add-or-fix-type-checking
ty, make typing, या make check-repo द्वारा पहचानी गई टूटी हुई टाइपिंग जाँचों को ठीक करता है। जब स्थानीय रन, CI, या PR लॉग में टाइपिंग त्रुटियाँ दिखाई दें तब उपयोग करें।
npx skills add https://github.com/huggingface/transformers --skill add-or-fix-type-checkingAdd Or Fix Type Checking
Input
<target>: module or directory to type-check (if known).- Optional
make typingor CI output showing typing failures.
Workflow
-
Identify scope from the failing run:
- If you already have
make typingor CI output, extract the failing file/module paths. - If not, run:
make typing - Choose the narrowest target that covers the failures.
- If you already have
-
Run
ty checkfor the target to get a focused baseline:ty check --respect-ignore-files --exclude '**/*_pb*' <target> -
Triage errors by category before fixing anything:
- Wrong/missing type annotations on signatures
- Attribute access on union types (for example
X | None) - Functions returning broad unions (for example
str | list | BatchEncoding) - Mixin/protocol self-type issues
- Dynamic attributes on objects or modules
- Third-party stub gaps (missing kwargs, missing
__version__, etc.)
-
Apply fixes using this priority order (simplest first):
a. Narrow unions with
isinstance()/if x is None/hasattr(). This is the primary tool for resolving union-type errors.tynarrows through all of these patterns, including the negative forms:# Narrow X | None — use `if ...: raise`, never `assert` if x is None: raise ValueError("x must not be None") x.method() # ty knows x is X here # Narrow str | UploadFile if isinstance(field, str): raise TypeError("Expected file upload, got string") await field.read() # ty knows field is UploadFile here # Narrow broad union parameters early in a function body # (common for methods accepting e.g. list | dict | BatchEncoding) if isinstance(encoded_inputs, (list, tuple)): raise TypeError("Expected a mapping, got sequence") encoded_inputs.keys() # ty sees only the dict/mapping types nowb. Use local variables to help ty track narrowing across closures. When
self.xisX | Noneand you need to pass it to nested functions or closures,tycannot track thatself.xstays non-None. Copy to a local variable and narrow the local:manager = self.batching_manager if manager is None: raise RuntimeError("Manager not initialized") # Use `manager` (not `self.batching_manager`) in nested functionsc. Split chained calls when the intermediate type is a broad union. If
func().method()fails becausefunc()returns a union, split it:# BAD: ty can't narrow through chained calls result = func(return_dict=True).to(device)["input_ids"] # GOOD: split, narrow, then chain result = func(return_dict=True) if not hasattr(result, "to"): raise TypeError("Expected dict-like result") inputs = result.to(device)["input_ids"]d. Fix incorrect type hints at the source. If a parameter is typed
X | Nonebut can never beNonewhen actually called, removeNonefrom the hint.e. Annotate untyped attributes. Add type annotations to instance variables set in
__init__or elsewhere (for exampleself.foo: list[int] = []). Declare class-level attributes that are set dynamically later (for example_cache: Cache,_token_tensor: torch.Tensor | None).f. Use
@overloadfor methods with input-dependent return types. When a method returns different types based on the input type (e.g.__getitem__with str vs int keys), use@overloadto declare each signature separately:from typing import overload @overload def __getitem__(self, item: str) -> ValueType: ... @overload def __getitem__(self, item: int) -> EncodingType: ... @overload def __getitem__(self, item: slice) -> dict[str, ValueType]: ... def __getitem__(self, item: int | str | slice) -> ValueType | EncodingType | dict[str, ValueType]: ... # actual implementationThis eliminates
cast()calls at usage sites by giving the checker precise return types for each call pattern.g. Make container classes generic to propagate value types. When a class like
UserDictholds values whose type changes after transformation (e.g. lists → tensors after.to()), make the class generic so methods can return narrowed types:from typing import Generic, overload from typing_extensions import TypeVar _V = TypeVar("_V", default=Any) # default=Any keeps existing code working class MyDict(UserDict, Generic[_V]): @overload def __getitem__(self, item: str) -> _V: ... # ... def to(self, device) -> MyDict[torch.Tensor]: # after .to(), values are tensors ... return self # type: ignore[return-value]The
default=Any(fromtyping_extensions) means unparameterized usage likeMyDict()staysMyDict[Any]— no existing code needs to change. Only methods that narrow the value type (like.to()) declare a specific return type. This eliminatescast()at all call sites.h. Use
self: "ProtocolType"for mixins. When a mixin accesses attributes from its host class, define a Protocol insrc/transformers/_typing.pyand annotateselfon methods that need it. Apply this consistently to all methods in the mixin. Import underTYPE_CHECKINGto avoid circular imports.i. Use
TypeGuardfunctions for dynamic module attributes (for exampletorch.npu,torch.xpu,torch.compiler). Instead ofgetattr(torch, "npu")orhasattr(torch, "npu") and torch.npu.is_available(), define a type guard function insrc/transformers/_typing.py:def has_torch_npu(mod: ModuleType) -> TypeGuard[Any]: return hasattr(mod, "npu") and mod.npu.is_available()Then use it as a narrowing check:
if has_torch_npu(torch): torch.npu.device_count(). After the guard,tytreats the module asAny, allowing attribute access withoutgetattr()orcast(). See existing guards in_typing.pyfor all device backends.Key rules for type guards:
- Use
TypeGuard[Any](not a Protocol) — this is the simplest form that works withtyand avoids losing the original module's known attributes. - The guard function must be called directly in an
ifcondition for narrowing to work.tydoes NOT narrow throughandconditions orif not guard: return. - Import guards with
from .._typing import has_torch_xxx(not via module attribute_typing.has_torch_xxx) —tyonly resolvesTypeGuardfrom direct imports.
j. Use
getattr()/setattr()for dynamic model/config attributes. For runtime-injected fields (for example config/model flags), usegetattr(obj, "field", default)for reads andsetattr(obj, "field", value)for writes. Also usegetattr()for third-party packages missing type stubs (for examplegetattr(safetensors, "__version__", "unknown")). Avoidgetattr(torch, "npu")style — use type guards instead (see above).k. Use
cast()as a last resort before# type: ignore. Use when you've structurally validated the type but the checker can't see it: pattern-matched AST nodes, known-typed dict values, or validated API responses.# After structural validation confirms the type: stmt = cast(cst.Assign, node.body[0]) annotations = cast(list[Annotation], [])Do not use
cast()for module attribute narrowing — use type guards. Do not usecast()when@overloador generics can solve it at the source.l. Use
# type: ignoreonly for third-party stub defects. This means cases where the third-party package's type stubs are wrong or incomplete and there is no way to narrow or cast around it. Examples:- A kwarg that exists at runtime but is missing from the stubs
- A method that exists but isn't declared in the stubs
Always add the specific error code:
# type: ignore[call-arg], not bare# type: ignore.
- Use
-
Things to never do:
- Never use
assertfor type narrowing. Asserts are stripped bypython -Oand must not be relied on for correctness. Useif ...: raiseinstead. - Never use
# type: ignoreas a first resort. Exhaust all approaches above first. - Do not use
getattr(torch, "backend")to access dynamic device backends (npu,xpu,hpu,musa,mlu,neuron,compiler) — use type guards - Do not use
cast()for module attribute narrowing — use type guards - Do not use
cast()when@overloador generics can eliminate it at the source - Do not add helper methods or abstractions just to satisfy the type checker (especially for only 1-2 occurrences)
- Do not pollute base classes with domain-specific fields; use Protocols
- Do not add
if x is not Noneguards for values guaranteed non-None by the call chain; fix the annotation instead - Do not use conditional inheritance patterns; annotate
selfinstead
- Never use
-
Organization:
- Keep shared Protocols and type aliases in
src/transformers/_typing.py - Import type-only symbols under
if TYPE_CHECKING:to avoid circular deps - Use
from __future__ import annotationsfor PEP 604 syntax (X | Y)
- Keep shared Protocols and type aliases in
-
Verify and close the PR loop:
- Re-run
ty checkon the same<target> - Re-run
make typingto confirm the type/model-rules step passes - If working toward merge readiness, run
make check-repo - Ensure runtime behavior did not change and run relevant tests
- Re-run
-
Update CI coverage when adding new typed areas:
- Update
ty_check_dirsinMakefileto include newly type-checked directories.
- Update