Make mypy opt-out instead of opt-in (#33205)

Signed-off-by: Harry Mellor <19981378+hmellor@users.noreply.github.com>

tools/pre_commit/mypy.py

@@ -23,39 +23,8 @@ import sys
 
 import regex as re
 
-FILES = [
-    "vllm/*.py",
-    "vllm/assets",
-    "vllm/compilation",
-    "vllm/distributed",
-    "vllm/engine",
-    "vllm/entrypoints",
-    "vllm/executor",
-    "vllm/inputs",
-    "vllm/logging_utils",
-    "vllm/multimodal",
-    "vllm/platforms",
-    "vllm/plugins",
-    "vllm/renderers",
-    "vllm/tokenizers",
-    "vllm/transformers_utils",
-    "vllm/triton_utils",
-    "vllm/usage",
-    "vllm/utils",
-    "vllm/worker",
-    "vllm/v1/attention",
-    "vllm/v1/core",
-    "vllm/v1/engine",
-    "vllm/v1/executor",
-    "vllm/v1/metrics",
-    "vllm/v1/pool",
-    "vllm/v1/sample",
-    "vllm/v1/structured_output",
-    "vllm/v1/worker",
-]
-
 # After fixing errors resulting from changing follow_imports
-# from "skip" to "silent", move the following directories to FILES
+# from "skip" to "silent", remove its directory from SEPARATE_GROUPS.
 SEPARATE_GROUPS = [
     "tests",
     # v0 related
@@ -74,6 +43,16 @@ EXCLUDE = [
     "vllm/model_executor/layers/fla/ops",
     # Ignore triton kernels in ops.
     "vllm/v1/attention/ops",
+    # TODO: Remove these entries after fixing mypy errors.
+    "vllm/benchmarks",
+    "vllm/config",
+    "vllm/device_allocator",
+    "vllm/profiler",
+    "vllm/reasoning",
+    "vllm/tool_parser",
+    "vllm/v1/cudagraph_dispatcher.py",
+    "vllm/outputs.py",
+    "vllm/logger.py",
 ]
 
 
@@ -88,7 +67,6 @@ def group_files(changed_files: list[str]) -> dict[str, list[str]]:
         A dictionary mapping file group names to lists of changed files.
     """
     exclude_pattern = re.compile(f"^{'|'.join(EXCLUDE)}.*")
-    files_pattern = re.compile(f"^({'|'.join(FILES)}).*")
     file_groups = {"": []}
     file_groups.update({k: [] for k in SEPARATE_GROUPS})
     for changed_file in changed_files:
@@ -96,14 +74,13 @@ def group_files(changed_files: list[str]) -> dict[str, list[str]]:
         if exclude_pattern.match(changed_file):
             continue
         # Group files by mypy call
-        if files_pattern.match(changed_file):
-            file_groups[""].append(changed_file)
-            continue
+        for directory in SEPARATE_GROUPS:
+            if re.match(f"^{directory}.*", changed_file):
+                file_groups[directory].append(changed_file)
+                break
         else:
-            for directory in SEPARATE_GROUPS:
-                if re.match(f"^{directory}.*", changed_file):
-                    file_groups[directory].append(changed_file)
-                    break
+            if changed_file.startswith("vllm/"):
+                file_groups[""].append(changed_file)
     return file_groups
 
 

vllm/_aiter_ops.py

@@ -349,7 +349,7 @@ def _rocm_aiter_mla_decode_fwd_impl(
 ) -> None:
     from aiter.mla import mla_decode_fwd
 
-    kwargs = {
+    kwargs: dict[str, float | torch.Tensor | None] = {
         "sm_scale": sm_scale,
         "logit_cap": logit_cap,
     }

vllm/config/compilation.py

@@ -570,7 +570,7 @@ class CompilationConfig:
     pass_config: PassConfig = field(default_factory=PassConfig)
     """Custom inductor passes, see PassConfig for more details"""
 
-    max_cudagraph_capture_size: int | None = field(default=None)
+    max_cudagraph_capture_size: int = field(default=None)
     """The maximum cudagraph capture size.
 
     If cudagraph_capture_sizes is specified, this will be set to the largest
@@ -743,6 +743,7 @@ class CompilationConfig:
         "level",
         "mode",
         "cudagraph_mode",
+        "max_cudagraph_capture_size",
         "use_inductor_graph_partition",
         mode="wrap",
     )

vllm/config/utils.py

@@ -9,7 +9,7 @@ import inspect
 import json
 import pathlib
 import textwrap
-from collections.abc import Callable, Iterable, Mapping, Sequence, Set
+from collections.abc import Callable, Mapping, Sequence, Set
 from dataclasses import MISSING, Field, dataclass, field, fields, is_dataclass, replace
 from itertools import pairwise
 from typing import TYPE_CHECKING, Any, Protocol, TypeVar
@@ -75,7 +75,7 @@ def get_field(cls: ConfigType, name: str) -> Field:
 
 def getattr_iter(
     object: object,
-    names: Iterable[str],
+    names: Sequence[str],
     default: Any | None = None,
     default_factory: Callable[[], Any] | None = None,
     warn: bool = False,

vllm/env_override.py

@@ -382,7 +382,7 @@ def _patch_get_raw_stream_if_needed():
         if hasattr(torch._C, "_cuda_getCurrentRawStream"):
             from torch._C import _cuda_getCurrentRawStream as _get_raw_stream
 
-            builtins.get_raw_stream = _get_raw_stream
+            builtins.get_raw_stream = _get_raw_stream  # type: ignore[attr-defined]
 
 
 _patch_get_raw_stream_if_needed()

vllm/envs.py

@@ -1680,6 +1680,7 @@ def disable_envs_cache() -> None:
     global __getattr__
     # If __getattr__ is wrapped by functions.cache, unwrap the caching layer.
     if _is_envs_cache_enabled():
+        assert hasattr(__getattr__, "__wrapped__")
         __getattr__ = __getattr__.__wrapped__
 
 

vllm/forward_context.py

@@ -270,7 +270,7 @@ def create_forward_context(
     cudagraph_runtime_mode: CUDAGraphMode = CUDAGraphMode.NONE,
     batch_descriptor: BatchDescriptor | None = None,
     ubatch_slices: UBatchSlices | None = None,
-    slot_mapping: dict[str, torch.Tensor] | None = None,
+    slot_mapping: dict[str, torch.Tensor] | list[dict[str, torch.Tensor]] | None = None,
     additional_kwargs: dict[str, Any] | None = None,
     skip_compiled: bool = False,
 ):

vllm/logger.py

@@ -157,7 +157,7 @@ _METHODS_TO_PATCH = {
 
 
 def _configure_vllm_root_logger() -> None:
-    logging_config = dict[str, Any]()
+    logging_config = dict[str, dict[str, Any] | Any]()
 
     if not envs.VLLM_CONFIGURE_LOGGING and envs.VLLM_LOGGING_CONFIG_PATH:
         raise RuntimeError(

vllm/logprobs.py

@@ -28,7 +28,7 @@ LogprobsOnePosition = dict[int, Logprob]
 
 
 @dataclass
-class FlatLogprobs(MutableSequence[LogprobsOnePosition]):
+class FlatLogprobs(MutableSequence[LogprobsOnePosition | None]):
     """
     Flat logprobs of a request into multiple primitive type lists.
 
@@ -140,7 +140,7 @@ class FlatLogprobs(MutableSequence[LogprobsOnePosition]):
     def __delitem__(self, item) -> None:
         raise TypeError("Cannot delete logprobs from FlatLogprobs")
 
-    def insert(self, item) -> None:
+    def insert(self, index: int, value: dict[int, Logprob] | None) -> None:
         raise TypeError("Cannot insert logprobs to FlatLogprobs")
 
     def __iter__(self) -> Iterator[LogprobsOnePosition]:
@@ -161,7 +161,7 @@ SampleLogprobs = FlatLogprobs | list[LogprobsOnePosition]
 
 def create_prompt_logprobs(flat_logprobs: bool) -> PromptLogprobs:
     """Creates a container to store prompt logprobs for a request"""
-    logprobs = FlatLogprobs() if flat_logprobs else []
+    logprobs: PromptLogprobs = FlatLogprobs() if flat_logprobs else []
     # NOTE: logprob of first prompt token is None.
     logprobs.append(None)
     return logprobs

vllm/profiler/utils.py

@@ -4,6 +4,7 @@
 import dataclasses
 from collections.abc import Callable
 
+from _typeshed import DataclassInstance
 from torch._C._profiler import _EventType, _ProfilerEvent, _TensorMetadata
 
 #
@@ -11,7 +12,7 @@ from torch._C._profiler import _EventType, _ProfilerEvent, _TensorMetadata
 #
 
 
-def trim_string_front(string, width):
+def trim_string_front(string: str, width: int) -> str:
     if len(string) > width:
         offset = len(string) - width + 3
         string = string[offset:]
@@ -20,7 +21,7 @@ def trim_string_front(string, width):
     return string
 
 
-def trim_string_back(string, width):
+def trim_string_back(string: str, width: int) -> str:
     if len(string) > width:
         offset = len(string) - width + 3
         string = string[:-offset]
@@ -30,15 +31,13 @@ def trim_string_back(string, width):
 
 
 class TablePrinter:
-    def __init__(
-        self, row_cls: type[dataclasses.dataclass], column_widths: dict[str, int]
-    ):
+    def __init__(self, row_cls: type[DataclassInstance], column_widths: dict[str, int]):
         self.row_cls = row_cls
         self.fieldnames = [x.name for x in dataclasses.fields(row_cls)]
         self.column_widths = column_widths
         assert set(self.column_widths.keys()) == set(self.fieldnames)
 
-    def print_table(self, rows: list[dataclasses.dataclass]):
+    def print_table(self, rows: list[DataclassInstance]):
         self._print_header()
         self._print_line()
         for row in rows:

vllm/v1/kv_cache_interface.py

@@ -98,7 +98,7 @@ class FullAttentionSpec(AttentionSpec):
     In this case, we use FullAttentionSpec and record the sliding window size.
     """
 
-    head_size_v: int | None = None
+    head_size_v: int = None  # type: ignore[assignment]
 
     sliding_window: int | None = None
     """