Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

[Typing][C-112,C-114] Add type annotations for python/paddle/base/{initializer,param_attr.py} #67001

Merged
merged 4 commits into from
Aug 5, 2024
Merged

[Typing][C-112,C-114] Add type annotations for python/paddle/base/{initializer,param_attr.py} #67001

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
ee0a40b
[Typing][C-114] Add type annotations for `python/paddle/base/param_at…
SigureMo Aug 4, 2024
bab5991
add c-112
SigureMo Aug 4, 2024
5da9c0c
add pep 563
SigureMo Aug 4, 2024
c388012
pep 585 upgrade
SigureMo Aug 5, 2024
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
19 changes: 14 additions & 5 deletions python/paddle/base/initializer.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -12,31 +12,40 @@
# See the License for the specific language governing permissions and
# limitations under the License.

from __future__ import annotations

from typing import TYPE_CHECKING

import paddle

from .data_feeder import check_type

if TYPE_CHECKING:
from paddle.nn.initializer import Initializer

__all__ = []

_global_weight_initializer_ = None
_global_bias_initializer_ = None


def _global_weight_initializer():
def _global_weight_initializer() -> Initializer | None:
"""
Return the global weight initializer, The user doesn't need to use it.
"""
return _global_weight_initializer_


def _global_bias_initializer():
def _global_bias_initializer() -> Initializer | None:
"""
Return the global weight initializer, The user doesn't need to use it.
"""
return _global_bias_initializer_


def set_global_initializer(weight_init, bias_init=None):
def set_global_initializer(
weight_init: Initializer | None, bias_init: Initializer | None = None
) -> None:
"""
This API is used to set up global model parameter initializer in framework.

Expand All @@ -53,8 +62,8 @@ def set_global_initializer(weight_init, bias_init=None):
If you want to cancel the global initializer in framework, please set global initializer to ``None`` .

Args:
weight_init (Initializer): set the global initializer for ``weight`` of model parameters.
bias_init (Initializer, optional): set the global initializer for ``bias`` of model parameters.
weight_init (Initializer|None): set the global initializer for ``weight`` of model parameters.
bias_init (Initializer|None, optional): set the global initializer for ``bias`` of model parameters.
Default: None.

Returns:
Expand Down
106 changes: 63 additions & 43 deletions python/paddle/base/param_attr.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -13,14 +13,19 @@
# limitations under the License.
from __future__ import annotations

from typing import TYPE_CHECKING, Sequence, overload
from typing import TYPE_CHECKING, Any, ClassVar, overload

import paddle
from paddle.base.data_feeder import check_type
from paddle.regularizer import WeightDecayRegularizer

if TYPE_CHECKING:
from collections.abc import Sequence

from paddle import Tensor
from paddle._typing import ParamAttrLike
from paddle.nn.initializer import Initializer


__all__ = []

Expand Down Expand Up @@ -58,38 +63,48 @@ class ParamAttr:
need_clip (bool, optional): Whether the parameter gradient need to be clipped in optimizer. Default is True.

Returns:
ParamAttr Object.
ParamAttr Object.

Examples:

.. code-block:: python

>>> import paddle

>>> weight_attr = paddle.ParamAttr(name="weight",
... learning_rate=0.5,
... regularizer=paddle.regularizer.L2Decay(1.0),
... trainable=True)
>>> weight_attr = paddle.ParamAttr(
... name="weight",
... learning_rate=0.5,
... regularizer=paddle.regularizer.L2Decay(1.0),
... trainable=True,
... )
>>> print(weight_attr.name)
weight
>>> paddle.nn.Linear(3, 4, weight_attr=weight_attr)
"""

name: str | None
initializer: Initializer | None
learning_rate: float
regularizer: WeightDecayRegularizer | None
trainable: bool
do_model_average: bool
need_clip: bool

def __init__(
self,
name=None,
initializer=None,
learning_rate=1.0,
regularizer=None,
trainable=True,
do_model_average=True,
need_clip=True,
):
name: str | None = None,
initializer: Initializer | None = None,
learning_rate: float = 1.0,
regularizer: WeightDecayRegularizer | None = None,
trainable: bool = True,
do_model_average: bool = True,
need_clip: bool = True,
) -> None:
check_type(name, "name", (str, type(None)), "ParamAttr")
check_type(learning_rate, "learning_rate", (float, int), "ParamAttr")
check_type(trainable, "trainable", (bool), "ParamAttr")
check_type(do_model_average, "do_model_average", (bool), "ParamAttr")
check_type(need_clip, "need_clip", (bool), "ParamAttr")
check_type(trainable, "trainable", bool, "ParamAttr")
check_type(do_model_average, "do_model_average", bool, "ParamAttr")
check_type(need_clip, "need_clip", bool, "ParamAttr")
check_type(
initializer,
"initializer",
Expand All @@ -114,7 +129,7 @@ def __init__(
self.do_model_average = do_model_average
self.need_clip = need_clip

def _set_default_initializer(self, initializer):
def _set_default_initializer(self, initializer: Initializer | None) -> None:
"""
Set the default initializer, the initializer should be Constant,
Uniform, Normal, Xavier, MSRA.
Expand All @@ -135,7 +150,7 @@ def _set_default_initializer(self, initializer):

self.initializer = initializer

def _set_default_param_initializer(self):
def _set_default_param_initializer(self) -> None:
"""
Set the default initializer for the parameter with Xavier.

Expand All @@ -147,7 +162,7 @@ def _set_default_param_initializer(self):
"""
self._set_default_initializer(paddle.nn.initializer.XavierUniform())

def _set_default_bias_initializer(self):
def _set_default_bias_initializer(self) -> None:
"""
Set the default initializer for the bias with Constant(0.0).

Expand Down Expand Up @@ -207,7 +222,7 @@ def _to_attr(arg):
else:
raise TypeError(f"{type(arg)} cast to ParamAttr")

def _to_kwargs(self, with_initializer=False):
def _to_kwargs(self, with_initializer: bool = False) -> dict[str, Any]:
"""
Returns the attributes of this parameter.

Expand Down Expand Up @@ -284,36 +299,41 @@ class WeightNormParamAttr(ParamAttr):

>>> data = paddle.static.data(name="data", shape=[3, 32, 32], dtype="float32")

>>> fc = paddle.static.nn.fc(x=data,
... size=1000,
... weight_attr=paddle.static.WeightNormParamAttr(
... dim=None,
... name='weight_norm_param',
... initializer=paddle.nn.initializer.Constant(1.0),
... learning_rate=1.0,
... regularizer=paddle.regularizer.L2Decay(0.1),
... trainable=True,
... do_model_average=False,
... need_clip=True))
...
>>> fc = paddle.static.nn.fc(
... x=data,
... size=1000,
... weight_attr=paddle.static.WeightNormParamAttr(
... dim=None,
... name='weight_norm_param',
... initializer=paddle.nn.initializer.Constant(1.0),
... learning_rate=1.0,
... regularizer=paddle.regularizer.L2Decay(0.1),
... trainable=True,
... do_model_average=False,
... need_clip=True,
... ),
... )
"""

# List to record the parameters reparameterized by weight normalization.
# If these parameters are treated as Variable rather than Parameter,
# it can be used to discriminate these parameters and help to serialize
# these parameters for inference.
params_with_weight_norm = []
params_with_weight_norm: ClassVar[list[Tensor]] = []

dim: int | None

def __init__(
self,
dim=None,
name=None,
initializer=None,
learning_rate=1.0,
regularizer=None,
trainable=True,
do_model_average=False,
need_clip=True,
):
dim: int | None = None,
name: str | None = None,
initializer: Initializer | None = None,
learning_rate: float = 1.0,
regularizer: WeightDecayRegularizer | None = None,
trainable: bool = True,
do_model_average: bool = False,
need_clip: bool = True,
) -> None:
super().__init__(
name=name,
initializer=initializer,
Expand Down