more and more

Author: Cyrilvallez

src/transformers/__init__.py

@@ -39,12 +39,9 @@
 # so that mypy, pylint or other static linters can recognize them,
 # given that they are not exported using `__all__` in this file.
 from .utils import is_bitsandbytes_available as is_bitsandbytes_available
-from .utils import is_flax_available as is_flax_available
-from .utils import is_keras_nlp_available as is_keras_nlp_available
 from .utils import is_scipy_available as is_scipy_available
 from .utils import is_sentencepiece_available as is_sentencepiece_available
 from .utils import is_speech_available as is_speech_available
-from .utils import is_tensorflow_text_available as is_tensorflow_text_available
 from .utils import is_timm_available as is_timm_available
 from .utils import is_tokenizers_available as is_tokenizers_available
 from .utils import is_torch_available as is_torch_available
@@ -223,8 +220,6 @@
         "is_bitsandbytes_available",
         "is_datasets_available",
         "is_faiss_available",
-        "is_flax_available",
-        "is_keras_nlp_available",
         "is_matplotlib_available",
         "is_phonemizer_available",
         "is_psutil_available",
@@ -236,7 +231,6 @@
         "is_sentencepiece_available",
         "is_sklearn_available",
         "is_speech_available",
-        "is_tensorflow_text_available",
         "is_timm_available",
         "is_tokenizers_available",
         "is_torch_available",
@@ -573,20 +567,6 @@
     from .generation import EpsilonLogitsWarper as EpsilonLogitsWarper
     from .generation import EtaLogitsWarper as EtaLogitsWarper
     from .generation import ExponentialDecayLengthPenalty as ExponentialDecayLengthPenalty
-    from .generation import FlaxForcedBOSTokenLogitsProcessor as FlaxForcedBOSTokenLogitsProcessor
-    from .generation import FlaxForcedEOSTokenLogitsProcessor as FlaxForcedEOSTokenLogitsProcessor
-    from .generation import FlaxForceTokensLogitsProcessor as FlaxForceTokensLogitsProcessor
-    from .generation import FlaxGenerationMixin as FlaxGenerationMixin
-    from .generation import FlaxLogitsProcessor as FlaxLogitsProcessor
-    from .generation import FlaxLogitsProcessorList as FlaxLogitsProcessorList
-    from .generation import FlaxLogitsWarper as FlaxLogitsWarper
-    from .generation import FlaxMinLengthLogitsProcessor as FlaxMinLengthLogitsProcessor
-    from .generation import FlaxSuppressTokensAtBeginLogitsProcessor as FlaxSuppressTokensAtBeginLogitsProcessor
-    from .generation import FlaxSuppressTokensLogitsProcessor as FlaxSuppressTokensLogitsProcessor
-    from .generation import FlaxTemperatureLogitsWarper as FlaxTemperatureLogitsWarper
-    from .generation import FlaxTopKLogitsWarper as FlaxTopKLogitsWarper
-    from .generation import FlaxTopPLogitsWarper as FlaxTopPLogitsWarper
-    from .generation import FlaxWhisperTimeStampLogitsProcessor as FlaxWhisperTimeStampLogitsProcessor
     from .generation import ForcedBOSTokenLogitsProcessor as ForcedBOSTokenLogitsProcessor
     from .generation import ForcedEOSTokenLogitsProcessor as ForcedEOSTokenLogitsProcessor
     from .generation import GenerationConfig as GenerationConfig
@@ -648,18 +628,14 @@
     from .integrations import is_wandb_available as is_wandb_available
     from .integrations.executorch import TorchExportableModuleWithStaticCache as TorchExportableModuleWithStaticCache
     from .integrations.executorch import convert_and_export_with_cache as convert_and_export_with_cache
-    from .keras_callbacks import KerasMetricCallback as KerasMetricCallback
-    from .keras_callbacks import PushToHubCallback as PushToHubCallback
     from .masking_utils import AttentionMaskInterface as AttentionMaskInterface
     from .model_debugging_utils import model_addition_debugger_context as model_addition_debugger_context
 
     # Model Cards
     from .modelcard import ModelCard as ModelCard
-    from .modeling_flax_utils import FlaxPreTrainedModel as FlaxPreTrainedModel
     from .modeling_layers import GradientCheckpointingLayer as GradientCheckpointingLayer
     from .modeling_rope_utils import ROPE_INIT_FUNCTIONS as ROPE_INIT_FUNCTIONS
     from .modeling_rope_utils import dynamic_rope_update as dynamic_rope_update
-
     from .modeling_utils import AttentionInterface as AttentionInterface
     from .modeling_utils import PreTrainedModel as PreTrainedModel
     from .models import *

src/transformers/data/data_collator.py

@@ -685,7 +685,6 @@ class DataCollatorForLanguageModeling(DataCollatorMixin):
     mask_replace_prob: float = 0.8
     random_replace_prob: float = 0.1
     pad_to_multiple_of: Optional[int] = None
-    tf_experimental_compile: bool = False
     return_tensors: str = "pt"
     seed: Optional[int] = None
 
@@ -1293,13 +1292,6 @@ def torch_call(self, examples: list[Union[list[int], Any, dict[str, Any]]]) -> d
         inputs, perm_mask, target_mapping, labels = self.torch_mask_tokens(batch)
         return {"input_ids": inputs, "perm_mask": perm_mask, "target_mapping": target_mapping, "labels": labels}
 
-    def tf_call(self, examples: list[Union[list[int], Any, dict[str, Any]]]) -> dict[str, Any]:
-        if isinstance(examples[0], Mapping):
-            examples = [e["input_ids"] for e in examples]
-        batch = _tf_collate_batch(examples, self.tokenizer)
-        inputs, perm_mask, target_mapping, labels = self.tf_mask_tokens(batch)
-        return {"input_ids": inputs, "perm_mask": perm_mask, "target_mapping": target_mapping, "labels": labels}
-
     def numpy_call(self, examples: list[Union[list[int], Any, dict[str, Any]]]) -> dict[str, Any]:
         if isinstance(examples[0], Mapping):
             examples = [e["input_ids"] for e in examples]

src/transformers/feature_extraction_sequence_utils.py

@@ -20,7 +20,7 @@
 import numpy as np
 
 from .feature_extraction_utils import BatchFeature, FeatureExtractionMixin
-from .utils import PaddingStrategy, TensorType, is_tf_tensor, is_torch_tensor, logging, to_numpy
+from .utils import PaddingStrategy, TensorType, is_torch_tensor, logging, to_numpy
 
 
 logger = logging.get_logger(__name__)
@@ -116,7 +116,6 @@ def pad(
             return_tensors (`str` or [`~utils.TensorType`], *optional*):
                 If set, will return tensors instead of list of python integers. Acceptable values are:
 
-                - `'tf'`: Return TensorFlow `tf.constant` objects.
                 - `'pt'`: Return PyTorch `torch.Tensor` objects.
                 - `'np'`: Return Numpy `np.ndarray` objects.
         """
@@ -145,7 +144,7 @@ def pad(
                 processed_features["attention_mask"] = []
             return processed_features
 
-        # If we have PyTorch/TF tensors or lists as inputs, we cast them as Numpy arrays
+        # If we have PyTorch tensors or lists as inputs, we cast them as Numpy arrays
         # and rebuild them afterwards if no return_tensors is specified
         # Note that we lose the specific device the tensor may be on for PyTorch
 
@@ -159,9 +158,7 @@ def pad(
                 first_element = required_input[index][0]
 
         if return_tensors is None:
-            if is_tf_tensor(first_element):
-                return_tensors = "tf"
-            elif is_torch_tensor(first_element):
+            if is_torch_tensor(first_element):
                 return_tensors = "pt"
             elif isinstance(first_element, (int, float, list, tuple, np.ndarray)):
                 return_tensors = "np"

src/transformers/file_utils.py

@@ -47,9 +47,6 @@
     TORCH_FX_REQUIRED_VERSION,
     TRANSFORMERS_CACHE,
     TRANSFORMERS_DYNAMIC_MODULE_NAME,
-    USE_JAX,
-    USE_TF,
-    USE_TORCH,
     WEIGHTS_INDEX_NAME,
     WEIGHTS_NAME,
     ContextManagers,
@@ -80,7 +77,6 @@
     is_datasets_available,
     is_detectron2_available,
     is_faiss_available,
-    is_flax_available,
     is_ftfy_available,
     is_g2p_en_available,
     is_in_notebook,

src/transformers/generation/__init__.py

@@ -14,7 +14,7 @@
 
 from typing import TYPE_CHECKING
 
-from ..utils import OptionalDependencyNotAvailable, _LazyModule, is_flax_available, is_tf_available, is_torch_available
+from ..utils import OptionalDependencyNotAvailable, _LazyModule, is_torch_available
 
 
 _import_structure = {
@@ -124,7 +124,6 @@
     ]
 
 
-
 if TYPE_CHECKING:
     from .configuration_utils import (
         BaseWatermarkingConfig,

src/transformers/integrations/integration_utils.py

@@ -48,16 +48,13 @@
     flatten_dict,
     is_datasets_available,
     is_pandas_available,
-    is_tf_available,
     is_torch_available,
     logging,
 )
 
 
 logger = logging.get_logger(__name__)
 
-if is_tf_available():
-    from .. import TFPreTrainedModel
 
 if is_torch_available():
     import torch
@@ -760,12 +757,6 @@ def save_model_architecture_to_file(model: Any, output_dir: str):
     with open(f"{output_dir}/model_architecture.txt", "w+") as f:
         if isinstance(model, PreTrainedModel):
             print(model, file=f)
-        elif is_tf_available() and isinstance(model, TFPreTrainedModel):
-
-            def print_to_file(s):
-                print(s, file=f)
-
-            model.summary(print_fn=print_to_file)
         elif is_torch_available() and (
             isinstance(model, (torch.nn.Module, PushToHubMixin)) and hasattr(model, "base_model")
         ):

src/transformers/models/conditional_detr/image_processing_conditional_detr.py

@@ -56,8 +56,6 @@
 from ...utils import (
     TensorType,
     is_scipy_available,
-    is_tf_available,
-    is_tf_tensor,
     is_torch_available,
     is_torch_tensor,
     is_vision_available,

src/transformers/models/data2vec/modeling_data2vec_text.py

@@ -50,7 +50,7 @@
 
 
 # Copied from transformers.models.roberta.modeling_roberta.RobertaEmbeddings with Roberta->Data2VecText
-class Data2VecTextEmbeddings(nn.Module):
+class Data2VecTextForTextEmbeddings(nn.Module):
     """
     Same as BertEmbeddings with a tiny tweak for positional embeddings indexing.
     """

src/transformers/models/decision_transformer/modeling_decision_transformer.py

@@ -15,7 +15,6 @@
 """PyTorch DecisionTransformer model."""
 
 import math
-import os
 from dataclasses import dataclass
 from typing import Callable, Optional, Union
 

src/transformers/models/luke/tokenization_luke.py

@@ -37,7 +37,7 @@
     TruncationStrategy,
     to_py_obj,
 )
-from ...utils import add_end_docstrings, is_tf_tensor, is_torch_tensor, logging
+from ...utils import add_end_docstrings, is_torch_tensor, logging
 
 
 logger = logging.get_logger(__name__)
@@ -1441,7 +1441,6 @@ def pad(
             return_tensors (`str` or [`~utils.TensorType`], *optional*):
                 If set, will return tensors instead of list of python integers. Acceptable values are:
 
-                - `'tf'`: Return TensorFlow `tf.constant` objects.
                 - `'pt'`: Return PyTorch `torch.Tensor` objects.
                 - `'np'`: Return Numpy `np.ndarray` objects.
             verbose (`bool`, *optional*, defaults to `True`):
@@ -1466,7 +1465,7 @@ def pad(
                 encoded_inputs["attention_mask"] = []
             return encoded_inputs
 
-        # If we have PyTorch/TF/NumPy tensors/arrays as inputs, we cast them as python objects
+        # If we have PyTorch/NumPy tensors/arrays as inputs, we cast them as python objects
         # and rebuild them afterwards if no return_tensors is specified
         # Note that we lose the specific device the tensor may be on for PyTorch
 
@@ -1480,9 +1479,7 @@ def pad(
                 first_element = required_input[index][0]
         # At this state, if `first_element` is still a list/tuple, it's an empty one so there is nothing to do.
         if not isinstance(first_element, (int, list, tuple)):
-            if is_tf_tensor(first_element):
-                return_tensors = "tf" if return_tensors is None else return_tensors
-            elif is_torch_tensor(first_element):
+            if is_torch_tensor(first_element):
                 return_tensors = "pt" if return_tensors is None else return_tensors
             elif isinstance(first_element, np.ndarray):
                 return_tensors = "np" if return_tensors is None else return_tensors