xinyun/vllm
1
0
Fork 0
mirror of https://git.datalinker.icu/vllm-project/vllm.git synced 2026-01-23 12:54:30 +08:00
Code Issues Packages Projects Releases Wiki Activity

[Misc] Upgrade to Python 3.9 typing for additional directories (#14492)

Browse Source 
Signed-off-by: DarkLight1337 <tlleungac@connect.ust.hk>
This commit is contained in:
Cyrus Leung 2025-03-09 01:35:50 +08:00 committed by GitHub
parent 0d5e73d30e
commit 9513290032
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
12 changed files with 78 additions and 78 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

5
pyproject.toml
View File

@ -75,11 +75,8 @@ exclude = [
"vllm/distributed/**/*.py" = ["UP006", "UP035"]
"vllm/engine/**/*.py" = ["UP006", "UP035"]
"vllm/executor/**/*.py" = ["UP006", "UP035"]
"vllm/inputs/**/*.py" = ["UP006", "UP035"]
"vllm/logging_utils/**/*.py" = ["UP006", "UP035"]
"vllm/lora/**/*.py" = ["UP006", "UP035"]
"vllm/model_executor/**/*.py" = ["UP006", "UP035"]
"vllm/multimodal/**/*.py" = ["UP006", "UP035"]
"vllm/platforms/**/*.py" = ["UP006", "UP035"]
"vllm/plugins/**/*.py" = ["UP006", "UP035"]
"vllm/profiler/**/*.py" = ["UP006", "UP035"]
@ -87,9 +84,7 @@ exclude = [
"vllm/spec_decode/**/*.py" = ["UP006", "UP035"]
"vllm/transformers_utils/**/*.py" = ["UP006", "UP035"]
"vllm/triton_utils/**/*.py" = ["UP006", "UP035"]
"vllm/usage/**/*.py" = ["UP006", "UP035"]
"vllm/vllm_flash_attn/**/*.py" = ["UP006", "UP035"]
"vllm/assets/**/*.py" = ["UP006", "UP035"]
"vllm/worker/**/*.py" = ["UP006", "UP035"]
[tool.ruff.lint]

6
vllm/assets/video.py
View File

@ -2,7 +2,7 @@
from dataclasses import dataclass
from functools import lru_cache
from typing import List, Literal
from typing import Literal
import cv2
import numpy as np
@ -58,7 +58,7 @@ def video_to_ndarrays(path: str, num_frames: int = -1) -> npt.NDArray:
def video_to_pil_images_list(path: str,
num_frames: int = -1) -> List[Image.Image]:
num_frames: int = -1) -> list[Image.Image]:
frames = video_to_ndarrays(path, num_frames)
return [
Image.fromarray(cv2.cvtColor(frame, cv2.COLOR_BGR2RGB))
@ -72,7 +72,7 @@ class VideoAsset:
num_frames: int = -1
@property
def pil_images(self) -> List[Image.Image]:
def pil_images(self) -> list[Image.Image]:
video_path = download_video_asset(self.name)
ret = video_to_pil_images_list(video_path, self.num_frames)
return ret

54
vllm/inputs/data.py
View File

@ -1,9 +1,9 @@
# SPDX-License-Identifier: Apache-2.0
from collections.abc import Iterable
from dataclasses import dataclass
from functools import cached_property
from typing import (TYPE_CHECKING, Any, Dict, Generic, Iterable, List, Literal,
Optional, Tuple, Union, cast)
from typing import TYPE_CHECKING, Any, Generic, Literal, Optional, Union, cast
import torch
from typing_extensions import NotRequired, TypedDict, TypeVar, assert_never
@ -26,7 +26,7 @@ class TextPrompt(TypedDict):
if the model supports it.
"""
mm_processor_kwargs: NotRequired[Dict[str, Any]]
mm_processor_kwargs: NotRequired[dict[str, Any]]
"""
Optional multi-modal processor kwargs to be forwarded to the
multimodal input mapper & processor. Note that if multiple modalities
@ -38,10 +38,10 @@ class TextPrompt(TypedDict):
class TokensPrompt(TypedDict):
"""Schema for a tokenized prompt."""
prompt_token_ids: List[int]
prompt_token_ids: list[int]
"""A list of token IDs to pass to the model."""
token_type_ids: NotRequired[List[int]]
token_type_ids: NotRequired[list[int]]
"""A list of token type IDs to pass to the cross encoder model."""
multi_modal_data: NotRequired["MultiModalDataDict"]
@ -50,7 +50,7 @@ class TokensPrompt(TypedDict):
if the model supports it.
"""
mm_processor_kwargs: NotRequired[Dict[str, Any]]
mm_processor_kwargs: NotRequired[dict[str, Any]]
"""
Optional multi-modal processor kwargs to be forwarded to the
multimodal input mapper & processor. Note that if multiple modalities
@ -115,7 +115,7 @@ class ExplicitEncoderDecoderPrompt(TypedDict, Generic[_T1_co, _T2_co]):
decoder_prompt: Optional[_T2_co]
mm_processor_kwargs: NotRequired[Dict[str, Any]]
mm_processor_kwargs: NotRequired[dict[str, Any]]
PromptType = Union[SingletonPrompt, ExplicitEncoderDecoderPrompt]
@ -136,10 +136,10 @@ class TokenInputs(TypedDict):
type: Literal["token"]
"""The type of inputs."""
prompt_token_ids: List[int]
prompt_token_ids: list[int]
"""The token IDs of the prompt."""
token_type_ids: NotRequired[List[int]]
token_type_ids: NotRequired[list[int]]
"""The token type IDs of the prompt."""
prompt: NotRequired[str]
@ -164,12 +164,12 @@ class TokenInputs(TypedDict):
Placeholder ranges for the multi-modal data.
"""
multi_modal_hashes: NotRequired[List[str]]
multi_modal_hashes: NotRequired[list[str]]
"""
The hashes of the multi-modal data.
"""
mm_processor_kwargs: NotRequired[Dict[str, Any]]
mm_processor_kwargs: NotRequired[dict[str, Any]]
"""
Optional multi-modal processor kwargs to be forwarded to the
multimodal input mapper & processor. Note that if multiple modalities
@ -179,14 +179,14 @@ class TokenInputs(TypedDict):
def token_inputs(
prompt_token_ids: List[int],
token_type_ids: Optional[List[int]] = None,
prompt_token_ids: list[int],
token_type_ids: Optional[list[int]] = None,
prompt: Optional[str] = None,
multi_modal_data: Optional["MultiModalDataDict"] = None,
multi_modal_inputs: Optional["MultiModalKwargs"] = None,
multi_modal_hashes: Optional[List[str]] = None,
multi_modal_hashes: Optional[list[str]] = None,
multi_modal_placeholders: Optional["MultiModalPlaceholderDict"] = None,
mm_processor_kwargs: Optional[Dict[str, Any]] = None,
mm_processor_kwargs: Optional[dict[str, Any]] = None,
) -> TokenInputs:
"""Construct :class:`TokenInputs` from optional values."""
inputs = TokenInputs(type="token", prompt_token_ids=prompt_token_ids)
@ -255,7 +255,7 @@ class SingletonInputsAdapter:
assert_never(inputs) # type: ignore[arg-type]
@cached_property
def prompt_token_ids(self) -> List[int]:
def prompt_token_ids(self) -> list[int]:
inputs = self.inputs
if inputs["type"] == "token" or inputs["type"] == "multimodal":
@ -264,7 +264,7 @@ class SingletonInputsAdapter:
assert_never(inputs) # type: ignore[arg-type]
@cached_property
def token_type_ids(self) -> List[int]:
def token_type_ids(self) -> list[int]:
inputs = self.inputs
if inputs["type"] == "token" or inputs["type"] == "multimodal":
@ -294,7 +294,7 @@ class SingletonInputsAdapter:
assert_never(inputs) # type: ignore[arg-type]
@cached_property
def multi_modal_inputs(self) -> Union[Dict, "MultiModalKwargs"]:
def multi_modal_inputs(self) -> Union[dict, "MultiModalKwargs"]:
inputs = self.inputs
if inputs["type"] == "token":
@ -306,7 +306,7 @@ class SingletonInputsAdapter:
assert_never(inputs) # type: ignore[arg-type]
@cached_property
def multi_modal_hashes(self) -> List[str]:
def multi_modal_hashes(self) -> list[str]:
inputs = self.inputs
if inputs["type"] == "token":
@ -331,7 +331,7 @@ class SingletonInputsAdapter:
assert_never(inputs) # type: ignore[arg-type]
@cached_property
def mm_processor_kwargs(self) -> Dict[str, Any]:
def mm_processor_kwargs(self) -> dict[str, Any]:
inputs = self.inputs
if inputs["type"] == "token":
@ -355,7 +355,7 @@ _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,
mm_processor_kwargs: Optional[dict[str, Any]] = None,
) -> ExplicitEncoderDecoderPrompt[_T1, _T2]:
if mm_processor_kwargs is None:
mm_processor_kwargs = {}
@ -368,9 +368,9 @@ 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,
) -> List[ExplicitEncoderDecoderPrompt[_T1, _T2]]:
mm_processor_kwargs: Optional[Union[Iterable[dict[str, Any]],
dict[str, Any]]] = None,
) -> list[ExplicitEncoderDecoderPrompt[_T1, _T2]]:
"""
Zip encoder and decoder prompts together into a list of
:class:`ExplicitEncoderDecoderPrompt` instances.
@ -380,12 +380,12 @@ def zip_enc_dec_prompts(
provided, it will be zipped with the encoder/decoder prompts.
"""
if mm_processor_kwargs is None:
mm_processor_kwargs = cast(Dict[str, Any], {})
mm_processor_kwargs = cast(dict[str, Any], {})
if isinstance(mm_processor_kwargs, dict):
return [
build_explicit_enc_dec_prompt(
encoder_prompt, decoder_prompt,
cast(Dict[str, Any], mm_processor_kwargs))
cast(dict[str, Any], mm_processor_kwargs))
for (encoder_prompt,
decoder_prompt) in zip(enc_prompts, dec_prompts)
]
@ -399,7 +399,7 @@ 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, Optional[_T2]]]:
return [(enc_dec_prompt["encoder_prompt"],
enc_dec_prompt["decoder_prompt"])
for enc_dec_prompt in enc_dec_prompts]

17
vllm/inputs/parse.py
View File

@ -1,6 +1,7 @@
# SPDX-License-Identifier: Apache-2.0
from typing import List, Literal, Sequence, TypedDict, Union, cast, overload
from collections.abc import Sequence
from typing import Literal, TypedDict, Union, cast, overload
from typing_extensions import TypeIs
@ -17,24 +18,24 @@ class ParsedText(TypedDict):
class ParsedTokens(TypedDict):
content: List[int]
content: list[int]
is_tokens: Literal[True]
@overload
def parse_and_batch_prompt(
prompt: Union[str, List[str]]) -> Sequence[ParsedText]:
prompt: Union[str, list[str]]) -> Sequence[ParsedText]:
...
@overload
def parse_and_batch_prompt(
prompt: Union[List[int], List[List[int]]]) -> Sequence[ParsedTokens]:
prompt: Union[list[int], list[list[int]]]) -> Sequence[ParsedTokens]:
...
def parse_and_batch_prompt(
prompt: Union[str, List[str], List[int], List[List[int]]],
prompt: Union[str, list[str], list[int], list[list[int]]],
) -> Union[Sequence[ParsedText], Sequence[ParsedTokens]]:
if isinstance(prompt, str):
# case 1: a string
@ -46,16 +47,16 @@ def parse_and_batch_prompt(
if is_list_of(prompt, str):
# case 2: array of strings
prompt = cast(List[str], prompt)
prompt = cast(list[str], prompt)
return [
ParsedText(content=elem, is_tokens=False) for elem in prompt
]
if is_list_of(prompt, int):
# case 3: array of tokens
prompt = cast(List[int], prompt)
prompt = cast(list[int], prompt)
return [ParsedTokens(content=prompt, is_tokens=True)]
if is_list_of(prompt, list):
prompt = cast(List[List[int]], prompt)
prompt = cast(list[list[int]], prompt)
if len(prompt[0]) == 0:
raise ValueError("please provide at least one prompt")

23
vllm/inputs/preprocess.py
View File

@ -1,7 +1,8 @@
# SPDX-License-Identifier: Apache-2.0
import asyncio
from typing import List, Mapping, Optional, Tuple, Union, cast
from collections.abc import Mapping
from typing import Optional, Union, cast
from typing_extensions import assert_never
@ -92,7 +93,7 @@ class InputPreprocessor:
return dec_start_token_id
def _get_default_enc_dec_decoder_prompt(self) -> List[int]:
def _get_default_enc_dec_decoder_prompt(self) -> list[int]:
'''
Specifically for encoder/decoder models:
generate a default decoder prompt for when
@ -130,8 +131,8 @@ class InputPreprocessor:
def _prepare_decoder_input_ids_for_generation(
self,
decoder_input_ids: Optional[List[int]],
) -> List[int]:
decoder_input_ids: Optional[list[int]],
) -> list[int]:
"""
Prepares `decoder_input_ids` for generation with encoder-decoder models.
@ -168,9 +169,9 @@ class InputPreprocessor:
def _apply_prompt_adapter(
self,
prompt_token_ids: List[int],
prompt_token_ids: list[int],
prompt_adapter_request: Optional[PromptAdapterRequest],
) -> List[int]:
) -> list[int]:
if prompt_adapter_request:
prompt_token_ids = (
[0] * prompt_adapter_request.prompt_adapter_num_virtual_tokens
@ -183,7 +184,7 @@ class InputPreprocessor:
prompt: str,
request_id: str,
lora_request: Optional[LoRARequest],
) -> List[int]:
) -> list[int]:
"""
Apply the model's tokenizer to a text prompt, returning the
corresponding token IDs.
@ -211,7 +212,7 @@ class InputPreprocessor:
prompt: str,
request_id: str,
lora_request: Optional[LoRARequest],
) -> List[int]:
) -> list[int]:
"""Async version of :meth:`_tokenize_prompt`."""
tokenizer = self.get_tokenizer_group()
add_special_tokens = None
@ -250,7 +251,7 @@ class InputPreprocessor:
def _process_multimodal(
self,
prompt: Union[str, List[int]],
prompt: Union[str, list[int]],
mm_data: MultiModalDataDict,
mm_processor_kwargs: Optional[Mapping[str, object]],
lora_request: Optional[LoRARequest],
@ -280,7 +281,7 @@ class InputPreprocessor:
async def _process_multimodal_async(
self,
prompt: Union[str, List[int]],
prompt: Union[str, list[int]],
mm_data: MultiModalDataDict,
mm_processor_kwargs: Optional[Mapping[str, object]],
lora_request: Optional[LoRARequest],
@ -511,7 +512,7 @@ class InputPreprocessor:
self,
inputs: SingletonInputs,
decoder_inputs_to_override: Optional[SingletonInputs] = None,
) -> Tuple[SingletonInputs, SingletonInputs]:
) -> tuple[SingletonInputs, SingletonInputs]:
"""
For encoder/decoder models only:
Separate Encoder/Decoder inputs from a MultiModalEncDecInputs

5
vllm/inputs/registry.py
View File

@ -2,9 +2,10 @@
import functools
from collections import UserDict
from collections.abc import Mapping
from dataclasses import dataclass
from typing import (TYPE_CHECKING, Any, Callable, Mapping, NamedTuple,
Optional, Protocol, Union)
from typing import (TYPE_CHECKING, Any, Callable, NamedTuple, Optional,
Protocol, Union)
from torch import nn
from transformers import BatchFeature, PretrainedConfig, ProcessorMixin

7
vllm/multimodal/base.py
View File

@ -2,9 +2,10 @@
from abc import ABC, abstractmethod
from collections import defaultdict
from collections.abc import Sequence
from pathlib import Path
from typing import (TYPE_CHECKING, Any, Callable, Generic, NamedTuple,
Optional, Sequence, Tuple, Type, TypeVar, Union)
Optional, TypeVar, Union)
from torch import nn
@ -39,7 +40,7 @@ model. This does not include tokens that correspond to the input text.
"""
_T = TypeVar("_T")
N = TypeVar("N", bound=Type[nn.Module])
N = TypeVar("N", bound=type[nn.Module])
class MultiModalPlugin(ABC):
@ -274,7 +275,7 @@ class MultiModalPlaceholderMap:
@classmethod
def from_seq_group(
cls, seq_group: "SequenceGroupMetadata", positions: range
) -> Tuple[Optional[MultiModalDataDict], dict[str,
) -> tuple[Optional[MultiModalDataDict], dict[str,
"MultiModalPlaceholderMap"]]:
"""
Returns the multi-modal items that intersect with the portion of a

3
vllm/multimodal/hasher.py
View File

@ -1,7 +1,8 @@
# SPDX-License-Identifier: Apache-2.0
import pickle
from typing import TYPE_CHECKING, Iterable, Mapping, Optional
from collections.abc import Iterable, Mapping
from typing import TYPE_CHECKING, Optional
import numpy as np
import torch

4
vllm/multimodal/image.py
View File

@ -3,7 +3,7 @@
import base64
from io import BytesIO
from pathlib import Path
from typing import TYPE_CHECKING, Any, Dict, Optional
from typing import TYPE_CHECKING, Any, Optional
import torch
from PIL import Image
@ -31,7 +31,7 @@ class ImagePlugin(MultiModalPlugin):
def _get_hf_image_processor(
self,
model_config: "ModelConfig",
mm_processor_kwargs: Optional[Dict[str, Any]] = None,
mm_processor_kwargs: Optional[dict[str, Any]] = None,
):
if mm_processor_kwargs is None:
mm_processor_kwargs = {}

14
vllm/multimodal/registry.py
View File

@ -2,9 +2,9 @@
import functools
from collections import UserDict
from collections.abc import Mapping, Sequence
from dataclasses import dataclass
from typing import (TYPE_CHECKING, Any, Dict, Generic, Mapping, Optional,
Protocol, Sequence, Type, TypeVar)
from typing import TYPE_CHECKING, Any, Generic, Optional, Protocol, TypeVar
import torch.nn as nn
@ -29,7 +29,7 @@ if TYPE_CHECKING:
logger = init_logger(__name__)
N = TypeVar("N", bound=Type[nn.Module])
N = TypeVar("N", bound=type[nn.Module])
_I = TypeVar("_I", bound=BaseProcessingInfo)
_I_co = TypeVar("_I_co", bound=BaseProcessingInfo, covariant=True)
@ -83,13 +83,13 @@ class _ProcessorFactories(Generic[_I]):
return self.processor(info, dummy_inputs_builder, cache=cache)
class _MultiModalLimits(UserDict["ModelConfig", Dict[str, int]]):
class _MultiModalLimits(UserDict["ModelConfig", dict[str, int]]):
"""
Wraps `_limits_by_model` for a more informative error message
when attempting to access a model that does not exist.
"""
def __getitem__(self, key: "ModelConfig") -> Dict[str, int]:
def __getitem__(self, key: "ModelConfig") -> dict[str, int]:
try:
return super().__getitem__(key)
except KeyError as exc:
@ -170,7 +170,7 @@ class MultiModalRegistry:
self,
model_config: "ModelConfig",
data: MultiModalDataDict,
mm_processor_kwargs: Optional[Dict[str, Any]] = None,
mm_processor_kwargs: Optional[dict[str, Any]] = None,
) -> MultiModalKwargs:
"""
Apply an input mapper to the data passed to the model.
@ -184,7 +184,7 @@ class MultiModalRegistry:
Note:
This should be called after :meth:`init_mm_limits_per_prompt`.
"""
merged_dict: Dict[str, NestedTensors] = {}
merged_dict = dict[str, NestedTensors]()
for data_key, data_value in data.items():
plugin = self._get_plugin(data_key)

4
vllm/multimodal/video.py
View File

@ -4,7 +4,7 @@ import base64
from functools import partial
from io import BytesIO
from pathlib import Path
from typing import TYPE_CHECKING, Any, Dict, Optional
from typing import TYPE_CHECKING, Any, Optional
import numpy as np
import numpy.typing as npt
@ -39,7 +39,7 @@ class VideoPlugin(ImagePlugin):
def _get_hf_video_processor(
self,
model_config: "ModelConfig",
mm_processor_kwargs: Optional[Dict[str, Any]] = None,
mm_processor_kwargs: Optional[dict[str, Any]] = None,
):
if mm_processor_kwargs is None:
mm_processor_kwargs = {}

14
vllm/usage/usage_lib.py
View File

@ -9,7 +9,7 @@ import time
from enum import Enum
from pathlib import Path
from threading import Thread
from typing import Any, Dict, Optional, Union
from typing import Any, Optional, Union
from uuid import uuid4
import cpuinfo
@ -27,7 +27,7 @@ _USAGE_STATS_DO_NOT_TRACK_PATH = os.path.join(_config_home, "do_not_track")
_USAGE_STATS_ENABLED = None
_USAGE_STATS_SERVER = envs.VLLM_USAGE_STATS_SERVER
_GLOBAL_RUNTIME_DATA: Dict[str, Union[str, int, bool]] = {}
_GLOBAL_RUNTIME_DATA = dict[str, Union[str, int, bool]]()
_USAGE_ENV_VARS_TO_COLLECT = [
"VLLM_USE_MODELSCOPE",
@ -150,7 +150,7 @@ class UsageMessage:
def report_usage(self,
model_architecture: str,
usage_context: UsageContext,
extra_kvs: Optional[Dict[str, Any]] = None) -> None:
extra_kvs: Optional[dict[str, Any]] = None) -> None:
t = Thread(target=self._report_usage_worker,
args=(model_architecture, usage_context, extra_kvs or {}),
daemon=True)
@ -158,13 +158,13 @@ class UsageMessage:
def _report_usage_worker(self, model_architecture: str,
usage_context: UsageContext,
extra_kvs: Dict[str, Any]) -> None:
extra_kvs: dict[str, Any]) -> None:
self._report_usage_once(model_architecture, usage_context, extra_kvs)
self._report_continous_usage()
def _report_usage_once(self, model_architecture: str,
usage_context: UsageContext,
extra_kvs: Dict[str, Any]) -> None:
extra_kvs: dict[str, Any]) -> None:
# Platform information
from vllm.platforms import current_platform
if current_platform.is_cuda_alike():
@ -227,7 +227,7 @@ class UsageMessage:
self._write_to_file(data)
self._send_to_server(data)
def _send_to_server(self, data: Dict[str, Any]) -> None:
def _send_to_server(self, data: dict[str, Any]) -> None:
try:
global_http_client = global_http_connection.get_sync_client()
global_http_client.post(_USAGE_STATS_SERVER, json=data)
@ -235,7 +235,7 @@ class UsageMessage:
# silently ignore unless we are using debug log
logging.debug("Failed to send usage data to server")
def _write_to_file(self, data: Dict[str, Any]) -> None:
def _write_to_file(self, data: dict[str, Any]) -> None:
os.makedirs(os.path.dirname(_USAGE_STATS_JSON_PATH), exist_ok=True)
Path(_USAGE_STATS_JSON_PATH).touch(exist_ok=True)
with open(_USAGE_STATS_JSON_PATH, "a") as f: