huggingface-hub 0.21.2__py3-none-any.whl → 0.22.0__py3-none-any.whl

huggingface_hub/inference/_generated/types/__init__.py

@@ -0,0 +1,115 @@
+# This file is auto-generated by `utils/generate_inference_types.py`.
+# Do not modify it manually.
+#
+# ruff: noqa: F401
+
+from .audio_classification import (
+    AudioClassificationInput,
+    AudioClassificationOutputElement,
+    AudioClassificationParameters,
+)
+from .audio_to_audio import AudioToAudioInput, AudioToAudioOutputElement
+from .automatic_speech_recognition import (
+    AutomaticSpeechRecognitionGenerationParameters,
+    AutomaticSpeechRecognitionInput,
+    AutomaticSpeechRecognitionOutput,
+    AutomaticSpeechRecognitionOutputChunk,
+    AutomaticSpeechRecognitionParameters,
+)
+from .base import BaseInferenceType
+from .chat_completion import (
+    ChatCompletionInput,
+    ChatCompletionInputMessage,
+    ChatCompletionOutput,
+    ChatCompletionOutputChoice,
+    ChatCompletionOutputChoiceMessage,
+    ChatCompletionStreamOutput,
+    ChatCompletionStreamOutputChoice,
+    ChatCompletionStreamOutputDelta,
+)
+from .depth_estimation import DepthEstimationInput, DepthEstimationOutput
+from .document_question_answering import (
+    DocumentQuestionAnsweringInput,
+    DocumentQuestionAnsweringInputData,
+    DocumentQuestionAnsweringOutputElement,
+    DocumentQuestionAnsweringParameters,
+)
+from .feature_extraction import FeatureExtractionInput
+from .fill_mask import FillMaskInput, FillMaskOutputElement, FillMaskParameters
+from .image_classification import (
+    ImageClassificationInput,
+    ImageClassificationOutputElement,
+    ImageClassificationParameters,
+)
+from .image_segmentation import ImageSegmentationInput, ImageSegmentationOutputElement, ImageSegmentationParameters
+from .image_to_image import ImageToImageInput, ImageToImageOutput, ImageToImageParameters, ImageToImageTargetSize
+from .image_to_text import ImageToTextGenerationParameters, ImageToTextInput, ImageToTextOutput, ImageToTextParameters
+from .object_detection import (
+    ObjectDetectionBoundingBox,
+    ObjectDetectionInput,
+    ObjectDetectionOutputElement,
+    ObjectDetectionParameters,
+)
+from .question_answering import (
+    QuestionAnsweringInput,
+    QuestionAnsweringInputData,
+    QuestionAnsweringOutputElement,
+    QuestionAnsweringParameters,
+)
+from .sentence_similarity import SentenceSimilarityInput, SentenceSimilarityInputData
+from .summarization import SummarizationGenerationParameters, SummarizationInput, SummarizationOutput
+from .table_question_answering import (
+    TableQuestionAnsweringInput,
+    TableQuestionAnsweringInputData,
+    TableQuestionAnsweringOutputElement,
+)
+from .text2text_generation import Text2TextGenerationInput, Text2TextGenerationOutput, Text2TextGenerationParameters
+from .text_classification import TextClassificationInput, TextClassificationOutputElement, TextClassificationParameters
+from .text_generation import (
+    TextGenerationInput,
+    TextGenerationOutput,
+    TextGenerationOutputDetails,
+    TextGenerationOutputSequenceDetails,
+    TextGenerationOutputToken,
+    TextGenerationParameters,
+    TextGenerationPrefillToken,
+    TextGenerationStreamDetails,
+    TextGenerationStreamOutput,
+)
+from .text_to_audio import TextToAudioGenerationParameters, TextToAudioInput, TextToAudioOutput, TextToAudioParameters
+from .text_to_image import TextToImageInput, TextToImageOutput, TextToImageParameters, TextToImageTargetSize
+from .token_classification import (
+    TokenClassificationInput,
+    TokenClassificationOutputElement,
+    TokenClassificationParameters,
+)
+from .translation import TranslationGenerationParameters, TranslationInput, TranslationOutput
+from .video_classification import (
+    VideoClassificationInput,
+    VideoClassificationOutputElement,
+    VideoClassificationParameters,
+)
+from .visual_question_answering import (
+    VisualQuestionAnsweringInput,
+    VisualQuestionAnsweringInputData,
+    VisualQuestionAnsweringOutputElement,
+    VisualQuestionAnsweringParameters,
+)
+from .zero_shot_classification import (
+    ZeroShotClassificationInput,
+    ZeroShotClassificationInputData,
+    ZeroShotClassificationOutputElement,
+    ZeroShotClassificationParameters,
+)
+from .zero_shot_image_classification import (
+    ZeroShotImageClassificationInput,
+    ZeroShotImageClassificationInputData,
+    ZeroShotImageClassificationOutputElement,
+    ZeroShotImageClassificationParameters,
+)
+from .zero_shot_object_detection import (
+    ZeroShotObjectDetectionBoundingBox,
+    ZeroShotObjectDetectionInput,
+    ZeroShotObjectDetectionInputData,
+    ZeroShotObjectDetectionOutputElement,
+)

huggingface_hub/inference/_generated/types/audio_classification.py

@@ -0,0 +1,43 @@
+# Inference code generated from the JSON schema spec in @huggingface/tasks.
+#
+# See:
+#   - script: https://github.com/huggingface/huggingface.js/blob/main/packages/tasks/scripts/inference-codegen.ts
+#   - specs:  https://github.com/huggingface/huggingface.js/tree/main/packages/tasks/src/tasks.
+from dataclasses import dataclass
+from typing import Any, Literal, Optional
+
+from .base import BaseInferenceType
+
+
+ClassificationOutputTransform = Literal["sigmoid", "softmax", "none"]
+
+
+@dataclass
+class AudioClassificationParameters(BaseInferenceType):
+    """Additional inference parameters
+    Additional inference parameters for Audio Classification
+    """
+
+    function_to_apply: Optional["ClassificationOutputTransform"] = None
+    top_k: Optional[int] = None
+    """When specified, limits the output to the top K most probable classes."""
+
+
+@dataclass
+class AudioClassificationInput(BaseInferenceType):
+    """Inputs for Audio Classification inference"""
+
+    inputs: Any
+    """The input audio data"""
+    parameters: Optional[AudioClassificationParameters] = None
+    """Additional inference parameters"""
+
+
+@dataclass
+class AudioClassificationOutputElement(BaseInferenceType):
+    """Outputs for Audio Classification inference"""
+
+    label: str
+    """The predicted class label."""
+    score: float
+    """The corresponding probability."""

huggingface_hub/inference/_generated/types/audio_to_audio.py

@@ -0,0 +1,31 @@
+# Inference code generated from the JSON schema spec in @huggingface/tasks.
+#
+# See:
+#   - script: https://github.com/huggingface/huggingface.js/blob/main/packages/tasks/scripts/inference-codegen.ts
+#   - specs:  https://github.com/huggingface/huggingface.js/tree/main/packages/tasks/src/tasks.
+from dataclasses import dataclass
+from typing import Any
+
+from .base import BaseInferenceType
+
+
+@dataclass
+class AudioToAudioInput(BaseInferenceType):
+    """Inputs for Audio to Audio inference"""
+
+    inputs: Any
+    """The input audio data"""
+
+
+@dataclass
+class AudioToAudioOutputElement(BaseInferenceType):
+    """Outputs of inference for the Audio To Audio task
+    A generated audio file with its label.
+    """
+
+    blob: Any
+    """The generated audio file."""
+    content_type: str
+    """The content type of audio file."""
+    label: str
+    """The label of the audio file."""

huggingface_hub/inference/_generated/types/automatic_speech_recognition.py

@@ -0,0 +1,116 @@
+# Inference code generated from the JSON schema spec in @huggingface/tasks.
+#
+# See:
+#   - script: https://github.com/huggingface/huggingface.js/blob/main/packages/tasks/scripts/inference-codegen.ts
+#   - specs:  https://github.com/huggingface/huggingface.js/tree/main/packages/tasks/src/tasks.
+from dataclasses import dataclass
+from typing import Any, List, Literal, Optional, Union
+
+from .base import BaseInferenceType
+
+
+EarlyStoppingEnum = Literal["never"]
+
+
+@dataclass
+class AutomaticSpeechRecognitionGenerationParameters(BaseInferenceType):
+    """Parametrization of the text generation process
+    Ad-hoc parametrization of the text generation process
+    """
+
+    do_sample: Optional[bool] = None
+    """Whether to use sampling instead of greedy decoding when generating new tokens."""
+    early_stopping: Optional[Union[bool, "EarlyStoppingEnum"]] = None
+    """Controls the stopping condition for beam-based methods."""
+    epsilon_cutoff: Optional[float] = None
+    """If set to float strictly between 0 and 1, only tokens with a conditional probability
+    greater than epsilon_cutoff will be sampled. In the paper, suggested values range from
+    3e-4 to 9e-4, depending on the size of the model. See [Truncation Sampling as Language
+    Model Desmoothing](https://hf.co/papers/2210.15191) for more details.
+    """
+    eta_cutoff: Optional[float] = None
+    """Eta sampling is a hybrid of locally typical sampling and epsilon sampling. If set to
+    float strictly between 0 and 1, a token is only considered if it is greater than either
+    eta_cutoff or sqrt(eta_cutoff) * exp(-entropy(softmax(next_token_logits))). The latter
+    term is intuitively the expected next token probability, scaled by sqrt(eta_cutoff). In
+    the paper, suggested values range from 3e-4 to 2e-3, depending on the size of the model.
+    See [Truncation Sampling as Language Model Desmoothing](https://hf.co/papers/2210.15191)
+    for more details.
+    """
+    max_length: Optional[int] = None
+    """The maximum length (in tokens) of the generated text, including the input."""
+    max_new_tokens: Optional[int] = None
+    """The maximum number of tokens to generate. Takes precedence over maxLength."""
+    min_length: Optional[int] = None
+    """The minimum length (in tokens) of the generated text, including the input."""
+    min_new_tokens: Optional[int] = None
+    """The minimum number of tokens to generate. Takes precedence over maxLength."""
+    num_beam_groups: Optional[int] = None
+    """Number of groups to divide num_beams into in order to ensure diversity among different
+    groups of beams. See [this paper](https://hf.co/papers/1610.02424) for more details.
+    """
+    num_beams: Optional[int] = None
+    """Number of beams to use for beam search."""
+    penalty_alpha: Optional[float] = None
+    """The value balances the model confidence and the degeneration penalty in contrastive
+    search decoding.
+    """
+    temperature: Optional[float] = None
+    """The value used to modulate the next token probabilities."""
+    top_k: Optional[int] = None
+    """The number of highest probability vocabulary tokens to keep for top-k-filtering."""
+    top_p: Optional[float] = None
+    """If set to float < 1, only the smallest set of most probable tokens with probabilities
+    that add up to top_p or higher are kept for generation.
+    """
+    typical_p: Optional[float] = None
+    """Local typicality measures how similar the conditional probability of predicting a target
+    token next is to the expected conditional probability of predicting a random token next,
+    given the partial text already generated. If set to float < 1, the smallest set of the
+    most locally typical tokens with probabilities that add up to typical_p or higher are
+    kept for generation. See [this paper](https://hf.co/papers/2202.00666) for more details.
+    """
+    use_cache: Optional[bool] = None
+    """Whether the model should use the past last key/values attentions to speed up decoding"""
+
+
+@dataclass
+class AutomaticSpeechRecognitionParameters(BaseInferenceType):
+    """Additional inference parameters
+    Additional inference parameters for Automatic Speech Recognition
+    """
+
+    generate: Optional[AutomaticSpeechRecognitionGenerationParameters] = None
+    """Parametrization of the text generation process"""
+    return_timestamps: Optional[bool] = None
+    """Whether to output corresponding timestamps with the generated text"""
+
+
+@dataclass
+class AutomaticSpeechRecognitionInput(BaseInferenceType):
+    """Inputs for Automatic Speech Recognition inference"""
+
+    inputs: Any
+    """The input audio data"""
+    parameters: Optional[AutomaticSpeechRecognitionParameters] = None
+    """Additional inference parameters"""
+
+
+@dataclass
+class AutomaticSpeechRecognitionOutputChunk(BaseInferenceType):
+    text: str
+    """A chunk of text identified by the model"""
+    timestamps: List[float]
+    """The start and end timestamps corresponding with the text"""
+
+
+@dataclass
+class AutomaticSpeechRecognitionOutput(BaseInferenceType):
+    """Outputs of inference for the Automatic Speech Recognition task"""
+
+    text: str
+    """The recognized text."""
+    chunks: Optional[List[AutomaticSpeechRecognitionOutputChunk]] = None
+    """When returnTimestamps is enabled, chunks contains a list of audio chunks identified by
+    the model.
+    """

huggingface_hub/inference/_generated/types/base.py

@@ -0,0 +1,149 @@
+# Copyright 2024 The HuggingFace Team. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+"""Contains a base class for all inference types."""
+
+import inspect
+import json
+import warnings
+from dataclasses import asdict, dataclass
+from typing import Any, Dict, List, Type, TypeVar, Union, get_args
+
+
+T = TypeVar("T", bound="BaseInferenceType")
+
+
+@dataclass
+class BaseInferenceType(dict):
+    """Base class for all inference types.
+
+    Object is a dataclass and a dict for backward compatibility but plan is to remove the dict part in the future.
+
+    Handle parsing from dict, list and json strings in a permissive way to ensure future-compatibility (e.g. all fields
+    are made optional, and non-expected fields are added as dict attributes).
+    """
+
+    @classmethod
+    def parse_obj_as_list(cls: Type[T], data: Union[bytes, str, List, Dict]) -> List[T]:
+        """Alias to parse server response and return a single instance.
+
+        See `parse_obj` for more details.
+        """
+        output = cls.parse_obj(data)
+        if not isinstance(output, list):
+            raise ValueError(f"Invalid input data for {cls}. Expected a list, but got {type(output)}.")
+        return output
+
+    @classmethod
+    def parse_obj_as_instance(cls: Type[T], data: Union[bytes, str, List, Dict]) -> T:
+        """Alias to parse server response and return a single instance.
+
+        See `parse_obj` for more details.
+        """
+        output = cls.parse_obj(data)
+        if isinstance(output, list):
+            raise ValueError(f"Invalid input data for {cls}. Expected a single instance, but got a list.")
+        return output
+
+    @classmethod
+    def parse_obj(cls: Type[T], data: Union[bytes, str, List, Dict]) -> Union[List[T], T]:
+        """Parse server response as a dataclass or list of dataclasses.
+
+        To enable future-compatibility, we want to handle cases where the server return more fields than expected.
+        In such cases, we don't want to raise an error but still create the dataclass object. Remaining fields are
+        added as dict attributes.
+        """
+        # Parse server response (from bytes)
+        if isinstance(data, bytes):
+            data = data.decode()
+        if isinstance(data, str):
+            data = json.loads(data)
+
+        # If a list, parse each item individually
+        if isinstance(data, List):
+            return [cls.parse_obj(d) for d in data]  # type: ignore [misc]
+
+        # At this point, we expect a dict
+        if not isinstance(data, dict):
+            raise ValueError(f"Invalid data type: {type(data)}")
+
+        init_values = {}
+        other_values = {}
+        for key, value in data.items():
+            key = normalize_key(key)
+            if key in cls.__dataclass_fields__ and cls.__dataclass_fields__[key].init:
+                if isinstance(value, dict) or isinstance(value, list):
+                    field_type = cls.__dataclass_fields__[key].type
+
+                    # if `field_type` is a `BaseInferenceType`, parse it
+                    if inspect.isclass(field_type) and issubclass(field_type, BaseInferenceType):
+                        value = field_type.parse_obj(value)
+
+                    # otherwise, recursively parse nested dataclasses (if possible)
+                    # `get_args` returns handle Union and Optional for us
+                    else:
+                        expected_types = get_args(field_type)
+                        for expected_type in expected_types:
+                            if getattr(expected_type, "_name", None) == "List":
+                                expected_type = get_args(expected_type)[
+                                    0
+                                ]  # assume same type for all items in the list
+                            if inspect.isclass(expected_type) and issubclass(expected_type, BaseInferenceType):
+                                value = expected_type.parse_obj(value)
+                                break
+                init_values[key] = value
+            else:
+                other_values[key] = value
+
+        # Make all missing fields default to None
+        # => ensure that dataclass initialization will never fail even if the server does not return all fields.
+        for key in cls.__dataclass_fields__:
+            if key not in init_values:
+                init_values[key] = None
+
+        # Initialize dataclass with expected values
+        item = cls(**init_values)
+
+        # Add remaining fields as dict attributes
+        item.update(other_values)
+        return item
+
+    def __post_init__(self):
+        self.update(asdict(self))
+
+    def __setitem__(self, __key: Any, __value: Any) -> None:
+        # Hacky way to keep dataclass values in sync when dict is updated
+        super().__setitem__(__key, __value)
+        if __key in self.__dataclass_fields__ and getattr(self, __key, None) != __value:
+            self.__setattr__(__key, __value)
+        return
+
+    def __setattr__(self, __name: str, __value: Any) -> None:
+        # Hacky way to keep dict values is sync when dataclass is updated
+        super().__setattr__(__name, __value)
+        if self.get(__name) != __value:
+            self[__name] = __value
+        return
+
+    def __getitem__(self, __key: Any) -> Any:
+        warnings.warn(
+            f"Accessing '{self.__class__.__name__}' values through dict is deprecated and "
+            "will be removed from version '0.25'. Use dataclass attributes instead.",
+            FutureWarning,
+        )
+        return super().__getitem__(__key)
+
+
+def normalize_key(key: str) -> str:
+    # e.g "content-type" -> "content_type", "Accept" -> "accept"
+    return key.replace("-", "_").replace(" ", "_").lower()

huggingface_hub/inference/_generated/types/chat_completion.py

@@ -0,0 +1,106 @@
+# Inference code generated from the JSON schema spec in @huggingface/tasks.
+#
+# See:
+#   - script: https://github.com/huggingface/huggingface.js/blob/main/packages/tasks/scripts/inference-codegen.ts
+#   - specs:  https://github.com/huggingface/huggingface.js/tree/main/packages/tasks/src/tasks.
+from dataclasses import dataclass
+from typing import List, Literal, Optional, Union
+
+from .base import BaseInferenceType
+
+
+ChatCompletionMessageRole = Literal["assistant", "system", "user"]
+
+
+@dataclass
+class ChatCompletionInputMessage(BaseInferenceType):
+    content: str
+    """The content of the message."""
+    role: "ChatCompletionMessageRole"
+
+
+@dataclass
+class ChatCompletionInput(BaseInferenceType):
+    """Inputs for ChatCompletion inference"""
+
+    messages: List[ChatCompletionInputMessage]
+    frequency_penalty: Optional[float] = None
+    """Number between -2.0 and 2.0. Positive values penalize new tokens based on their existing
+    frequency in the text so far, decreasing the model's likelihood to repeat the same line
+    verbatim.
+    """
+    max_tokens: Optional[int] = None
+    """The maximum number of tokens that can be generated in the chat completion."""
+    seed: Optional[int] = None
+    """The random sampling seed."""
+    stop: Optional[Union[List[str], str]] = None
+    """Stop generating tokens if a stop token is generated."""
+    stream: Optional[bool] = None
+    """If set, partial message deltas will be sent."""
+    temperature: Optional[float] = None
+    """The value used to modulate the logits distribution."""
+    top_p: Optional[float] = None
+    """If set to < 1, only the smallest set of most probable tokens with probabilities that add
+    up to `top_p` or higher are kept for generation.
+    """
+
+
+ChatCompletionFinishReason = Literal["length", "eos_token", "stop_sequence"]
+
+
+@dataclass
+class ChatCompletionOutputChoiceMessage(BaseInferenceType):
+    content: str
+    """The content of the chat completion message."""
+    role: "ChatCompletionMessageRole"
+
+
+@dataclass
+class ChatCompletionOutputChoice(BaseInferenceType):
+    finish_reason: "ChatCompletionFinishReason"
+    """The reason why the generation was stopped."""
+    index: int
+    """The index of the choice in the list of choices."""
+    message: ChatCompletionOutputChoiceMessage
+
+
+@dataclass
+class ChatCompletionOutput(BaseInferenceType):
+    """Outputs for Chat Completion inference"""
+
+    choices: List[ChatCompletionOutputChoice]
+    """A list of chat completion choices."""
+    created: int
+    """The Unix timestamp (in seconds) of when the chat completion was created."""
+
+
+@dataclass
+class ChatCompletionStreamOutputDelta(BaseInferenceType):
+    """A chat completion delta generated by streamed model responses."""
+
+    content: Optional[str] = None
+    """The contents of the chunk message."""
+    role: Optional[str] = None
+    """The role of the author of this message."""
+
+
+@dataclass
+class ChatCompletionStreamOutputChoice(BaseInferenceType):
+    delta: ChatCompletionStreamOutputDelta
+    """A chat completion delta generated by streamed model responses."""
+    index: int
+    """The index of the choice in the list of choices."""
+    finish_reason: Optional["ChatCompletionFinishReason"] = None
+    """The reason why the generation was stopped."""
+
+
+@dataclass
+class ChatCompletionStreamOutput(BaseInferenceType):
+    """Chat Completion Stream Output"""
+
+    choices: List[ChatCompletionStreamOutputChoice]
+    """A list of chat completion choices."""
+    created: int
+    """The Unix timestamp (in seconds) of when the chat completion was created. Each chunk has
+    the same timestamp.
+    """

huggingface_hub/inference/_generated/types/depth_estimation.py

@@ -0,0 +1,29 @@
+# Inference code generated from the JSON schema spec in @huggingface/tasks.
+#
+# See:
+#   - script: https://github.com/huggingface/huggingface.js/blob/main/packages/tasks/scripts/inference-codegen.ts
+#   - specs:  https://github.com/huggingface/huggingface.js/tree/main/packages/tasks/src/tasks.
+from dataclasses import dataclass
+from typing import Any, Dict, Optional
+
+from .base import BaseInferenceType
+
+
+@dataclass
+class DepthEstimationInput(BaseInferenceType):
+    """Inputs for Depth Estimation inference"""
+
+    inputs: Any
+    """The input image data"""
+    parameters: Optional[Dict[str, Any]] = None
+    """Additional inference parameters"""
+
+
+@dataclass
+class DepthEstimationOutput(BaseInferenceType):
+    """Outputs of inference for the Depth Estimation task"""
+
+    depth: Any
+    """The predicted depth as an image"""
+    predicted_depth: Any
+    """The predicted depth as a tensor"""

huggingface_hub/inference/_generated/types/document_question_answering.py

@@ -0,0 +1,85 @@
+# Inference code generated from the JSON schema spec in @huggingface/tasks.
+#
+# See:
+#   - script: https://github.com/huggingface/huggingface.js/blob/main/packages/tasks/scripts/inference-codegen.ts
+#   - specs:  https://github.com/huggingface/huggingface.js/tree/main/packages/tasks/src/tasks.
+from dataclasses import dataclass
+from typing import Any, List, Optional, Union
+
+from .base import BaseInferenceType
+
+
+@dataclass
+class DocumentQuestionAnsweringInputData(BaseInferenceType):
+    """One (document, question) pair to answer"""
+
+    image: Any
+    """The image on which the question is asked"""
+    question: str
+    """A question to ask of the document"""
+
+
+@dataclass
+class DocumentQuestionAnsweringParameters(BaseInferenceType):
+    """Additional inference parameters
+    Additional inference parameters for Document Question Answering
+    """
+
+    doc_stride: Optional[int] = None
+    """If the words in the document are too long to fit with the question for the model, it will
+    be split in several chunks with some overlap. This argument controls the size of that
+    overlap.
+    """
+    handle_impossible_answer: Optional[bool] = None
+    """Whether to accept impossible as an answer"""
+    lang: Optional[str] = None
+    """Language to use while running OCR. Defaults to english."""
+    max_answer_len: Optional[int] = None
+    """The maximum length of predicted answers (e.g., only answers with a shorter length are
+    considered).
+    """
+    max_question_len: Optional[int] = None
+    """The maximum length of the question after tokenization. It will be truncated if needed."""
+    max_seq_len: Optional[int] = None
+    """The maximum length of the total sentence (context + question) in tokens of each chunk
+    passed to the model. The context will be split in several chunks (using doc_stride as
+    overlap) if needed.
+    """
+    top_k: Optional[int] = None
+    """The number of answers to return (will be chosen by order of likelihood). Can return less
+    than top_k answers if there are not enough options available within the context.
+    """
+    word_boxes: Optional[List[Union[List[float], str]]] = None
+    """A list of words and bounding boxes (normalized 0->1000). If provided, the inference will
+    skip the OCR step and use the provided bounding boxes instead.
+    """
+
+
+@dataclass
+class DocumentQuestionAnsweringInput(BaseInferenceType):
+    """Inputs for Document Question Answering inference"""
+
+    inputs: DocumentQuestionAnsweringInputData
+    """One (document, question) pair to answer"""
+    parameters: Optional[DocumentQuestionAnsweringParameters] = None
+    """Additional inference parameters"""
+
+
+@dataclass
+class DocumentQuestionAnsweringOutputElement(BaseInferenceType):
+    """Outputs of inference for the Document Question Answering task"""
+
+    answer: str
+    """The answer to the question."""
+    end: int
+    """The end word index of the answer (in the OCR’d version of the input or provided word
+    boxes).
+    """
+    score: float
+    """The probability associated to the answer."""
+    start: int
+    """The start word index of the answer (in the OCR’d version of the input or provided word
+    boxes).
+    """
+    words: List[int]
+    """The index of each word/box pair that is in the answer"""