transformers 5.0.0__py3-none-any.whl → 5.0.0rc0__py3-none-any.whl

transformers/data/data_collator.py

@@ -711,6 +711,9 @@ class DataCollatorForLanguageModeling(DataCollatorMixin):
         if self.random_replace_prob < 0 or self.random_replace_prob > 1:
             raise ValueError("random_replace_prob should be between 0 and 1.")
 
+        self.mask_replace_prob = float(self.mask_replace_prob)
+        self.random_replace_prob = float(self.random_replace_prob)
+
         if self.whole_word_mask:
             if not self.tokenizer.is_fast:
                 warnings.warn(
@@ -726,9 +729,6 @@ class DataCollatorForLanguageModeling(DataCollatorMixin):
                 self.mask_replace_prob = 1
                 self.random_replace_prob = 0
 
-        self.mask_replace_prob = float(self.mask_replace_prob)
-        self.random_replace_prob = float(self.random_replace_prob)
-
         self.generator = None
 
     def get_generator(self, seed):
@@ -1413,17 +1413,9 @@ class DataCollatorWithFlattening(DefaultDataCollator):
             max_length = 0
         for seq_idx, sample in enumerate(features):
             input_ids = sample["input_ids"]
-            # Convert to list if tensor
-            if hasattr(input_ids, "tolist"):
-                input_ids = input_ids.tolist()
             batch["input_ids"] += input_ids
-
             if is_labels_provided:
-                labels = sample["labels"]
-                # Convert to list if tensor
-                if hasattr(labels, "tolist"):
-                    labels = labels.tolist()
-                batch["labels"] += [separator_id] + labels[1:]
+                batch["labels"] += [separator_id] + sample["labels"][1:]
             else:
                 batch["labels"] += [separator_id] + input_ids[1:]
             if self.return_position_ids:

transformers/data/processors/glue.py

@@ -1,3 +1,4 @@
+# coding=utf-8
 # Copyright 2018 The Google AI Language Team Authors and The HuggingFace Inc. team.
 # Copyright (c) 2018, NVIDIA CORPORATION.  All rights reserved.
 #

transformers/data/processors/utils.py

@@ -1,3 +1,4 @@
+# coding=utf-8
 # Copyright 2018 The Google AI Language Team Authors and The HuggingFace Inc. team.
 # Copyright (c) 2018, NVIDIA CORPORATION.  All rights reserved.
 #

transformers/data/processors/xnli.py

@@ -1,3 +1,4 @@
+# coding=utf-8
 # Copyright 2018 The Google AI Language Team Authors and The HuggingFace Inc. team.
 # Copyright (c) 2018, NVIDIA CORPORATION.  All rights reserved.
 #

transformers/dependency_versions_check.py

@@ -26,6 +26,7 @@ pkgs_to_check_at_runtime = [
     "python",
     "tqdm",
     "regex",
+    "requests",
     "packaging",
     "filelock",
     "numpy",

transformers/dependency_versions_table.py

@@ -1,6 +1,6 @@
 # THIS FILE HAS BEEN AUTOGENERATED. To update:
 # 1. modify the `_deps` dict in setup.py
-# 2. run `make fix-repo``
+# 2. run `make deps_table_update``
 deps = {
     "Pillow": "Pillow>=10.0.1,<=15.0",
     "accelerate": "accelerate>=1.1.0",
@@ -9,6 +9,7 @@ deps = {
     "blobfile": "blobfile",
     "codecarbon": "codecarbon>=2.8.1",
     "cookiecutter": "cookiecutter==1.7.3",
+    "dataclasses": "dataclasses",
     "datasets": "datasets>=2.15.0",
     "deepspeed": "deepspeed>=0.9.3",
     "diffusers": "diffusers",
@@ -22,7 +23,7 @@ deps = {
     "GitPython": "GitPython<3.1.19",
     "hf-doc-builder": "hf-doc-builder>=0.3.0",
     "hf_xet": "hf_xet",
-    "huggingface-hub": "huggingface-hub>=1.3.0,<2.0",
+    "huggingface-hub": "huggingface-hub>=1.0.0,<2.0",
     "importlib_metadata": "importlib_metadata",
     "ipadic": "ipadic>=1.0.0,<2.0",
     "jinja2": "jinja2>=3.1.0",
@@ -56,10 +57,11 @@ deps = {
     "python": "python>=3.10.0",
     "ray[tune]": "ray[tune]>=2.7.0",
     "regex": "regex!=2019.12.17",
+    "requests": "requests",
     "rhoknp": "rhoknp>=1.1.0,<1.3.1",
     "rjieba": "rjieba",
     "rouge-score": "rouge-score!=0.0.7,!=0.0.8,!=0.1,!=0.1.1",
-    "ruff": "ruff==0.14.10",
+    "ruff": "ruff==0.13.1",
     "sacrebleu": "sacrebleu>=1.4.12,<2.0.0",
     "sacremoses": "sacremoses",
     "safetensors": "safetensors>=0.4.3",
@@ -74,7 +76,7 @@ deps = {
     "tensorboard": "tensorboard",
     "timeout-decorator": "timeout-decorator",
     "tiktoken": "tiktoken",
-    "timm": "timm>=1.0.23",
+    "timm": "timm<=1.0.19,!=1.0.18",
     "tokenizers": "tokenizers>=0.22.0,<=0.23.0",
     "torch": "torch>=2.2",
     "torchaudio": "torchaudio",
@@ -90,5 +92,5 @@ deps = {
     "libcst": "libcst",
     "rich": "rich",
     "opentelemetry-api": "opentelemetry-api",
-    "mistral-common[image]": "mistral-common[image]>=1.8.8",
+    "mistral-common[opencv]": "mistral-common[opencv]>=1.6.3",
 }

transformers/distributed/configuration_utils.py

@@ -75,7 +75,8 @@ class DistributedConfig:
     # Copied from transformers.utils.quantization_config.QuantizationConfigMixin.__iter__
     def __iter__(self):
         """allows `dict(obj)` for situations where obj may be a dict or QuantizationConfigMixin"""
-        yield from copy.deepcopy(self.__dict__).items()
+        for attr, value in copy.deepcopy(self.__dict__).items():
+            yield attr, value
 
     # Copied from transformers.utils.quantization_config.QuantizationConfigMixin.__repr__
     def __repr__(self):

transformers/dynamic_module_utils.py

@@ -28,9 +28,9 @@ import sys
 import threading
 from pathlib import Path
 from types import ModuleType
-from typing import Any
+from typing import Any, Optional, Union
 
-from huggingface_hub import is_offline_mode, try_to_load_from_cache
+from huggingface_hub import try_to_load_from_cache
 from packaging import version
 
 from .utils import (
@@ -38,6 +38,7 @@ from .utils import (
     TRANSFORMERS_DYNAMIC_MODULE_NAME,
     cached_file,
     extract_commit_hash,
+    is_offline_mode,
     logging,
 )
 from .utils.import_utils import VersionComparison, split_package_version
@@ -98,7 +99,7 @@ def init_hf_modules():
         importlib.invalidate_caches()
 
 
-def create_dynamic_module(name: str | os.PathLike) -> None:
+def create_dynamic_module(name: Union[str, os.PathLike]) -> None:
     """
     Creates a dynamic module in the cache directory for modules.
 
@@ -120,7 +121,7 @@ def create_dynamic_module(name: str | os.PathLike) -> None:
         importlib.invalidate_caches()
 
 
-def get_relative_imports(module_file: str | os.PathLike) -> list[str]:
+def get_relative_imports(module_file: Union[str, os.PathLike]) -> list[str]:
     """
     Get the list of modules that are relatively imported in a module file.
 
@@ -141,7 +142,7 @@ def get_relative_imports(module_file: str | os.PathLike) -> list[str]:
     return list(set(relative_imports))
 
 
-def get_relative_import_files(module_file: str | os.PathLike) -> list[str]:
+def get_relative_import_files(module_file: Union[str, os.PathLike]) -> list[str]:
     """
     Get the list of all files that are needed for a given module. Note that this function recurses through the relative
     imports (if a imports b and b imports c, it will return module files for b and c).
@@ -173,7 +174,7 @@ def get_relative_import_files(module_file: str | os.PathLike) -> list[str]:
     return all_relative_imports
 
 
-def get_imports(filename: str | os.PathLike) -> list[str]:
+def get_imports(filename: Union[str, os.PathLike]) -> list[str]:
     """
     Extracts all the libraries (not relative imports this time) that are imported in a file.
 
@@ -228,7 +229,7 @@ def get_imports(filename: str | os.PathLike) -> list[str]:
     return sorted(imported_modules)
 
 
-def check_imports(filename: str | os.PathLike) -> list[str]:
+def check_imports(filename: Union[str, os.PathLike]) -> list[str]:
     """
     Check if the current Python environment contains all the libraries that are imported in a file. Will raise if a
     library is missing.
@@ -265,7 +266,7 @@ def check_imports(filename: str | os.PathLike) -> list[str]:
 
 def get_class_in_module(
     class_name: str,
-    module_path: str | os.PathLike,
+    module_path: Union[str, os.PathLike],
     *,
     force_reload: bool = False,
 ) -> type:
@@ -290,7 +291,7 @@ def get_class_in_module(
         if force_reload:
             sys.modules.pop(name, None)
             importlib.invalidate_caches()
-        cached_module: ModuleType | None = sys.modules.get(name)
+        cached_module: Optional[ModuleType] = sys.modules.get(name)
         module_spec = importlib.util.spec_from_file_location(name, location=module_file)
 
         # Hash the module file and all its relative imports to check if we need to reload it
@@ -312,16 +313,16 @@ def get_class_in_module(
 
 
 def get_cached_module_file(
-    pretrained_model_name_or_path: str | os.PathLike,
+    pretrained_model_name_or_path: Union[str, os.PathLike],
     module_file: str,
-    cache_dir: str | os.PathLike | None = None,
+    cache_dir: Optional[Union[str, os.PathLike]] = None,
     force_download: bool = False,
-    proxies: dict[str, str] | None = None,
-    token: bool | str | None = None,
-    revision: str | None = None,
+    proxies: Optional[dict[str, str]] = None,
+    token: Optional[Union[bool, str]] = None,
+    revision: Optional[str] = None,
     local_files_only: bool = False,
-    repo_type: str | None = None,
-    _commit_hash: str | None = None,
+    repo_type: Optional[str] = None,
+    _commit_hash: Optional[str] = None,
     **deprecated_kwargs,
 ) -> str:
     """
@@ -475,15 +476,15 @@ def get_cached_module_file(
 
 def get_class_from_dynamic_module(
     class_reference: str,
-    pretrained_model_name_or_path: str | os.PathLike,
-    cache_dir: str | os.PathLike | None = None,
+    pretrained_model_name_or_path: Union[str, os.PathLike],
+    cache_dir: Optional[Union[str, os.PathLike]] = None,
     force_download: bool = False,
-    proxies: dict[str, str] | None = None,
-    token: bool | str | None = None,
-    revision: str | None = None,
+    proxies: Optional[dict[str, str]] = None,
+    token: Optional[Union[bool, str]] = None,
+    revision: Optional[str] = None,
     local_files_only: bool = False,
-    repo_type: str | None = None,
-    code_revision: str | None = None,
+    repo_type: Optional[str] = None,
+    code_revision: Optional[str] = None,
     **kwargs,
 ) -> type:
     """
@@ -583,7 +584,7 @@ def get_class_from_dynamic_module(
     return get_class_in_module(class_name, final_module, force_reload=force_download)
 
 
-def custom_object_save(obj: Any, folder: str | os.PathLike, config: dict | None = None) -> list[str]:
+def custom_object_save(obj: Any, folder: Union[str, os.PathLike], config: Optional[dict] = None) -> list[str]:
     """
     Save the modeling files corresponding to a custom model/configuration/tokenizer etc. in a given folder. Optionally
     adds the proper fields in a config.

transformers/feature_extraction_sequence_utils.py

@@ -15,6 +15,8 @@
 Sequence feature extraction class for common feature extractors to preprocess sequences.
 """
 
+from typing import Optional, Union
+
 import numpy as np
 
 from .audio_utils import is_valid_audio, load_audio
@@ -50,17 +52,19 @@ class SequenceFeatureExtractor(FeatureExtractionMixin):
 
     def pad(
         self,
-        processed_features: BatchFeature
-        | list[BatchFeature]
-        | dict[str, BatchFeature]
-        | dict[str, list[BatchFeature]]
-        | list[dict[str, BatchFeature]],
-        padding: bool | str | PaddingStrategy = True,
-        max_length: int | None = None,
+        processed_features: Union[
+            BatchFeature,
+            list[BatchFeature],
+            dict[str, BatchFeature],
+            dict[str, list[BatchFeature]],
+            list[dict[str, BatchFeature]],
+        ],
+        padding: Union[bool, str, PaddingStrategy] = True,
+        max_length: Optional[int] = None,
         truncation: bool = False,
-        pad_to_multiple_of: int | None = None,
-        return_attention_mask: bool | None = None,
-        return_tensors: str | TensorType | None = None,
+        pad_to_multiple_of: Optional[int] = None,
+        return_attention_mask: Optional[bool] = None,
+        return_tensors: Optional[Union[str, TensorType]] = None,
     ) -> BatchFeature:
         """
         Pad input values / input vectors or a batch of input values / input vectors up to predefined length or to the
@@ -220,11 +224,11 @@ class SequenceFeatureExtractor(FeatureExtractionMixin):
 
     def _pad(
         self,
-        processed_features: dict[str, np.ndarray] | BatchFeature,
-        max_length: int | None = None,
+        processed_features: Union[dict[str, np.ndarray], BatchFeature],
+        max_length: Optional[int] = None,
         padding_strategy: PaddingStrategy = PaddingStrategy.DO_NOT_PAD,
-        pad_to_multiple_of: int | None = None,
-        return_attention_mask: bool | None = None,
+        pad_to_multiple_of: Optional[int] = None,
+        return_attention_mask: Optional[bool] = None,
     ) -> dict:
         """
         Pad inputs (on left/right and up to predefined length or max length in the batch)
@@ -292,10 +296,10 @@ class SequenceFeatureExtractor(FeatureExtractionMixin):
 
     def _truncate(
         self,
-        processed_features: dict[str, np.ndarray] | BatchFeature,
-        max_length: int | None = None,
-        pad_to_multiple_of: int | None = None,
-        truncation: bool | None = None,
+        processed_features: Union[dict[str, np.ndarray], BatchFeature],
+        max_length: Optional[int] = None,
+        pad_to_multiple_of: Optional[int] = None,
+        truncation: Optional[bool] = None,
     ):
         """
         Truncate inputs to predefined length or max length in the batch
@@ -365,7 +369,7 @@ class SequenceFeatureExtractor(FeatureExtractionMixin):
 
         return padding_strategy
 
-    def fetch_audio(self, audio_url_or_urls: str | list[str] | list[list[str]]):
+    def fetch_audio(self, audio_url_or_urls: Union[str, list[str], list[list[str]]]):
         """
         Convert a single or a list of urls into the corresponding `np.ndarray` objects.
 

transformers/feature_extraction_utils.py

@@ -19,10 +19,10 @@ import copy
 import json
 import os
 from collections import UserDict
-from typing import TYPE_CHECKING, Any, TypeVar, Union
+from typing import TYPE_CHECKING, Any, Optional, TypeVar, Union
 
 import numpy as np
-from huggingface_hub import create_repo, is_offline_mode
+from huggingface_hub import create_repo
 
 from .dynamic_module_utils import custom_object_save
 from .utils import (
@@ -30,9 +30,9 @@ from .utils import (
     PROCESSOR_NAME,
     PushToHubMixin,
     TensorType,
-    _is_tensor_or_array_like,
     copy_func,
     is_numpy_array,
+    is_offline_mode,
     is_torch_available,
     is_torch_device,
     is_torch_dtype,
@@ -68,18 +68,11 @@ class BatchFeature(UserDict):
         tensor_type (`Union[None, str, TensorType]`, *optional*):
             You can give a tensor_type here to convert the lists of integers in PyTorch/Numpy Tensors at
             initialization.
-        skip_tensor_conversion (`list[str]` or `set[str]`, *optional*):
-            List or set of keys that should NOT be converted to tensors, even when `tensor_type` is specified.
     """
 
-    def __init__(
-        self,
-        data: dict[str, Any] | None = None,
-        tensor_type: None | str | TensorType = None,
-        skip_tensor_conversion: list[str] | set[str] | None = None,
-    ):
+    def __init__(self, data: Optional[dict[str, Any]] = None, tensor_type: Union[None, str, TensorType] = None):
         super().__init__(data)
-        self.convert_to_tensors(tensor_type=tensor_type, skip_tensor_conversion=skip_tensor_conversion)
+        self.convert_to_tensors(tensor_type=tensor_type)
 
     def __getitem__(self, item: str) -> Any:
         """
@@ -104,7 +97,7 @@ class BatchFeature(UserDict):
         if "data" in state:
             self.data = state["data"]
 
-    def _get_is_as_tensor_fns(self, tensor_type: str | TensorType | None = None):
+    def _get_is_as_tensor_fns(self, tensor_type: Optional[Union[str, TensorType]] = None):
         if tensor_type is None:
             return None, None
 
@@ -118,14 +111,6 @@ class BatchFeature(UserDict):
             import torch
 
             def as_tensor(value):
-                if torch.is_tensor(value):
-                    return value
-
-                # stack list of tensors if tensor_type is PyTorch (# torch.tensor() does not support list of tensors)
-                if isinstance(value, (list, tuple)) and len(value) > 0 and torch.is_tensor(value[0]):
-                    return torch.stack(value)
-
-                # convert list of numpy arrays to numpy array (stack) if tensor_type is Numpy
                 if isinstance(value, (list, tuple)) and len(value) > 0:
                     if isinstance(value[0], np.ndarray):
                         value = np.array(value)
@@ -154,11 +139,7 @@ class BatchFeature(UserDict):
             is_tensor = is_numpy_array
         return is_tensor, as_tensor
 
-    def convert_to_tensors(
-        self,
-        tensor_type: str | TensorType | None = None,
-        skip_tensor_conversion: list[str] | set[str] | None = None,
-    ):
+    def convert_to_tensors(self, tensor_type: Optional[Union[str, TensorType]] = None):
         """
         Convert the inner content to tensors.
 
@@ -166,13 +147,6 @@ class BatchFeature(UserDict):
             tensor_type (`str` or [`~utils.TensorType`], *optional*):
                 The type of tensors to use. If `str`, should be one of the values of the enum [`~utils.TensorType`]. If
                 `None`, no modification is done.
-            skip_tensor_conversion (`list[str]` or `set[str]`, *optional*):
-                List or set of keys that should NOT be converted to tensors, even when `tensor_type` is specified.
-
-        Note:
-            Values that don't have an array-like structure (e.g., strings, dicts, lists of strings) are
-            automatically skipped and won't be converted to tensors. Ragged arrays (lists of arrays with
-            different lengths) are still attempted, though they may raise errors during conversion.
         """
         if tensor_type is None:
             return self
@@ -181,30 +155,18 @@ class BatchFeature(UserDict):
 
         # Do the tensor conversion in batch
         for key, value in self.items():
-            # Skip keys explicitly marked for no conversion
-            if skip_tensor_conversion and key in skip_tensor_conversion:
-                continue
-
-            # Skip values that are not array-like
-            if not _is_tensor_or_array_like(value):
-                continue
-
             try:
                 if not is_tensor(value):
                     tensor = as_tensor(value)
+
                     self[key] = tensor
-            except Exception as e:
+            except:  # noqa E722
                 if key == "overflowing_values":
-                    raise ValueError(
-                        f"Unable to create tensor for '{key}' with overflowing values of different lengths. "
-                        f"Original error: {str(e)}"
-                    ) from e
+                    raise ValueError("Unable to create tensor returning overflowing values of different lengths. ")
                 raise ValueError(
-                    f"Unable to convert output '{key}' (type: {type(value).__name__}) to tensor: {str(e)}\n"
-                    f"You can try:\n"
-                    f"  1. Use padding=True to ensure all outputs have the same shape\n"
-                    f"  2. Set return_tensors=None to return Python objects instead of tensors"
-                ) from e
+                    "Unable to create tensor, you should probably activate padding "
+                    "with 'padding=True' to have batched tensors with the same length."
+                )
 
         return self
 
@@ -243,15 +205,12 @@ class BatchFeature(UserDict):
 
         # We cast only floating point tensors to avoid issues with tokenizers casting `LongTensor` to `FloatTensor`
         def maybe_to(v):
-            # check if v is a floating point tensor
+            # check if v is a floating point
             if isinstance(v, torch.Tensor) and torch.is_floating_point(v):
                 # cast and send to device
                 return v.to(*args, **kwargs)
             elif isinstance(v, torch.Tensor) and device is not None:
                 return v.to(device=device, non_blocking=non_blocking)
-            # recursively handle lists and tuples
-            elif isinstance(v, (list, tuple)):
-                return type(v)(maybe_to(item) for item in v)
             else:
                 return v
 
@@ -269,8 +228,8 @@ class FeatureExtractionMixin(PushToHubMixin):
 
     def __init__(self, **kwargs):
         """Set elements of `kwargs` as attributes."""
-        # Pop "processor_class", it should not be saved in feature extractor config
-        kwargs.pop("processor_class", None)
+        # Pop "processor_class" as it should be saved as private attribute
+        self._processor_class = kwargs.pop("processor_class", None)
         # Additional attributes without default values
         for key, value in kwargs.items():
             try:
@@ -279,14 +238,18 @@ class FeatureExtractionMixin(PushToHubMixin):
                 logger.error(f"Can't set {key} with value {value} for {self}")
                 raise err
 
+    def _set_processor_class(self, processor_class: str):
+        """Sets processor class as an attribute."""
+        self._processor_class = processor_class
+
     @classmethod
     def from_pretrained(
         cls: type[SpecificFeatureExtractorType],
-        pretrained_model_name_or_path: str | os.PathLike,
-        cache_dir: str | os.PathLike | None = None,
+        pretrained_model_name_or_path: Union[str, os.PathLike],
+        cache_dir: Optional[Union[str, os.PathLike]] = None,
         force_download: bool = False,
         local_files_only: bool = False,
-        token: str | bool | None = None,
+        token: Optional[Union[str, bool]] = None,
         revision: str = "main",
         **kwargs,
     ) -> SpecificFeatureExtractorType:
@@ -376,7 +339,7 @@ class FeatureExtractionMixin(PushToHubMixin):
 
         return cls.from_dict(feature_extractor_dict, **kwargs)
 
-    def save_pretrained(self, save_directory: str | os.PathLike, push_to_hub: bool = False, **kwargs):
+    def save_pretrained(self, save_directory: Union[str, os.PathLike], push_to_hub: bool = False, **kwargs):
         """
         Save a feature_extractor object to the directory `save_directory`, so that it can be re-loaded using the
         [`~feature_extraction_utils.FeatureExtractionMixin.from_pretrained`] class method.
@@ -426,7 +389,7 @@ class FeatureExtractionMixin(PushToHubMixin):
 
     @classmethod
     def get_feature_extractor_dict(
-        cls, pretrained_model_name_or_path: str | os.PathLike, **kwargs
+        cls, pretrained_model_name_or_path: Union[str, os.PathLike], **kwargs
     ) -> tuple[dict[str, Any], dict[str, Any]]:
         """
         From a `pretrained_model_name_or_path`, resolve to a dictionary of parameters, to be used for instantiating a
@@ -591,7 +554,7 @@ class FeatureExtractionMixin(PushToHubMixin):
         return output
 
     @classmethod
-    def from_json_file(cls, json_file: str | os.PathLike) -> "FeatureExtractionMixin":
+    def from_json_file(cls, json_file: Union[str, os.PathLike]) -> "FeatureExtractionMixin":
         """
         Instantiates a feature extractor of type [`~feature_extraction_utils.FeatureExtractionMixin`] from the path to
         a JSON file of parameters.
@@ -622,9 +585,15 @@ class FeatureExtractionMixin(PushToHubMixin):
             if isinstance(value, np.ndarray):
                 dictionary[key] = value.tolist()
 
+        # make sure private name "_processor_class" is correctly
+        # saved as "processor_class"
+        _processor_class = dictionary.pop("_processor_class", None)
+        if _processor_class is not None:
+            dictionary["processor_class"] = _processor_class
+
         return json.dumps(dictionary, indent=2, sort_keys=True) + "\n"
 
-    def to_json_file(self, json_file_path: str | os.PathLike):
+    def to_json_file(self, json_file_path: Union[str, os.PathLike]):
         """
         Save this instance to a JSON file.
 

transformers/file_utils.py

@@ -68,6 +68,7 @@ from .utils import (
     is_in_notebook,
     is_ipex_available,
     is_librosa_available,
+    is_offline_mode,
     is_onnx_available,
     is_pandas_available,
     is_phonemizer_available,

transformers/generation/__init__.py

@@ -86,11 +86,7 @@ else:
         "StopStringCriteria",
     ]
     _import_structure["continuous_batching"] = [
-        "ContinuousBatchingManager",
         "ContinuousMixin",
-        "FIFOScheduler",
-        "PrefillFirstScheduler",
-        "Scheduler",
     ]
     _import_structure["utils"] = [
         "GenerationMixin",
@@ -131,13 +127,7 @@ if TYPE_CHECKING:
             EarlyExitCandidateGenerator,
             PromptLookupCandidateGenerator,
         )
-        from .continuous_batching import (
-            ContinuousBatchingManager,
-            ContinuousMixin,
-            FIFOScheduler,
-            PrefillFirstScheduler,
-            Scheduler,
-        )
+        from .continuous_batching import ContinuousMixin
         from .logits_process import (
             AlternatingCodebooksLogitsProcessor,
             ClassifierFreeGuidanceLogitsProcessor,