[Typing][C-95, C-96, C-99]Add annotations for `python/paddle/incubate/optimizer/{lookahead, modelaverage, lbfgs}.py` #67448

inaomIIsfarell · 2024-08-14T13:50:20Z

PR Category

User Experience

PR Types

Improvements

Description

类型标注：

python/paddle/incubate/optimizer/{lookahead, modelaverage, bfgs}.py

Related links

[Type Hints] 为公开 API 标注类型提示信息 #65008

@SigureMo @megemini

…ookahead.py`

…odelaverage.py`

…bfgs.py`

… c959699

paddle-bot · 2024-08-14T13:50:24Z

你的PR提交成功，感谢你对开源项目的贡献!
请关注后续CI自动化测试结果，详情请参考Paddle-CI手册。
Your PR has been submitted. Thanks for your contribution!
Please wait for the result of CI firstly. See Paddle CI Manual for details.

megemini

这几个类中实例的参数都没写，如：

class A:
    a: int
    def __init__(self)->None:
        self.a = 1
        self._b = 'b'

上面 A 中的 a ～只需要添加可公开属性即可，如 self._b 不需要标注～

另外，PR 标题中的标号写错了，应该是 C-95

megemini · 2024-08-15T05:38:06Z

python/paddle/incubate/optimizer/lbfgs.py


 from collections import defaultdict
 from functools import reduce
+from typing import TYPE_CHECKING, Callable


从 collections.abc 导入 Callable

megemini · 2024-08-15T05:40:21Z

python/paddle/incubate/optimizer/lbfgs.py

+        tolerance_change: float = 1e-9,
+        history_size: int = 100,
+        line_search_fn: str | None = None,
+        parameters: list[Tensor] | tuple[Tensor] | None = None,


Suggested change

parameters: list[Tensor] | tuple[Tensor] | None = None,

parameters: Sequence[Tensor] | Sequence[_ParameterConfig] | None = None,

参考 Optimizer 的初始化参数

megemini · 2024-08-15T05:45:47Z

python/paddle/incubate/optimizer/lbfgs.py

@@ -226,7 +232,7 @@ def _directional_evaluate(self, closure, x, alpha, d):
        self._set_param(x)
        return loss, flat_grad

-    def step(self, closure):
+    def step(self, closure: Callable) -> Tensor:


Suggested change

def step(self, closure: Callable) -> Tensor:

def step(self, closure: Callable[[], _T]) -> _T:

_T = TypeVar('_T', covariant=True)

megemini · 2024-08-15T05:50:23Z

python/paddle/incubate/optimizer/modelaverage.py

-        name=None,
-    ):
+        average_window_rate: float,
+        parameters: list[Tensor] | None = None,


megemini · 2024-08-15T05:51:59Z

python/paddle/incubate/optimizer/modelaverage.py

+        startup_program: Program | None = None,
+        parameters: list[Tensor] | list[str] | None = None,
+        no_grad_set: set[Tensor] | set[str] | None = None,
+    ) -> tuple[list[Operator], list[tuple[Tensor, Tensor]]]:


Suggested change

) -> tuple[list[Operator], list[tuple[Tensor, Tensor]]]:

) -> None:

应该是方法没写完整

megemini · 2024-08-15T05:53:57Z

python/paddle/incubate/optimizer/modelaverage.py

-    def apply(self, executor=None, need_restore=True):
+    def apply(
+        self, executor: Executor | None = None, need_restore: bool = True
+    ):


Suggested change

):

) -> Generator[None, None, None]:

…bfgs.py`

…ookahead.py`

…odelaverage.py`

… c959699

SigureMo · 2024-08-18T17:26:45Z

python/paddle/incubate/optimizer/lbfgs.py

+    tolerance_change: float
+    history_size: int
+    line_search_fn: str | None
+    state: dict[_KT, _VT]


这个 _KT、_VT 的意义是什么，单独使用一个泛型是没有任何约束的，和 dict[Any, Any] 没有区别

SigureMo · 2024-08-18T17:29:16Z

python/paddle/incubate/optimizer/lbfgs.py

+    tolerance_grad: float
+    tolerance_change: float
+    history_size: int
+    line_search_fn: str | None


Suggested change

line_search_fn: str | None

line_search_fn: Literal['strong_wolfe'] | None

下同

SigureMo · 2024-08-18T17:32:45Z

python/paddle/incubate/optimizer/lbfgs.py

+        weight_decay: NotRequired[float | WeightDecayRegularizer | None]
+        learning_rate: NotRequired[float | Tensor | LRScheduler | None]
+
+    _T = TypeVar('_T', covariant=True)


Suggested change

_T = TypeVar('_T', covariant=True)

_T_co = TypeVar('_T_co', covariant=True)

协变、逆变应该从名字上直接体现

SigureMo · 2024-08-18T17:33:28Z

python/paddle/incubate/optimizer/lookahead.py


 import paddle
 from paddle.base import framework, unique_name
 from paddle.base.dygraph import base as imperative_base
 from paddle.base.framework import Variable
-from paddle.base.layer_helper import LayerHelper
+from paddle.base.layer_helper import LayerHelper, LayerHelperBase


非运行时相关 import
放到 if TYPE_CHECKING: 下

SigureMo · 2024-08-18T17:33:48Z

python/paddle/incubate/optimizer/lookahead.py

 from paddle.framework import in_pir_mode
 from paddle.optimizer import Optimizer
 from paddle.pir.core import create_parameter

+if TYPE_CHECKING:
+    from paddle import Tensor
+    from paddle.base.framework import Operator, Program


Program 从 paddle.static 下 import

SigureMo · 2024-08-18T17:35:46Z

python/paddle/incubate/optimizer/modelaverage.py

+    from paddle.regularizer import WeightDecayRegularizer
+    from paddle.static import Executor
+
+    class _ParameterConfig(TypedDict):


这几个 _ParameterConfig，有什么理由不复用 python/paddle/optimizer/optimizer.py 中的 _ParameterConfig？

…ookahead.py`

…odelaverage.py`

…bfgs.py`

… c959699

megemini · 2024-08-20T10:28:17Z

python/paddle/incubate/optimizer/lbfgs.py

+    tolerance_change: float
+    history_size: int
+    line_search_fn: Literal['strong_wolfe'] | None
+    state: defaultdict


Suggested change

state: defaultdict

state: dict[str, dict[str, Any]]

megemini · 2024-08-20T10:30:35Z

python/paddle/incubate/optimizer/lbfgs.py

@@ -165,7 +185,7 @@ def __init__(

        self._numel_cache = None

-    def state_dict(self):
+    def state_dict(self) -> dict[str, Tensor]:


Suggested change

def state_dict(self) -> dict[str, Tensor]:

def state_dict(self) -> dict[str, dict[str, Any]]:

megemini · 2024-08-20T10:43:22Z

python/paddle/incubate/optimizer/lookahead.py

+    alpha: float
+    k: int
+    type: str
+    helper: LayerHelperBase | None


Suggested change

helper: LayerHelperBase | None

helper: LayerHelper

没必要用 LayerHelperBase 吧？

megemini · 2024-08-20T10:50:31Z

python/paddle/incubate/optimizer/modelaverage.py


 import paddle
 from paddle import _C_ops
 from paddle.base import framework
 from paddle.base.dygraph import base as imperative_base
-from paddle.base.layer_helper import LayerHelper
+from paddle.base.layer_helper import LayerHelper, LayerHelperBase


Suggested change

from paddle.base.layer_helper import LayerHelper, LayerHelperBase

from paddle.base.layer_helper import LayerHelper

貌似不需要 LayerHelperBase 吧

megemini · 2024-08-20T10:50:57Z

python/paddle/incubate/optimizer/modelaverage.py

+    from collections.abc import Generator, Sequence
+
+    from paddle import Tensor
+    from paddle.base.framework import Program


Suggested change

from paddle.base.framework import Program

from paddle.static import Program

megemini · 2024-08-20T10:51:58Z

python/paddle/incubate/optimizer/modelaverage.py

@@ -169,14 +181,22 @@ class ModelAverage(Optimizer):

    """

+    helper: LayerHelperBase | None


Suggested change

helper: LayerHelperBase | None

helper: LayerHelper

没有必要用 LayerHelperBase 吧？

…ookahead.py`

…odelaverage.py`

…bfgs.py`

… c959699

megemini

LGTM ~

inaomIIsfarell added 4 commits August 14, 2024 21:31

[Typing][C-95]Add annotations for `python/paddle/incubate/optimizer/l…

c58b86c

…ookahead.py`

[Typing][C-96]Add annotations for `python/paddle/incubate/optimizer/m…

9194f48

…odelaverage.py`

[Typing][C-99]Add annotations for `python/paddle/incubate/optimizer/l…

40a7daf

…bfgs.py`

Merge branch 'develop' of https://github.com/PaddlePaddle/Paddle into…

7961d6e

… c959699

paddle-bot bot added the contributor External developers label Aug 14, 2024

luotao1 added the HappyOpenSource 快乐开源活动issue与PR label Aug 15, 2024

luotao1 assigned luotao1 and SigureMo Aug 15, 2024

megemini reviewed Aug 15, 2024

View reviewed changes

This comment was marked as duplicate.

Sign in to view

megemini mentioned this pull request Aug 15, 2024

[Type Hints] 为公开 API 标注类型提示信息 #65008

Closed

inaomIIsfarell changed the title ~~[Typing][C-96, C-96, C-99]Add annotations for python/paddle/incubate/optimizer/{lookahead, modelaverage, lbfgs].py~~ [Typing][C-95, C-96, C-99]Add annotations for python/paddle/incubate/optimizer/{lookahead, modelaverage, lbfgs}.py Aug 15, 2024

megemini mentioned this pull request Aug 15, 2024

[Typing][C-40][BUAA] Add type annotations for 1 file in python/paddle/distributed/fleet/base/distributed_strategy.py #67405

Merged

inaomIIsfarell added 5 commits August 17, 2024 21:57

[Typing][C-99]Add annotations for `python/paddle/incubate/optimizer/l…

80e95f9

…bfgs.py`

[Typing][C-99]Add annotations for `python/paddle/incubate/optimizer/l…

8dd1d08

…bfgs.py`

[Typing][C-95]Add annotations for `python/paddle/incubate/optimizer/l…

c302da7

…ookahead.py`

[Typing][C-96]Add annotations for `python/paddle/incubate/optimizer/m…

d912ef0

…odelaverage.py`

Merge branch 'develop' of https://github.com/PaddlePaddle/Paddle into…

43a4f2b

… c959699

SigureMo reviewed Aug 18, 2024

View reviewed changes

inaomIIsfarell added 4 commits August 20, 2024 15:41

[Typing][C-95]Add annotations for `python/paddle/incubate/optimizer/l…

31e6dfd

…ookahead.py`

[Typing][C-96]Add annotations for `python/paddle/incubate/optimizer/m…

74c38ff

…odelaverage.py`

[Typing][C-99]Add annotations for `python/paddle/incubate/optimizer/l…

e6e0e4e

…bfgs.py`

Merge branch 'develop' of https://github.com/PaddlePaddle/Paddle into…

96ae0bf

… c959699

megemini reviewed Aug 20, 2024

View reviewed changes

inaomIIsfarell added 4 commits August 21, 2024 09:55

[Typing][C-95]Add annotations for `python/paddle/incubate/optimizer/l…

63f2e09

…ookahead.py`

[Typing][C-96]Add annotations for `python/paddle/incubate/optimizer/m…

3d061ce

…odelaverage.py`

[Typing][C-99]Add annotations for `python/paddle/incubate/optimizer/l…

e602923

…bfgs.py`

Merge branch 'develop' of https://github.com/PaddlePaddle/Paddle into…

2b49e70

… c959699

megemini approved these changes Aug 21, 2024

View reviewed changes

luotao1 approved these changes Aug 21, 2024

View reviewed changes

luotao1 merged commit 99daec4 into PaddlePaddle:develop Aug 21, 2024
28 checks passed

inaomIIsfarell mentioned this pull request Aug 22, 2024

[WeeklyReports] 2024.08.10~2024.08.23 周报收集 PFCCLab/Starter#340

Closed

35 tasks

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

[Typing][C-95, C-96, C-99]Add annotations for `python/paddle/incubate/optimizer/{lookahead, modelaverage, lbfgs}.py` #67448

[Typing][C-95, C-96, C-99]Add annotations for `python/paddle/incubate/optimizer/{lookahead, modelaverage, lbfgs}.py` #67448

inaomIIsfarell commented Aug 14, 2024 •

edited

Loading

paddle-bot bot commented Aug 14, 2024

megemini left a comment

megemini Aug 15, 2024

megemini Aug 15, 2024

megemini Aug 15, 2024

megemini Aug 15, 2024

megemini Aug 15, 2024

megemini Aug 15, 2024

This comment was marked as duplicate.

SigureMo Aug 18, 2024

SigureMo Aug 18, 2024

SigureMo Aug 18, 2024

SigureMo Aug 18, 2024

SigureMo Aug 18, 2024

SigureMo Aug 18, 2024

megemini Aug 20, 2024

megemini Aug 20, 2024

megemini Aug 20, 2024

megemini Aug 20, 2024

megemini Aug 20, 2024

megemini Aug 20, 2024

megemini left a comment

	parameters: list[Tensor] \| tuple[Tensor] \| None = None,
	parameters: Sequence[Tensor] \| Sequence[_ParameterConfig] \| None = None,

	def step(self, closure: Callable) -> Tensor:
	def step(self, closure: Callable[[], _T]) -> _T:

	) -> tuple[list[Operator], list[tuple[Tensor, Tensor]]]:
	) -> None:

	line_search_fn: str \| None
	line_search_fn: Literal['strong_wolfe'] \| None

	_T = TypeVar('_T', covariant=True)
	_T_co = TypeVar('_T_co', covariant=True)

	def state_dict(self) -> dict[str, Tensor]:
	def state_dict(self) -> dict[str, dict[str, Any]]:

	from paddle.base.layer_helper import LayerHelper, LayerHelperBase
	from paddle.base.layer_helper import LayerHelper

	from paddle.base.framework import Program
	from paddle.static import Program

		@@ -169,14 +181,22 @@ class ModelAverage(Optimizer):

		"""

		helper: LayerHelperBase \| None

[Typing][C-95, C-96, C-99]Add annotations for python/paddle/incubate/optimizer/{lookahead, modelaverage, lbfgs}.py #67448

[Typing][C-95, C-96, C-99]Add annotations for python/paddle/incubate/optimizer/{lookahead, modelaverage, lbfgs}.py #67448

Conversation

inaomIIsfarell commented Aug 14, 2024 • edited Loading

PR Category

PR Types

Description

Related links

paddle-bot bot commented Aug 14, 2024

megemini left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

This comment was marked as duplicate.

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

megemini left a comment

Choose a reason for hiding this comment

[Typing][C-95, C-96, C-99]Add annotations for `python/paddle/incubate/optimizer/{lookahead, modelaverage, lbfgs}.py` #67448

[Typing][C-95, C-96, C-99]Add annotations for `python/paddle/incubate/optimizer/{lookahead, modelaverage, lbfgs}.py` #67448

inaomIIsfarell commented Aug 14, 2024 •

edited

Loading