Update Optional[x] -> x | None and Union[x, y] to x | y (#26633)

Signed-off-by: Harry Mellor <19981378+hmellor@users.noreply.github.com>
This commit is contained in:
Harry Mellor
2025-10-12 09:51:31 -07:00
committed by GitHub
parent 9bb38130cb
commit 8fcaaf6a16
944 changed files with 9490 additions and 10121 deletions
+26 -24
View File
@@ -1,7 +1,7 @@
# SPDX-License-Identifier: Apache-2.0
# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
from collections.abc import Iterable
from typing import TYPE_CHECKING, Any, Generic, Literal, Optional, Union, cast
from typing import TYPE_CHECKING, Any, Generic, Literal, TypeAlias, cast
import torch
from typing_extensions import NotRequired, TypedDict, TypeIs, TypeVar
@@ -12,6 +12,10 @@ if TYPE_CHECKING:
MultiModalInputs,
MultiModalUUIDDict,
)
else:
MultiModalDataDict = object
MultiModalInputs = object
MultiModalUUIDDict = object
class TextPrompt(TypedDict):
@@ -20,13 +24,13 @@ class TextPrompt(TypedDict):
prompt: str
"""The input text to be tokenized before passing to the model."""
multi_modal_data: NotRequired[Optional["MultiModalDataDict"]]
multi_modal_data: NotRequired[MultiModalDataDict | None]
"""
Optional multi-modal data to pass to the model,
if the model supports it.
"""
mm_processor_kwargs: NotRequired[Optional[dict[str, Any]]]
mm_processor_kwargs: NotRequired[dict[str, Any] | None]
"""
Optional multi-modal processor kwargs to be forwarded to the
multimodal input mapper & processor. Note that if multiple modalities
@@ -34,7 +38,7 @@ class TextPrompt(TypedDict):
to pass the mm_processor_kwargs to each of them.
"""
multi_modal_uuids: NotRequired["MultiModalUUIDDict"]
multi_modal_uuids: NotRequired[MultiModalUUIDDict]
"""
Optional user-specified UUIDs for multimodal items, mapped by modality.
Lists must match the number of items per modality and may contain `None`.
@@ -61,13 +65,13 @@ class TokensPrompt(TypedDict):
token_type_ids: NotRequired[list[int]]
"""A list of token type IDs to pass to the cross encoder model."""
multi_modal_data: NotRequired[Optional["MultiModalDataDict"]]
multi_modal_data: NotRequired[MultiModalDataDict | None]
"""
Optional multi-modal data to pass to the model,
if the model supports it.
"""
mm_processor_kwargs: NotRequired[Optional[dict[str, Any]]]
mm_processor_kwargs: NotRequired[dict[str, Any] | None]
"""
Optional multi-modal processor kwargs to be forwarded to the
multimodal input mapper & processor. Note that if multiple modalities
@@ -75,7 +79,7 @@ class TokensPrompt(TypedDict):
to pass the mm_processor_kwargs to each of them.
"""
multi_modal_uuids: NotRequired["MultiModalUUIDDict"]
multi_modal_uuids: NotRequired[MultiModalUUIDDict]
"""
Optional user-specified UUIDs for multimodal items, mapped by modality.
Lists must match the number of items per modality and may contain `None`.
@@ -111,7 +115,7 @@ class DataPrompt(TypedDict):
"""The input data format"""
SingletonPrompt = Union[str, TextPrompt, TokensPrompt, EmbedsPrompt]
SingletonPrompt: TypeAlias = str | TextPrompt | TokensPrompt | EmbedsPrompt
"""
Set of possible schemas for a single prompt:
@@ -185,12 +189,12 @@ class ExplicitEncoderDecoderPrompt(TypedDict, Generic[_T1_co, _T2_co]):
encoder_prompt: _T1_co
decoder_prompt: Optional[_T2_co]
decoder_prompt: _T2_co | None
mm_processor_kwargs: NotRequired[dict[str, Any]]
PromptType = Union[SingletonPrompt, ExplicitEncoderDecoderPrompt]
PromptType: TypeAlias = SingletonPrompt | ExplicitEncoderDecoderPrompt
"""
Set of possible schemas for an LLM input, including
both decoder-only and encoder/decoder input types:
@@ -220,7 +224,7 @@ class TokenInputs(TypedDict):
def token_inputs(
prompt_token_ids: list[int],
cache_salt: Optional[str] = None,
cache_salt: str | None = None,
) -> TokenInputs:
"""Construct [`TokenInputs`][vllm.inputs.data.TokenInputs] from optional
values."""
@@ -249,7 +253,7 @@ class EmbedsInputs(TypedDict):
def embeds_inputs(
prompt_embeds: torch.Tensor,
cache_salt: Optional[str] = None,
cache_salt: str | None = None,
) -> EmbedsInputs:
"""Construct [`EmbedsInputs`][vllm.inputs.data.EmbedsInputs] from optional
values."""
@@ -261,7 +265,7 @@ def embeds_inputs(
return inputs
DecoderOnlyInputs = Union[TokenInputs, EmbedsInputs, "MultiModalInputs"]
DecoderOnlyInputs: TypeAlias = TokenInputs | EmbedsInputs | MultiModalInputs
"""
The inputs in [`LLMEngine`][vllm.engine.llm_engine.LLMEngine] before they are
passed to the model executor.
@@ -277,20 +281,20 @@ class EncoderDecoderInputs(TypedDict):
This specifies the required data for encoder-decoder models.
"""
encoder: Union[TokenInputs, "MultiModalInputs"]
encoder: TokenInputs | MultiModalInputs
"""The inputs for the encoder portion."""
decoder: Union[TokenInputs, "MultiModalInputs"]
decoder: TokenInputs | MultiModalInputs
"""The inputs for the decoder portion."""
SingletonInputs = Union[TokenInputs, EmbedsInputs, "MultiModalInputs"]
SingletonInputs: TypeAlias = TokenInputs | EmbedsInputs | MultiModalInputs
"""
A processed [`SingletonPrompt`][vllm.inputs.data.SingletonPrompt] which can be
passed to [`Sequence`][collections.abc.Sequence].
"""
ProcessorInputs = Union[DecoderOnlyInputs, EncoderDecoderInputs]
ProcessorInputs: TypeAlias = DecoderOnlyInputs | EncoderDecoderInputs
"""
The outputs from [`vllm.inputs.preprocess.InputPreprocessor`][].
"""
@@ -301,8 +305,8 @@ _T2 = TypeVar("_T2", bound=SingletonPrompt, default=SingletonPrompt)
def build_explicit_enc_dec_prompt(
encoder_prompt: _T1,
decoder_prompt: Optional[_T2],
mm_processor_kwargs: Optional[dict[str, Any]] = None,
decoder_prompt: _T2 | None,
mm_processor_kwargs: dict[str, Any] | None = None,
) -> ExplicitEncoderDecoderPrompt[_T1, _T2]:
if mm_processor_kwargs is None:
mm_processor_kwargs = {}
@@ -315,10 +319,8 @@ def build_explicit_enc_dec_prompt(
def zip_enc_dec_prompts(
enc_prompts: Iterable[_T1],
dec_prompts: Iterable[Optional[_T2]],
mm_processor_kwargs: Optional[
Union[Iterable[dict[str, Any]], dict[str, Any]]
] = None,
dec_prompts: Iterable[_T2 | None],
mm_processor_kwargs: Iterable[dict[str, Any]] | dict[str, Any] | None = None,
) -> list[ExplicitEncoderDecoderPrompt[_T1, _T2]]:
"""
Zip encoder and decoder prompts together into a list of
@@ -350,7 +352,7 @@ def zip_enc_dec_prompts(
def to_enc_dec_tuple_list(
enc_dec_prompts: Iterable[ExplicitEncoderDecoderPrompt[_T1, _T2]],
) -> list[tuple[_T1, Optional[_T2]]]:
) -> list[tuple[_T1, _T2 | None]]:
return [
(enc_dec_prompt["encoder_prompt"], enc_dec_prompt["decoder_prompt"])
for enc_dec_prompt in enc_dec_prompts
+10 -10
View File
@@ -1,7 +1,7 @@
# SPDX-License-Identifier: Apache-2.0
# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
from collections.abc import Sequence
from typing import TYPE_CHECKING, Literal, NamedTuple, Optional, TypedDict, Union, cast
from typing import TYPE_CHECKING, Literal, NamedTuple, TypeAlias, TypedDict, cast
from typing_extensions import TypeIs
@@ -23,8 +23,8 @@ if TYPE_CHECKING:
def parse_raw_prompts(
prompt: Union[str, list[str], list[int], list[list[int]]],
) -> Union[Sequence[TextPrompt], Sequence[TokensPrompt]]:
prompt: str | list[str] | list[int] | list[list[int]],
) -> Sequence[TextPrompt] | Sequence[TokensPrompt]:
if isinstance(prompt, str):
# case 1: a string
return [TextPrompt(prompt=prompt)]
@@ -76,9 +76,9 @@ class ParsedEmbedsPrompt(TypedDict):
content: EmbedsPrompt
ParsedSingletonPrompt = Union[
ParsedStrPrompt, ParsedTextPrompt, ParsedTokensPrompt, ParsedEmbedsPrompt
]
ParsedSingletonPrompt: TypeAlias = (
ParsedStrPrompt | ParsedTextPrompt | ParsedTokensPrompt | ParsedEmbedsPrompt
)
def parse_singleton_prompt(prompt: SingletonPrompt) -> ParsedSingletonPrompt:
@@ -106,7 +106,7 @@ def is_explicit_encoder_decoder_prompt(
def split_enc_dec_inputs(
inputs: ProcessorInputs,
) -> tuple[Optional[SingletonInputs], SingletonInputs]:
) -> tuple[SingletonInputs | None, SingletonInputs]:
if "encoder" in inputs and "decoder" in inputs:
# NOTE: This passes pyright but not mypy
return (
@@ -118,9 +118,9 @@ def split_enc_dec_inputs(
class PromptComponents(NamedTuple):
text: Optional[str] = None
token_ids: Optional[list[int]] = None
embeds: Optional["torch.Tensor"] = None
text: str | None = None
token_ids: list[int] | None = None
embeds: "torch.Tensor | None" = None
def get_prompt_components(prompt: PromptType) -> PromptComponents:
+42 -44
View File
@@ -2,7 +2,7 @@
# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
from collections.abc import Mapping
from typing import Any, Optional, Union, cast
from typing import Any, cast
from typing_extensions import assert_never
@@ -46,9 +46,9 @@ class InputPreprocessor:
def __init__(
self,
model_config: ModelConfig,
tokenizer: Optional[AnyTokenizer],
tokenizer: AnyTokenizer | None,
mm_registry: MultiModalRegistry = MULTIMODAL_REGISTRY,
mm_processor_cache: Optional[BaseMultiModalProcessorCache] = None,
mm_processor_cache: BaseMultiModalProcessorCache | None = None,
) -> None:
super().__init__()
@@ -67,7 +67,7 @@ class InputPreprocessor:
return self.tokenizer
def get_bos_token_id(self) -> Optional[int]:
def get_bos_token_id(self) -> int | None:
if self.tokenizer is None:
logger.warning_once(
"Using None for BOS token id because tokenizer is not initialized"
@@ -76,7 +76,7 @@ class InputPreprocessor:
return self.tokenizer.bos_token_id
def get_eos_token_id(self) -> Optional[int]:
def get_eos_token_id(self) -> int | None:
if self.tokenizer is None:
logger.warning_once(
"Using None for EOS token id because tokenizer is not initialized"
@@ -85,7 +85,7 @@ class InputPreprocessor:
return self.tokenizer.eos_token_id
def get_decoder_start_token_id(self) -> Optional[int]:
def get_decoder_start_token_id(self) -> int | None:
"""
Obtain the decoder start token id employed by an encoder/decoder
model. Returns None for non-encoder/decoder models or if the
@@ -157,7 +157,7 @@ class InputPreprocessor:
def _prepare_decoder_input_ids_for_generation(
self,
decoder_input_ids: Optional[list[int]],
decoder_input_ids: list[int] | None,
) -> list[int]:
"""
Prepares `decoder_input_ids` for generation with encoder-decoder models.
@@ -194,7 +194,7 @@ class InputPreprocessor:
def _get_tokenization_kw(
self,
overrides: Optional[dict[str, Any]] = None,
overrides: dict[str, Any] | None = None,
) -> dict[str, Any]:
kwargs = dict[str, Any]()
@@ -212,7 +212,7 @@ class InputPreprocessor:
def _tokenize_prompt(
self,
prompt: str,
tokenization_kwargs: Optional[dict[str, Any]] = None,
tokenization_kwargs: dict[str, Any] | None = None,
) -> list[int]:
"""
Apply the model's tokenizer to a text prompt, returning the
@@ -251,12 +251,12 @@ class InputPreprocessor:
def _process_multimodal(
self,
prompt: Union[str, list[int]],
prompt: str | list[int],
mm_data: MultiModalDataDict,
mm_processor_kwargs: Optional[Mapping[str, object]],
tokenization_kwargs: Optional[dict[str, Any]] = None,
mm_processor_kwargs: Mapping[str, object] | None,
tokenization_kwargs: dict[str, Any] | None = None,
*,
mm_uuids: Optional[MultiModalUUIDDict] = None,
mm_uuids: MultiModalUUIDDict | None = None,
) -> MultiModalInputs:
"""
Apply the model's multi-modal processor to a multi-modal prompt,
@@ -320,7 +320,7 @@ class InputPreprocessor:
)
def _truncate_inputs(
self, inputs: list[int], tokenization_kwargs: Optional[dict[str, Any]] = None
self, inputs: list[int], tokenization_kwargs: dict[str, Any] | None = None
) -> list[int]:
if (
not tokenization_kwargs
@@ -339,15 +339,15 @@ class InputPreprocessor:
def _process_tokens(
self,
parsed_content: TokensPrompt,
tokenization_kwargs: Optional[dict[str, Any]] = None,
tokenization_kwargs: dict[str, Any] | None = None,
*,
mm_uuids: Optional[MultiModalUUIDDict] = None,
) -> Union[TokenInputs, MultiModalInputs]:
mm_uuids: MultiModalUUIDDict | None = None,
) -> TokenInputs | MultiModalInputs:
prompt_token_ids = self._truncate_inputs(
parsed_content["prompt_token_ids"], tokenization_kwargs
)
inputs: Union[TokenInputs, MultiModalInputs]
inputs: TokenInputs | MultiModalInputs
if self.model_config.is_multimodal_model:
inputs = self._process_multimodal(
prompt_token_ids,
@@ -370,13 +370,13 @@ class InputPreprocessor:
def _process_text(
self,
parsed_content: TextPrompt,
tokenization_kwargs: Optional[dict[str, Any]] = None,
tokenization_kwargs: dict[str, Any] | None = None,
*,
mm_uuids: Optional[MultiModalUUIDDict] = None,
) -> Union[TokenInputs, MultiModalInputs]:
mm_uuids: MultiModalUUIDDict | None = None,
) -> TokenInputs | MultiModalInputs:
prompt_text = parsed_content["prompt"]
inputs: Union[TokenInputs, MultiModalInputs]
inputs: TokenInputs | MultiModalInputs
if self.model_config.is_multimodal_model:
inputs = self._process_multimodal(
prompt_text,
@@ -403,9 +403,9 @@ class InputPreprocessor:
def _prompt_to_llm_inputs(
self,
prompt: SingletonPrompt,
tokenization_kwargs: Optional[dict[str, Any]] = None,
tokenization_kwargs: dict[str, Any] | None = None,
*,
mm_uuids: Optional[MultiModalUUIDDict] = None,
mm_uuids: MultiModalUUIDDict | None = None,
) -> SingletonInputs:
"""
Extract the singleton inputs from a prompt.
@@ -445,7 +445,7 @@ class InputPreprocessor:
def _build_enc_dec_llm_inputs(
self,
encoder_inputs: SingletonInputs,
decoder_inputs: Optional[SingletonInputs],
decoder_inputs: SingletonInputs | None,
) -> EncoderDecoderInputs:
if (
encoder_inputs["type"] == "embeds"
@@ -457,10 +457,8 @@ class InputPreprocessor:
)
# Needed for mypy
encoder_inputs = cast(Union[TokenInputs, MultiModalInputs], encoder_inputs)
decoder_inputs = cast(
Optional[Union[TokenInputs, MultiModalInputs]], decoder_inputs
)
encoder_inputs = cast(TokenInputs | MultiModalInputs, encoder_inputs)
decoder_inputs = cast(TokenInputs | MultiModalInputs | None, decoder_inputs)
if decoder_inputs is None:
if self.model_config.hf_config.model_type == "whisper":
@@ -491,8 +489,8 @@ class InputPreprocessor:
def _split_enc_dec_mm_inputs(
self,
inputs: Union[SingletonInputs, MultiModalEncDecInputs],
decoder_inputs_to_override: Optional[SingletonInputs] = None,
inputs: SingletonInputs | MultiModalEncDecInputs,
decoder_inputs_to_override: SingletonInputs | None = None,
) -> tuple[SingletonInputs, SingletonInputs]:
"""
For encoder/decoder models only:
@@ -509,11 +507,11 @@ class InputPreprocessor:
# Needed for mypy
inputs = cast(
Union[TokenInputs, MultiModalInputs, MultiModalEncDecInputs],
TokenInputs | MultiModalInputs | MultiModalEncDecInputs,
inputs,
)
decoder_inputs_to_override = cast(
Optional[Union[TokenInputs, MultiModalInputs]],
TokenInputs | MultiModalInputs | None,
decoder_inputs_to_override,
)
@@ -553,9 +551,9 @@ class InputPreprocessor:
def _process_encoder_decoder_prompt(
self,
prompt: PromptType,
tokenization_kwargs: Optional[dict[str, Any]] = None,
tokenization_kwargs: dict[str, Any] | None = None,
*,
mm_uuids: Optional[MultiModalUUIDDict] = None,
mm_uuids: MultiModalUUIDDict | None = None,
) -> EncoderDecoderInputs:
"""
For encoder/decoder models only:
@@ -591,7 +589,7 @@ class InputPreprocessor:
instance
"""
encoder_inputs: SingletonInputs
decoder_inputs: Optional[SingletonInputs]
decoder_inputs: SingletonInputs | None
if is_explicit_encoder_decoder_prompt(prompt):
# `cast` is needed for mypy, but not pyright
@@ -633,7 +631,7 @@ class InputPreprocessor:
) -> DecoderOnlyInputs:
if "prompt_token_ids" in prompt_inputs:
prompt_inputs = cast(
Union[TokenInputs, MultiModalInputs], prompt_inputs
TokenInputs | MultiModalInputs, prompt_inputs
) # Needed for mypy
return prompt_inputs
@@ -641,9 +639,9 @@ class InputPreprocessor:
def _process_decoder_only_prompt(
self,
prompt: SingletonPrompt,
tokenization_kwargs: Optional[dict[str, Any]] = None,
tokenization_kwargs: dict[str, Any] | None = None,
*,
mm_uuids: Optional[MultiModalUUIDDict] = None,
mm_uuids: MultiModalUUIDDict | None = None,
) -> DecoderOnlyInputs:
"""
For decoder-only models:
@@ -670,9 +668,9 @@ class InputPreprocessor:
def _preprocess(
self,
prompt: PromptType,
tokenization_kwargs: Optional[dict[str, Any]] = None,
tokenization_kwargs: dict[str, Any] | None = None,
*,
mm_uuids: Optional[MultiModalUUIDDict] = None,
mm_uuids: MultiModalUUIDDict | None = None,
) -> ProcessorInputs:
if self.model_config.is_encoder_decoder:
# Encoder-decoder model requires special mapping of
@@ -699,9 +697,9 @@ class InputPreprocessor:
def preprocess(
self,
prompt: PromptType,
tokenization_kwargs: Optional[dict[str, Any]] = None,
tokenization_kwargs: dict[str, Any] | None = None,
*,
mm_uuids: Optional[MultiModalUUIDDict] = None,
mm_uuids: MultiModalUUIDDict | None = None,
) -> ProcessorInputs:
"""Preprocess the input prompt."""
res = self._preprocess(
@@ -718,7 +716,7 @@ class InputPreprocessor:
return res
def stat_mm_cache(self) -> Optional[MultiModalCacheStats]:
def stat_mm_cache(self) -> MultiModalCacheStats | None:
mm_cache_stats = self.mm_cache_stats
if mm_cache_stats is None:
return None