[Typing][B-10] Add type annotations for python/paddle/distribution/distribution.py
#65769
Conversation
|
你的PR提交成功，感谢你对开源项目的贡献! |
| @@ -18,7 +18,10 @@ | |||
| # 'Normal', | |||
| # 'Uniform'] | |||
There was a problem hiding this comment.
这块 TODO 和 __all__ 的注释删掉吧，看着也没啥用
| def _to_tensor(self, *args): | ||
| def _to_tensor( | ||
| self, *args: TensorLike | NestedNumbericSequence | ||
| ) -> tuple[Tensor, ...]: |
There was a problem hiding this comment.
应该是 Tensor？
from typing import TypeGuard
from typing_extensions import TypeIs
def is_str_seq(*args: str | int) -> TypeGuard[tuple[str, ...]]:
return all(isinstance(arg, str) for arg in args)
def foo(x: str | int):
if is_str_seq(x, x):
reveal_type(x) # tuple[str, ...]
else:
reveal_type(x)这里好像 guard 住的是每个元素的类型
There was a problem hiding this comment.
应该是 Tensor？
？输出？
return tuple(variable_args)from typing import TypeGuard from typing_extensions import TypeIs def is_str_seq(self, *args: str | int) -> TypeGuard[tuple[str, ...]]: return all(isinstance(arg, str) for arg in args) def foo(x: str | int): if is_str_seq(x, x): reveal_type(x) # tuple[str, ...] else: reveal_type(x)这里好像 guard 住的是每个元素的类型
这是指这里？
def _validate_args(
self, *args: TensorLike | NestedNumbericSequence
) -> TypeGuard[tuple[Tensor, ...]]:There was a problem hiding this comment.
def _validate_args(
self, *args: TensorLike | NestedNumbericSequence
) -> TypeGuard[tuple[Tensor, ...]]:
"""
Argument validation for distribution args
Args:
value (float, list, numpy.ndarray, Tensor)
Raises
ValueError: if one argument is Tensor, all arguments should be Tensor
"""
is_variable = False
is_number = False
for arg in args:
if isinstance(arg, (Variable, paddle.pir.Value)):
is_variable = True
else:
is_number = True
if is_variable and is_number:
raise ValueError(
'if one argument is Tensor, all arguments should be Tensor'
)
return is_variable是 guard 每个元素 ～ 有啥问题？
There was a problem hiding this comment.
每个元素都是 Tensor？那为啥是 TypeGuard[tuple[Tensor, ...]？
There was a problem hiding this comment.
from typing import TypeGuard
from typing_extensions import TypeIs
def is_str_seq(*args: str | int) -> TypeGuard[str]:
return all(isinstance(arg, str) for arg in args)
def foo(x: str | int, y: str | int):
if is_str_seq(x, y):
reveal_type(x) # str
reveal_type(y) # str | int
else:
reveal_type(x)
reveal_type(y)我试了下 guard 不住，pyright 只能保证第一个参数 guard 住，这是 TypeGuard 规范中写的
mypy 甚至连第一个参数都 guard 不住
There was a problem hiding this comment.
每个元素都是 Tensor？那为啥是
TypeGuard[tuple[Tensor, ...]]？
这个方法的作用不就是保证每一个元素都是 Tensor ？ 输入是 *args ，所以用的 tuple[Tensor, ...] ～
我试了下 guard 不住，pyright 只能保证第一个参数 guard 住，这是 TypeGuard 规范中写的
那应该是实现的还不够完善吧 ～ 那咋办？写 TypeGuard[Tensor] ?
There was a problem hiding this comment.
就 TypeGuard[Tensor] 吧，大多数用例是没问题的，而且是非公开 API，也还好
PR Category
User Experience
PR Types
Improvements
Description
类型标注：
python/paddle/distribution/distribution.pyRelated links
@SigureMo @megemini