Utils.initializer(初始化) 模块

ppsci.utils.initializer

The initialization method under this module is aligned with pytorch initialization. If you need to use the initialization method of PaddlePaddle, please refer to paddle.nn.initializer

This code is based on torch.nn.init Ths copyright of pytorch/pytorch is a BSD-style license, as found in the LICENSE file.

uniform_(tensor, a, b)

Modify tensor inplace using uniform_.

Parameters:

Name	Type	Description	Default
tensor	Tensor	Paddle Tensor.	required
a	float	min value.	required
b	float	max value.	required

Returns:

Type	Description
Tensor	paddle.Tensor: Initialized tensor.

Examples:

>>> import paddle
>>> import ppsci
>>> param = paddle.empty((128, 256), "float32")
>>> param = ppsci.utils.initializer.uniform_(param, -1, 1)

Source code in ppsci/utils/initializer.py

def uniform_(tensor: paddle.Tensor, a: float, b: float) -> paddle.Tensor:
    """Modify tensor inplace using uniform_.

    Args:
        tensor (paddle.Tensor): Paddle Tensor.
        a (float): min value.
        b (float): max value.

    Returns:
        paddle.Tensor: Initialized tensor.

    Examples:
        >>> import paddle
        >>> import ppsci
        >>> param = paddle.empty((128, 256), "float32")
        >>> param = ppsci.utils.initializer.uniform_(param, -1, 1)
    """
    return _no_grad_uniform_(tensor, a, b)

normal_(tensor, mean=0.0, std=1.0)

Modify tensor inplace using normal_.

Parameters:

Name	Type	Description	Default
tensor	Tensor	Paddle Tensor.	required
mean	float	mean value. Defaults to 0.0.	0.0
std	float	std value. Defaults to 1.0.	1.0

Returns:

Type	Description
Tensor	paddle.Tensor: Initialized tensor.

Examples:

>>> import paddle
>>> import ppsci
>>> param = paddle.empty((128, 256), "float32")
>>> param = ppsci.utils.initializer.normal_(param, 0, 1)

Source code in ppsci/utils/initializer.py

def normal_(
    tensor: paddle.Tensor, mean: float = 0.0, std: float = 1.0
) -> paddle.Tensor:
    """Modify tensor inplace using normal_.

    Args:
        tensor (paddle.Tensor): Paddle Tensor.
        mean (float, optional): mean value. Defaults to 0.0.
        std (float, optional): std value. Defaults to 1.0.

    Returns:
        paddle.Tensor: Initialized tensor.

    Examples:
        >>> import paddle
        >>> import ppsci
        >>> param = paddle.empty((128, 256), "float32")
        >>> param = ppsci.utils.initializer.normal_(param, 0, 1)
    """
    return _no_grad_normal_(tensor, mean, std)

trunc_normal_(tensor, mean=0.0, std=1.0, a=-2.0, b=2.0)

Modify tensor inplace using trunc_normal_.

Parameters:

Name	Type	Description	Default
tensor	Tensor	Paddle Tensor.	required
mean	float	The mean of the normal distribution. Defaults to 0.0.	0.0
std	float	The standard deviation of the normal distribution. Defaults to 1.0.	1.0
a	float	The minimum cutoff value. Defaults to -2.0.	-2.0
b	float	The maximum cutoff value. Defaults to 2.0.	2.0

Returns:

Type	Description
Tensor	paddle.Tensor: Initialized tensor.

Examples:

>>> import paddle
>>> import ppsci
>>> param = paddle.empty((128, 256), "float32")
>>> param = ppsci.utils.initializer.trunc_normal_(param, 0.0, 1.0)

Source code in ppsci/utils/initializer.py

def trunc_normal_(
    tensor: paddle.Tensor,
    mean: float = 0.0,
    std: float = 1.0,
    a: float = -2.0,
    b: float = 2.0,
) -> paddle.Tensor:
    """Modify tensor inplace using trunc_normal_.

    Args:
        tensor (paddle.Tensor): Paddle Tensor.
        mean (float, optional): The mean of the normal distribution. Defaults to 0.0.
        std (float, optional): The standard deviation of the normal distribution. Defaults to 1.0.
        a (float, optional): The minimum cutoff value. Defaults to -2.0.
        b (float, optional): The maximum cutoff value. Defaults to 2.0.

    Returns:
        paddle.Tensor: Initialized tensor.

    Examples:
        >>> import paddle
        >>> import ppsci
        >>> param = paddle.empty((128, 256), "float32")
        >>> param = ppsci.utils.initializer.trunc_normal_(param, 0.0, 1.0)
    """
    return _no_grad_trunc_normal_(tensor, mean, std, a, b)

constant_(tensor, value=0.0)

Modify tensor inplace using constant_.

Parameters:

Name	Type	Description	Default
tensor	Tensor	Paddle Tensor.	required
value	float	value to fill tensor. Defaults to 0.0.	0.0

Returns:

Type	Description
Tensor	paddle.Tensor: Initialized tensor.

Examples:

>>> import paddle
>>> import ppsci
>>> param = paddle.empty((128, 256), "float32")
>>> param = ppsci.utils.initializer.constant_(param, 2)

Source code in ppsci/utils/initializer.py

def constant_(tensor: paddle.Tensor, value: float = 0.0) -> paddle.Tensor:
    """Modify tensor inplace using constant_.

    Args:
        tensor (paddle.Tensor): Paddle Tensor.
        value (float, optional): value to fill tensor. Defaults to 0.0.

    Returns:
        paddle.Tensor: Initialized tensor.

    Examples:
        >>> import paddle
        >>> import ppsci
        >>> param = paddle.empty((128, 256), "float32")
        >>> param = ppsci.utils.initializer.constant_(param, 2)
    """
    return _no_grad_fill_(tensor, value)

ones_(tensor)

Modify tensor inplace using ones_.

Parameters:

Name	Type	Description	Default
tensor	Tensor	Paddle Tensor.	required

Returns:

Type	Description
Tensor	paddle.Tensor: Initialized tensor.

Examples:

>>> import paddle
>>> import ppsci
>>> param = paddle.empty((128, 256), "float32")
>>> param = ppsci.utils.initializer.ones_(param)

Source code in ppsci/utils/initializer.py

def ones_(tensor: paddle.Tensor) -> paddle.Tensor:
    """Modify tensor inplace using ones_.

    Args:
        tensor (paddle.Tensor): Paddle Tensor.

    Returns:
        paddle.Tensor: Initialized tensor.

    Examples:
        >>> import paddle
        >>> import ppsci
        >>> param = paddle.empty((128, 256), "float32")
        >>> param = ppsci.utils.initializer.ones_(param)
    """
    return _no_grad_fill_(tensor, 1)

zeros_(tensor)

Modify tensor inplace using zeros_.

Parameters:

Name	Type	Description	Default
tensor	Tensor	Paddle Tensor.	required

Returns:

Type	Description
Tensor	paddle.Tensor: Initialized tensor.

Examples:

>>> import paddle
>>> import ppsci
>>> param = paddle.empty((128, 256), "float32")
>>> param = ppsci.utils.initializer.zeros_(param)

Source code in ppsci/utils/initializer.py

def zeros_(tensor: paddle.Tensor) -> paddle.Tensor:
    """Modify tensor inplace using zeros_.

    Args:
        tensor (paddle.Tensor): Paddle Tensor.

    Returns:
        paddle.Tensor: Initialized tensor.

    Examples:
        >>> import paddle
        >>> import ppsci
        >>> param = paddle.empty((128, 256), "float32")
        >>> param = ppsci.utils.initializer.zeros_(param)
    """
    return _no_grad_fill_(tensor, 0)

xavier_uniform_(tensor, gain=1.0, reverse=False)

Modify tensor inplace using xavier_uniform_.

Parameters:

Name	Type	Description	Default
tensor	Tensor	Paddle Tensor.	required
gain	float	Hyperparameter. Defaults to 1.0.	1.0
reverse	bool	Tensor data format order, False by default as [fout, fin, ...].. Defaults to False.	False

Returns:

Type	Description
Tensor	paddle.Tensor: Initialized tensor.

Examples:

>>> import paddle
>>> import ppsci
>>> param = paddle.empty((128, 256), "float32")
>>> param = ppsci.utils.initializer.xavier_uniform_(param)

Source code in ppsci/utils/initializer.py

def xavier_uniform_(
    tensor: paddle.Tensor, gain: float = 1.0, reverse: bool = False
) -> paddle.Tensor:
    """Modify tensor inplace using xavier_uniform_.

    Args:
        tensor (paddle.Tensor): Paddle Tensor.
        gain (float, optional): Hyperparameter. Defaults to 1.0.
        reverse (bool, optional): Tensor data format order, False by default as
            [fout, fin, ...].. Defaults to False.

    Returns:
        paddle.Tensor: Initialized tensor.

    Examples:
        >>> import paddle
        >>> import ppsci
        >>> param = paddle.empty((128, 256), "float32")
        >>> param = ppsci.utils.initializer.xavier_uniform_(param)
    """
    fan_in, fan_out = _calculate_fan_in_and_fan_out(tensor, reverse=reverse)
    std = gain * math.sqrt(2.0 / float(fan_in + fan_out))
    k = math.sqrt(3.0) * std
    return _no_grad_uniform_(tensor, -k, k)

xavier_normal_(tensor, gain=1.0, reverse=False)

Modify tensor inplace using xavier_normal_.

Parameters:

Name	Type	Description	Default
tensor	Tensor	Paddle Tensor.	required
gain	float	Hyperparameter. Defaults to 1.0.	1.0
reverse	bool	tensor data format order, False by default as [fout, fin, ...]. Defaults to False.	False

Returns:

Type	Description
Tensor	paddle.Tensor: Initialized tensor.

Examples:

>>> import paddle
>>> import ppsci
>>> param = paddle.empty((128, 256), "float32")
>>> param = ppsci.utils.initializer.xavier_normal_(param)

Source code in ppsci/utils/initializer.py

def xavier_normal_(
    tensor: paddle.Tensor, gain: float = 1.0, reverse: bool = False
) -> paddle.Tensor:
    """Modify tensor inplace using xavier_normal_.

    Args:
        tensor (paddle.Tensor): Paddle Tensor.
        gain (float, optional): Hyperparameter. Defaults to 1.0.
        reverse (bool, optional): tensor data format order, False by
            default as [fout, fin, ...]. Defaults to False.

    Returns:
        paddle.Tensor: Initialized tensor.

    Examples:
        >>> import paddle
        >>> import ppsci
        >>> param = paddle.empty((128, 256), "float32")
        >>> param = ppsci.utils.initializer.xavier_normal_(param)
    """
    fan_in, fan_out = _calculate_fan_in_and_fan_out(tensor, reverse=reverse)
    std = gain * math.sqrt(2.0 / float(fan_in + fan_out))
    return _no_grad_normal_(tensor, 0, std)

kaiming_uniform_(tensor, a=0, mode='fan_in', nonlinearity='leaky_relu', reverse=False)

Modify tensor inplace using kaiming_uniform method.

Parameters:

Name	Type	Description	Default
tensor	Tensor	Paddle Tensor.	required
a	float	The negative slope of the rectifier used after this layer. Defaults to 0.	0
mode	Literal["fan_in", "fan_out"]	["fan_in", "fan_out"]. Defaults to "fan_in".	'fan_in'
nonlinearity	str	Nonlinearity method name. Defaults to "leaky_relu".	'leaky_relu'
reverse	bool	tensor data format order, False by default as [fout, fin, ...].. Defaults to False.	False

Returns:

Type	Description
Tensor	paddle.Tensor: Initialized tensor.

Examples:

>>> import paddle
>>> import ppsci
>>> param = paddle.empty((128, 256), "float32")
>>> param = ppsci.utils.initializer.kaiming_uniform_(param)

Source code in ppsci/utils/initializer.py

def kaiming_uniform_(
    tensor: paddle.Tensor,
    a: float = 0,
    mode: Literal["fan_in", "fan_out"] = "fan_in",
    nonlinearity: str = "leaky_relu",
    reverse: bool = False,
) -> paddle.Tensor:
    """Modify tensor inplace using kaiming_uniform method.

    Args:
        tensor (paddle.Tensor):  Paddle Tensor.
        a (float, optional): The negative slope of the rectifier used after this layer.
            Defaults to 0.
        mode (Literal["fan_in", "fan_out"], optional):
            ["fan_in", "fan_out"]. Defaults to "fan_in".
        nonlinearity (str, optional): Nonlinearity method name. Defaults to "leaky_relu".
        reverse (bool, optional): tensor data format order, False by default as
            [fout, fin, ...].. Defaults to False.

    Returns:
        paddle.Tensor: Initialized tensor.

    Examples:
        >>> import paddle
        >>> import ppsci
        >>> param = paddle.empty((128, 256), "float32")
        >>> param = ppsci.utils.initializer.kaiming_uniform_(param)
    """
    fan = _calculate_correct_fan(tensor, mode, reverse)
    gain = _calculate_gain(nonlinearity, a)
    std = gain / math.sqrt(fan)
    k = math.sqrt(3.0) * std
    return _no_grad_uniform_(tensor, -k, k)

kaiming_normal_(tensor, a=0, mode='fan_in', nonlinearity='leaky_relu', reverse=False)

Modify tensor inplace using kaiming_normal_.

Parameters:

Name	Type	Description	Default
tensor	Tensor	Paddle Tensor.	required
a	float	The negative slope of the rectifier used after this layer. Defaults to 0.	0
mode	Literal["fan_in", "fan_out"]	Either 'fan_in' (default) or 'fan_out'. Defaults to "fan_in".	'fan_in'
nonlinearity	str	Nonlinearity method name. Defaults to "leaky_relu".	'leaky_relu'
reverse	bool	Tensor data format order. Defaults to False.	False

Returns:

Type	Description
Tensor	paddle.Tensor: Initialized tensor.

Examples:

>>> import paddle
>>> import ppsci
>>> param = paddle.empty((128, 256), "float32")
>>> param = ppsci.utils.initializer.kaiming_normal_(param)

Source code in ppsci/utils/initializer.py

def kaiming_normal_(
    tensor: paddle.Tensor,
    a: float = 0,
    mode: Literal["fan_in", "fan_out"] = "fan_in",
    nonlinearity: str = "leaky_relu",
    reverse: bool = False,
) -> paddle.Tensor:
    """Modify tensor inplace using kaiming_normal_.

    Args:
        tensor (paddle.Tensor): Paddle Tensor.
        a (float, optional): The negative slope of the rectifier used after this layer.
            Defaults to 0.
        mode (Literal["fan_in", "fan_out"], optional): Either
            'fan_in' (default) or 'fan_out'. Defaults to "fan_in".
        nonlinearity (str, optional): Nonlinearity method name. Defaults to "leaky_relu".
        reverse (bool, optional): Tensor data format order. Defaults to False.

    Returns:
        paddle.Tensor: Initialized tensor.

    Examples:
        >>> import paddle
        >>> import ppsci
        >>> param = paddle.empty((128, 256), "float32")
        >>> param = ppsci.utils.initializer.kaiming_normal_(param)
    """
    fan = _calculate_correct_fan(tensor, mode, reverse)
    gain = _calculate_gain(nonlinearity, a)
    std = gain / math.sqrt(fan)
    return _no_grad_normal_(tensor, 0, std)

linear_init_(module)

Initialize module's weight and bias as it is a linear layer.

Parameters:

Name	Type	Description	Default
module	Layer	Linear Layer to be initialized.	required

Source code in ppsci/utils/initializer.py

def linear_init_(module: nn.Layer) -> None:
    """Initialize module's weight and bias as it is a linear layer.

    Args:
        module (nn.Layer): Linear Layer to be initialized.
    """
    kaiming_uniform_(module.weight, a=math.sqrt(5))
    if module.bias is not None:
        fan_in, _ = _calculate_fan_in_and_fan_out(module.weight, reverse=True)
        bound = 1 / math.sqrt(fan_in) if fan_in > 0 else 0
        uniform_(module.bias, -bound, bound)

conv_init_(module)

Initialize module's weight and bias as it is a conv layer.

Parameters:

Name	Type	Description	Default
module	Layer	Convolution Layer to be initialized.	required

Source code in ppsci/utils/initializer.py

def conv_init_(module: nn.Layer) -> None:
    """Initialize module's weight and bias as it is a conv layer.

    Args:
        module (nn.Layer): Convolution Layer to be initialized.
    """
    kaiming_uniform_(module.weight, a=math.sqrt(5))
    if module.bias is not None:
        fan_in, _ = _calculate_fan_in_and_fan_out(module.weight, reverse=False)
        if fan_in != 0:
            bound = 1 / math.sqrt(fan_in)
            uniform_(module.bias, -bound, bound)

---

最后更新: November 17, 2023
创建日期: November 6, 2023