GroupNorm

class paddle.nn. GroupNorm ( num_groups: int, num_channels: int, epsilon: float = 1e-05, *, affine: bool = True, device: PlaceLike | None = None, dtype: DTypeLike | None = None, weight_attr: bool | ParamAttr | None = None, bias_attr: bool | ParamAttr | None = None, data_format: DataLayout1D | DataLayout2D | DataLayout3D = 'NCHW', name: str | None = None ) [source]

This interface is used to construct a callable object of the GroupNorm class. For more details, refer to code examples. It implements the function of the Group Normalization Layer. Refer to Group Normalization .

Parameters

  • num_groups (int) – The number of groups that divided from channels.

  • num_channels (int) – The number of channels of input.

  • epsilon (float, optional) – The small value added to the variance to prevent division by zero. Default: 1e-05. alias: eps.

  • affine (bool, optional) – Whether this module has learnable affine parameters (weight and bias). If set to False, no learnable parameters will be created, regardless of the settings of weight_attr and bias_attr. Defaults to True. Note: This argument must be passed as a keyword argument.

  • device (PlaceLike, optional) – Device where the computation takes place. Default: None. Note: This argument must be passed as a keyword argument.

  • dtype (DTypeLike, optional) – Data type of the weights and bias. Default: None. Note: This argument must be passed as a keyword argument.

  • weight_attr (ParamAttr|bool|None, optional) –

    The parameter attribute for the learnable scale \(g\). This setting only takes effect when affine is True. - If set to False, no scale parameter will be created. - If set to True or a ParamAttr object, a learnable scale parameter will be created.

    System Message: ERROR/3 (/usr/local/lib/python3.10/site-packages/paddle/nn/layer/norm.py:docstring of paddle.nn.layer.norm.GroupNorm, line 29)

    Unexpected indentation.

    When set to True, it is equivalent to ParamAttr() with default initialization.

    System Message: WARNING/2 (/usr/local/lib/python3.10/site-packages/paddle/nn/layer/norm.py:docstring of paddle.nn.layer.norm.GroupNorm, line 24)

    Block quote ends without a blank line; unexpected unindent.

    • If set to None, a learnable scale parameter will be created and initialized to one.

    System Message: WARNING/2 (/usr/local/lib/python3.10/site-packages/paddle/nn/layer/norm.py:docstring of paddle.nn.layer.norm.GroupNorm, line 25)

    Bullet list ends without a blank line; unexpected unindent.

    Default: None. Note: This argument must be passed as a keyword argument.

  • bias_attr (ParamAttr|bool|None, optional) –

    The parameter attribute for the learnable bias \(b\). This setting only takes effect when affine is True. - If set to False, no bias parameter will be created. - If set to True or a ParamAttr object, a learnable bias parameter will be created.

    System Message: ERROR/3 (/usr/local/lib/python3.10/site-packages/paddle/nn/layer/norm.py:docstring of paddle.nn.layer.norm.GroupNorm, line 38)

    Unexpected indentation.

    When set to True, it is equivalent to ParamAttr() with default initialization.

    System Message: WARNING/2 (/usr/local/lib/python3.10/site-packages/paddle/nn/layer/norm.py:docstring of paddle.nn.layer.norm.GroupNorm, line 33)

    Block quote ends without a blank line; unexpected unindent.

    • If set to None, a learnable bias parameter will be created and initialized to zero.

    System Message: WARNING/2 (/usr/local/lib/python3.10/site-packages/paddle/nn/layer/norm.py:docstring of paddle.nn.layer.norm.GroupNorm, line 34)

    Bullet list ends without a blank line; unexpected unindent.

    Default: None. Note: This argument must be passed as a keyword argument.

  • data_format (str, optional) – Specify the input data format. Support “NCL”, “NCHW”, “NCDHW”, “NLC”, “NHWC” or “NDHWC”. Default: “NCHW”. Note: This argument must be passed as a keyword argument.

  • name (str|None, optional) – Name for the GroupNorm, default is None. For more information, please refer to api_guide_Name. Note: This argument must be passed as a keyword argument.

Shape:

  • x: Tensor with shape: attr:(batch, num_features, *).

  • output: The same shape as input x.

Returns

None

Examples

>>> import paddle
>>> paddle.seed(100)
>>> x = paddle.arange(48, dtype="float32").reshape((2, 6, 2, 2))
>>> group_norm = paddle.nn.GroupNorm(num_channels=6, num_groups=6)
>>> group_norm_out = group_norm(x)

>>> print(group_norm_out)
Tensor(shape=[2, 6, 2, 2], dtype=float32, place=Place(cpu), stop_gradient=False,
[[[[-1.34163547, -0.44721183],
   [ 0.44721183,  1.34163547]],
  [[-1.34163547, -0.44721183],
   [ 0.44721183,  1.34163547]],
  [[-1.34163547, -0.44721183],
   [ 0.44721183,  1.34163547]],
  [[-1.34163547, -0.44721183],
   [ 0.44721183,  1.34163547]],
  [[-1.34163547, -0.44721183],
   [ 0.44721183,  1.34163547]],
  [[-1.34163547, -0.44721183],
   [ 0.44721183,  1.34163547]]],
 [[[-1.34163547, -0.44721183],
   [ 0.44721183,  1.34163547]],
  [[-1.34163547, -0.44721183],
   [ 0.44721183,  1.34163547]],
  [[-1.34163547, -0.44721183],
   [ 0.44721183,  1.34163547]],
  [[-1.34163547, -0.44721183],
   [ 0.44721183,  1.34163547]],
  [[-1.34163547, -0.44721183],
   [ 0.44721183,  1.34163547]],
  [[-1.34163547, -0.44721183],
   [ 0.44721183,  1.34163547]]]])
forward ( input: Tensor ) → Tensor

forward

Defines the computation performed at every call. Should be overridden by all subclasses.

Parameters

  • *inputs (tuple) – unpacked tuple arguments

  • **kwargs (dict) – unpacked dict arguments

extra_repr ( ) → str

extra_repr

Extra representation of this layer, you can have custom implementation of your own layer.