[Doc] Convert Sphinx directives ( {class}, {meth}, {attr}, ...) to MkDocs format for better documentation linking (#18663)

Signed-off-by: Zerohertz <ohg3417@gmail.com>
This commit is contained in:
Hyogeun Oh (오효근)
2025-05-27 01:44:20 -07:00
committed by GitHub
parent 6881107948
commit a68e293cb9
37 changed files with 360 additions and 247 deletions
+45 -29
View File
@@ -29,14 +29,14 @@ _T = TypeVar("_T")
HfImageItem: TypeAlias = Union["Image", np.ndarray, "torch.Tensor"]
"""
A {class}`transformers.image_utils.ImageInput` representing a single image
A `transformers.image_utils.ImageInput` representing a single image
item, which can be passed to a HuggingFace `ImageProcessor`.
"""
HfVideoItem: TypeAlias = Union[list["Image"], np.ndarray, "torch.Tensor",
list[np.ndarray], list["torch.Tensor"]]
"""
A {class}`transformers.image_utils.VideoInput` representing a single video
A `transformers.image_utils.VideoInput` representing a single video
item, which can be passed to a HuggingFace `VideoProcessor`.
"""
@@ -48,7 +48,7 @@ item, which can be passed to a HuggingFace `AudioProcessor`.
ImageItem: TypeAlias = Union[HfImageItem, "torch.Tensor"]
"""
A {class}`transformers.image_utils.ImageInput` representing a single image
A `transformers.image_utils.ImageInput` representing a single image
item, which can be passed to a HuggingFace `ImageProcessor`.
Alternatively, a 3-D tensor or batch of 2-D tensors,
@@ -58,7 +58,7 @@ these are directly passed to the model without HF processing.
VideoItem: TypeAlias = Union[HfVideoItem, "torch.Tensor"]
"""
A {class}`transformers.image_utils.VideoInput` representing a single video
A `transformers.image_utils.VideoInput` representing a single video
item, which can be passed to a HuggingFace `VideoProcessor`.
Alternatively, a 3-D tensor or batch of 2-D tensors,
@@ -108,7 +108,8 @@ MultiModalDataDict: TypeAlias = Mapping[str, ModalityData[Any]]
"""
A dictionary containing an entry for each modality type to input.
The built-in modalities are defined by {class}`MultiModalDataBuiltins`.
The built-in modalities are defined by
[`MultiModalDataBuiltins`][vllm.multimodal.inputs.MultiModalDataBuiltins].
"""
@@ -169,7 +170,8 @@ Uses a list instead of a tensor if the dimensions of each element do not match.
def nested_tensors_equal(a: NestedTensors, b: NestedTensors) -> bool:
"""Equality check between {data}`NestedTensors` objects."""
"""Equality check between
[`NestedTensors`][vllm.multimodal.inputs.NestedTensors] objects."""
if isinstance(a, torch.Tensor):
return isinstance(b, torch.Tensor) and torch.equal(a, b)
elif isinstance(b, torch.Tensor):
@@ -189,7 +191,7 @@ def nested_tensors_equal(a: NestedTensors, b: NestedTensors) -> bool:
BatchedTensorInputs: TypeAlias = Mapping[str, NestedTensors]
"""
A dictionary containing nested tensors which have been batched via
{meth}`MultiModalKwargs.batch`.
[`MultiModalKwargs.batch`][vllm.multimodal.inputs.MultiModalKwargs.batch].
"""
@@ -197,7 +199,7 @@ A dictionary containing nested tensors which have been batched via
class MultiModalFieldElem:
"""
Represents a keyword argument corresponding to a multi-modal item
in {class}`MultiModalKwargs`.
in [`MultiModalKwargs`][vllm.multimodal.inputs.MultiModalKwargs].
"""
modality: str
@@ -208,13 +210,15 @@ class MultiModalFieldElem:
key: str
"""
The key of this field in {class}`MultiModalKwargs`,
The key of this field in
[`MultiModalKwargs`][vllm.multimodal.inputs.MultiModalKwargs],
i.e. the name of the keyword argument to be passed to the model.
"""
data: NestedTensors
"""
The tensor data of this field in {class}`MultiModalKwargs`,
The tensor data of this field in
[`MultiModalKwargs`][vllm.multimodal.inputs.MultiModalKwargs],
i.e. the value of the keyword argument to be passed to the model.
"""
@@ -237,7 +241,8 @@ class MultiModalFieldElem:
class BaseMultiModalField(ABC):
"""
Defines how to interpret tensor data belonging to a keyword argument in
{class}`MultiModalKwargs` for multiple multi-modal items, and vice versa.
[`MultiModalKwargs`][vllm.multimodal.inputs.MultiModalKwargs] for multiple
multi-modal items, and vice versa.
"""
def _field_factory(self, *, modality: str, key: str):
@@ -262,10 +267,12 @@ class BaseMultiModalField(ABC):
data: NestedTensors,
) -> Sequence[MultiModalFieldElem]:
"""
Construct {class}`MultiModalFieldElem` instances to represent
the provided data.
Construct
[`MultiModalFieldElem`][vllm.multimodal.inputs.MultiModalFieldElem]
instances to represent the provided data.
This is the inverse of {meth}`reduce_data`.
This is the inverse of
[`reduce_data`][vllm.multimodal.inputs.BaseMultiModalField.reduce_data].
"""
raise NotImplementedError
@@ -275,9 +282,11 @@ class BaseMultiModalField(ABC):
def reduce_data(self, elems: list[MultiModalFieldElem]) -> NestedTensors:
"""
Merge the data from multiple instances of {class}`MultiModalFieldElem`.
Merge the data from multiple instances of
[`MultiModalFieldElem`][vllm.multimodal.inputs.MultiModalFieldElem].
This is the inverse of {meth}`build_elems`.
This is the inverse of
[`build_elems`][vllm.multimodal.inputs.BaseMultiModalField.build_elems].
"""
field_types = [type(item.field) for item in elems]
if len(set(field_types)) > 1:
@@ -290,7 +299,7 @@ class BaseMultiModalField(ABC):
class MultiModalBatchedField(BaseMultiModalField):
"""
Info:
[MultiModalFieldConfig.batched][]
[`MultiModalFieldConfig.batched`][vllm.multimodal.inputs.MultiModalFieldConfig.batched]
"""
def build_elems(
@@ -320,8 +329,8 @@ class MultiModalBatchedField(BaseMultiModalField):
class MultiModalFlatField(BaseMultiModalField):
"""
Info:
[MultiModalFieldConfig.flat][]
[MultiModalFieldConfig.flat_from_sizes][]
[`MultiModalFieldConfig.flat`][vllm.multimodal.inputs.MultiModalFieldConfig.flat]
[`MultiModalFieldConfig.flat_from_sizes`][vllm.multimodal.inputs.MultiModalFieldConfig.flat_from_sizes]
"""
slices: Union[Sequence[slice], Sequence[Sequence[slice]]]
dim: int = 0
@@ -362,7 +371,7 @@ class MultiModalFlatField(BaseMultiModalField):
class MultiModalSharedField(BaseMultiModalField):
"""
Info:
[MultiModalFieldConfig.shared][]
[`MultiModalFieldConfig.shared`][vllm.multimodal.inputs.MultiModalFieldConfig.shared]
"""
batch_size: int
@@ -508,7 +517,7 @@ class MultiModalFieldConfig:
```
Info:
[MultiModalFieldConfig.flat][]
[`MultiModalFieldConfig.flat`][vllm.multimodal.inputs.MultiModalFieldConfig.flat]
"""
if size_per_item.ndim != 1:
@@ -572,8 +581,10 @@ class MultiModalFieldConfig:
class MultiModalKwargsItem(UserDict[str, MultiModalFieldElem]):
"""
A collection of {class}`MultiModalFieldElem`
corresponding to a data item in {class}`MultiModalDataItems`.
A collection of
[`MultiModalFieldElem`][vllm.multimodal.inputs.MultiModalFieldElem]
corresponding to a data item in
[`MultiModalDataItems`][vllm.multimodal.parse.MultiModalDataItems].
"""
@staticmethod
@@ -592,11 +603,13 @@ class MultiModalKwargsItem(UserDict[str, MultiModalFieldElem]):
class MultiModalKwargs(UserDict[str, NestedTensors]):
"""
A dictionary that represents the keyword arguments to
{meth}`~torch.nn.Module.forward`.
[`torch.nn.Module.forward`][].
The metadata `items` enables us to obtain the keyword arguments
corresponding to each data item in {class}`MultiModalDataItems`, via
{meth}`get_item` and {meth}`get_items`.
corresponding to each data item in
[`MultiModalDataItems`][vllm.multimodal.parse.MultiModalDataItems], via
[`get_item`][vllm.multimodal.inputs.MultiModalKwargs.get_item] and
[`get_items`][vllm.multimodal.inputs.MultiModalKwargs.get_items].
"""
@staticmethod
@@ -635,7 +648,9 @@ class MultiModalKwargs(UserDict[str, NestedTensors]):
@staticmethod
def from_items(items: Sequence[MultiModalKwargsItem]):
"""Construct a new {class}`MultiModalKwargs` from multiple items."""
"""Construct a new
[`MultiModalKwargs`][vllm.multimodal.inputs.MultiModalKwargs]
from multiple items."""
elems_by_key = defaultdict[str, list[MultiModalFieldElem]](list)
for item in items:
for key, elem in item.items():
@@ -800,7 +815,7 @@ A dictionary containing placeholder ranges for each modality.
class MultiModalInputs(TypedDict):
"""
Represents the outputs of
{class}`vllm.multimodal.processing.BaseMultiModalProcessor`,
[`BaseMultiModalProcessor`][vllm.multimodal.processing.BaseMultiModalProcessor],
ready to be passed to vLLM internals.
"""
@@ -836,7 +851,8 @@ class MultiModalInputs(TypedDict):
class MultiModalEncDecInputs(MultiModalInputs):
"""
Represents the outputs of {class}`vllm.multimodal.EncDecMultiModalProcessor`
Represents the outputs of
[`EncDecMultiModalProcessor`][vllm.multimodal.processing.EncDecMultiModalProcessor]
ready to be passed to vLLM internals.
"""