNextSentencePrediction",
     ]
     _import_structure["deepspeed"] = []
-    _import_structure["generation_beam_search"] = ["BeamScorer", "BeamSearchScorer"]
+    _import_structure["generation_beam_constraints"] = [
+        "Constraint",
+        "ConstraintListState",
+        "PhrasalConstraint",
+    ]
+    _import_structure["generation_beam_search"] = ["BeamScorer", "BeamSearchScorer", "ConstrainedBeamSearchScorer"]
     _import_structure["generation_logits_process"] = [
         "ForcedBOSTokenLogitsProcessor",
         "ForcedEOSTokenLogitsProcessor",
@@ -2748,7 +2753,8 @@
             TextDataset,
             TextDatasetForNextSentencePrediction,
         )
-        from .generation_beam_search import BeamScorer, BeamSearchScorer
+        from .generation_beam_constraints import Constraint, ConstraintListState, PhrasalConstraint
+        from .generation_beam_search import BeamScorer, BeamSearchScorer, ConstrainedBeamSearchScorer
         from .generation_logits_process import (
             ForcedBOSTokenLogitsProcessor,
             ForcedEOSTokenLogitsProcessor,