sweeep ready for tests

Author: manueldeprada

docs/source/en/internal/generation_utils.md

@@ -198,18 +198,6 @@ A [`StoppingCriteria`] can be used to change when to stop generation (other than
 [[autodoc]] EosTokenCriteria
     - __call__
 
-## Constraints
-
-A [`Constraint`] can be used to force the generation to include specific tokens or sequences in the output. Please note that this is exclusively available to our PyTorch implementations.
-
-[[autodoc]] Constraint
-
-[[autodoc]] PhrasalConstraint
-
-[[autodoc]] DisjunctiveConstraint
-
-[[autodoc]] ConstraintListState
-
 ## Streamers
 
 [[autodoc]] TextStreamer

docs/source/ja/internal/generation_utils.md

@@ -300,18 +300,6 @@ generation_output[:2]
 [[autodoc]] MaxTimeCriteria
     - __call__
 
-## Constraints
-
-[`Constraint`] を使用すると、生成時に出力に特定のトークンまたはシーケンスが含まれるように強制できます。これは PyTorch 実装でのみ利用可能であることに注意してください。
-
-[[autodoc]] Constraint
-
-[[autodoc]] PhrasalConstraint
-
-[[autodoc]] DisjunctiveConstraint
-
-[[autodoc]] ConstraintListState
-
 ## Streamers
 
 [[autodoc]] TextStreamer

docs/source/ko/internal/generation_utils.md

@@ -305,18 +305,6 @@ generation_output[:2]
 [[autodoc]] EosTokenCriteria
     - __call__
 
-## Constraint [[transformers.Constraint]]
-
-[`Constraint`]는 생성 출력에 특정 토큰이나 시퀀스를 강제로 포함시키는 데 사용됩니다. 이 기능은 PyTorch 구현에만 제공됩니다.
-
-[[autodoc]] Constraint
-
-[[autodoc]] PhrasalConstraint
-
-[[autodoc]] DisjunctiveConstraint
-
-[[autodoc]] ConstraintListState
-
 ## 스트리머 (Streamers) [[transformers.TextStreamer]]
 
 [[autodoc]] TextStreamer

docs/source/zh/internal/generation_utils.md

@@ -295,18 +295,6 @@ generation_output[:2]
 [[autodoc]] MaxTimeCriteria
     - __call__
 
-## Constraints
-
-可以使用[`Constraint`]来强制生成结果包含输出中的特定tokens或序列。请注意，这仅适用于我们的PyTorch实现。
-
-[[autodoc]] Constraint
-
-[[autodoc]] PhrasalConstraint
-
-[[autodoc]] DisjunctiveConstraint
-
-[[autodoc]] ConstraintListState
-
 ## Streamers
 
 [[autodoc]] TextStreamer

src/transformers/generation/beam_constraints.py

@@ -7,6 +7,7 @@
 logger = logging.get_logger(__name__)
 
 
+# TODO joao, manuel: remove in v4.58.0
 class Constraint(ABC):
     r"""Abstract base class for all constraints that can be applied during generation.
     It must define how the constraint can be satisfied.

src/transformers/generation/configuration_utils.py

@@ -89,7 +89,6 @@ class GenerationConfig(PushToHubMixin):
         - *multinomial sampling* if `num_beams=1` and `do_sample=True`
         - *beam-search decoding* if `num_beams>1` and `do_sample=False`
         - *beam-search multinomial sampling* if `num_beams>1` and `do_sample=True`
-        - *constrained beam-search decoding* if `constraints!=None` or `force_words_ids!=None`
         - *assisted decoding* if `assistant_model` or `prompt_lookup_num_tokens` is passed to `.generate()`
 
     To learn more about decoding strategies refer to the [text generation strategies guide](../generation_strategies).
@@ -202,18 +201,10 @@ class GenerationConfig(PushToHubMixin):
         bad_words_ids (`list[list[int]]`, *optional*):
             List of list of token ids that are not allowed to be generated. Check
             [`~generation.NoBadWordsLogitsProcessor`] for further documentation and examples.
-        force_words_ids (`list[list[int]]` or `list[list[list[int]]]`, *optional*):
-            List of token ids that must be generated. If given a `list[list[int]]`, this is treated as a simple list of
-            words that must be included, the opposite to `bad_words_ids`. If given `list[list[list[int]]]`, this
-            triggers a [disjunctive constraint](https://github.com/huggingface/transformers/issues/14081), where one
-            can allow different forms of each word.
         renormalize_logits (`bool`, *optional*, defaults to `False`):
             Whether to renormalize the logits after applying all the logits processors (including the custom
             ones). It's highly recommended to set this flag to `True` as the search algorithms suppose the score logits
             are normalized but some logit processors break the normalization.
-        constraints (`list[Constraint]`, *optional*):
-            Custom constraints that can be added to the generation to ensure that the output will contain the use of
-            certain tokens as defined by `Constraint` objects, in the most sensible way possible.
         forced_bos_token_id (`int`, *optional*, defaults to `model.config.forced_bos_token_id`):
             The id of the token to force as the first generated token after the `decoder_start_token_id`. Useful for
             multilingual models like [mBART](../model_doc/mbart) where the first generated token needs to be the target
@@ -374,9 +365,7 @@ def __init__(self, **kwargs):
         self.length_penalty = kwargs.pop("length_penalty", 1.0)
         self.no_repeat_ngram_size = kwargs.pop("no_repeat_ngram_size", 0)
         self.bad_words_ids = kwargs.pop("bad_words_ids", None)
-        self.force_words_ids = kwargs.pop("force_words_ids", None)
         self.renormalize_logits = kwargs.pop("renormalize_logits", False)
-        self.constraints = kwargs.pop("constraints", None)
         self.forced_bos_token_id = kwargs.pop("forced_bos_token_id", None)
         self.forced_eos_token_id = kwargs.pop("forced_eos_token_id", None)
         self.remove_invalid_values = kwargs.pop("remove_invalid_values", False)
@@ -434,6 +423,8 @@ def __init__(self, **kwargs):
         self.dola_layers = kwargs.pop("dola_layers", None)
         self.diversity_penalty = kwargs.pop("diversity_penalty", 0.0)
         self.num_beam_groups = kwargs.pop("num_beam_groups", 1)
+        self.constraints = kwargs.pop("constraints", None)
+        self.force_words_ids = kwargs.pop("force_words_ids", None)
 
         # The remaining attributes do not parametrize `.generate()`, but are informative and/or used by the hub
         # interface.
@@ -625,24 +616,6 @@ def validate(self, strict=False):
                 minor_issues["length_penalty"] = single_beam_wrong_parameter_msg.format(
                     flag_name="length_penalty", flag_value=self.length_penalty
                 )
-            if self.constraints is not None:
-                minor_issues["constraints"] = single_beam_wrong_parameter_msg.format(
-                    flag_name="constraints", flag_value=self.constraints
-                )
-
-        # 2.3. detect incorrect parameterization specific to advanced beam modes
-        else:
-            # constrained beam search
-            if self.constraints is not None or self.force_words_ids is not None:
-                constrained_wrong_parameter_msg = (
-                    "one of `constraints`, `force_words_ids` is not `None`, triggering constrained beam search. "
-                    "However, `{flag_name}` is set to `{flag_value}`, which is incompatible with this generation "
-                    "mode. Set `constraints` and `force_words_ids` to `None` or unset `{flag_name}` to continue."
-                )
-                if self.do_sample is True:
-                    raise ValueError(
-                        constrained_wrong_parameter_msg.format(flag_name="do_sample", flag_value=self.do_sample)
-                    )
 
         # 2.4. check `num_return_sequences`
         if self.num_return_sequences != 1:

src/transformers/generation/utils.py

@@ -367,7 +367,6 @@ class GenerationMixin(ContinuousMixin):
         - *multinomial sampling* if `num_beams=1` and `do_sample=True`
         - *beam-search decoding* if `num_beams>1` and `do_sample=False`
         - *beam-search multinomial sampling* if `num_beams>1` and `do_sample=True`
-        - *constrained beam-search decoding* if `constraints!=None` or `force_words_ids!=None`
         - *assisted decoding* if `assistant_model` or `prompt_lookup_num_tokens` is passed to `.generate()`
 
     To learn more about decoding strategies refer to the [text generation strategies guide](../generation_strategies).