TF generate refactor - Greedy Search

docs/source/internal/generation_utils.mdx

@@ -148,6 +148,24 @@ generation.
 [[autodoc]] InfNanRemoveLogitsProcessor
     - __call__
 
+[[autodoc]] TFLogitsProcessor
+    - __call__
+
+[[autodoc]] TFLogitsProcessorList
+    - __call__
+
+[[autodoc]] TFMinLengthLogitsProcessor
+    - __call__
+
+[[autodoc]] TFNoBadWordsLogitsProcessor
+    - __call__
+
+[[autodoc]] TFNoRepeatNGramLogitsProcessor
+    - __call__
+
+[[autodoc]] TFRepetitionPenaltyLogitsProcessor
+    - __call__
+
 [[autodoc]] FlaxLogitsProcessor
     - __call__
 

src/transformers/__init__.py

@@ -1592,6 +1592,14 @@
     _import_structure["activations_tf"] = []
     _import_structure["benchmark.benchmark_args_tf"] = ["TensorFlowBenchmarkArguments"]
     _import_structure["benchmark.benchmark_tf"] = ["TensorFlowBenchmark"]
+    _import_structure["generation_tf_logits_process"] = [
+        "TFLogitsProcessor",
+        "TFLogitsProcessorList",
+        "TFMinLengthLogitsProcessor",
+        "TFNoBadWordsLogitsProcessor",
+        "TFNoRepeatNGramLogitsProcessor",
+        "TFRepetitionPenaltyLogitsProcessor",
+    ]
     _import_structure["generation_tf_utils"] = ["tf_top_k_top_p_filtering"]
     _import_structure["keras_callbacks"] = ["KerasMetricCallback", "PushToHubCallback"]
     _import_structure["modeling_tf_outputs"] = []
@@ -2046,6 +2054,7 @@
         ]
     )
     _import_structure["optimization_tf"] = ["AdamWeightDecay", "GradientAccumulator", "WarmUp", "create_optimizer"]
+    _import_structure["tf_utils"] = []
     _import_structure["trainer_tf"] = ["TFTrainer"]
 
 else:
@@ -3572,6 +3581,14 @@
 
         # Benchmarks
         from .benchmark.benchmark_tf import TensorFlowBenchmark
+        from .generation_tf_logits_process import (
+            TFLogitsProcessor,
+            TFLogitsProcessorList,
+            TFMinLengthLogitsProcessor,
+            TFNoBadWordsLogitsProcessor,
+            TFNoRepeatNGramLogitsProcessor,
+            TFRepetitionPenaltyLogitsProcessor,
+        )
         from .generation_tf_utils import tf_top_k_top_p_filtering
         from .keras_callbacks import KerasMetricCallback, PushToHubCallback
         from .modeling_tf_layoutlm import (

src/transformers/generation_flax_logits_process.py

@@ -14,7 +14,6 @@
 # limitations under the License.
 
 import inspect
-from abc import ABC
 
 import jax
 import jax.lax as lax
@@ -48,7 +47,7 @@
 """
 
 
-class FlaxLogitsProcessor(ABC):
+class FlaxLogitsProcessor:
     """Abstract base class for all logit processors that can be applied during generation."""
 
     @add_start_docstrings(LOGITS_PROCESSOR_INPUTS_DOCSTRING)
@@ -59,7 +58,7 @@ def __call__(self, input_ids: jnp.ndarray, scores: jnp.ndarray) -> jnp.ndarray:
         )
 
 
-class FlaxLogitsWarper(ABC):
+class FlaxLogitsWarper:
     """Abstract base class for all logit warpers that can be applied during generation with multinomial sampling."""
 
     @add_start_docstrings(LOGITS_PROCESSOR_INPUTS_DOCSTRING)

src/transformers/generation_logits_process.py

@@ -15,7 +15,6 @@
 
 import inspect
 import math
-from abc import ABC
 from typing import Callable, Iterable, List, Optional
 
 import numpy as np
@@ -49,7 +48,7 @@
 """
 
 
-class LogitsProcessor(ABC):
+class LogitsProcessor:
     """Abstract base class for all logit processors that can be applied during generation."""
 
     @add_start_docstrings(LOGITS_PROCESSOR_INPUTS_DOCSTRING)
@@ -60,7 +59,7 @@ def __call__(self, input_ids: torch.LongTensor, scores: torch.FloatTensor) -> to
         )
 
 
-class LogitsWarper(ABC):
+class LogitsWarper:
     """Abstract base class for all logit warpers that can be applied during generation with multinomial sampling."""
 
     @add_start_docstrings(LOGITS_PROCESSOR_INPUTS_DOCSTRING)