nn.yml

Conv1D:
  desc: "1维卷积"
  paddle:
    api_name: "paddle.nn.Conv1D"
    params:
      data:
        random: true
        type: "Tensor"
        dtype: "float32"
        shape: [2, 3, 4]
        range: [ -1, 1]
      kernel_size:
        value: [3]
      in_channels: 3
      out_channels: 1

Conv1DTranspose:
  desc: "1维反卷积"
  paddle:
    api_name: "paddle.nn.Conv1DTranspose"
    params:
      data:
        random: true
        type: "Tensor"
        dtype: "float32"
        shape: [2, 3, 4]
        range: [ -1, 1]
      kernel_size:
        value: [2]
      in_channels: 3
      out_channels: 1
      stride: 2
      padding: 1
      output_padding: 1
      groups: 1
      dilation: 1

Conv2D:
  desc: "2维卷积"
  paddle:
    api_name: "paddle.nn.Conv2D"
    params:
      data:
        random: true
        type: "Tensor"
        dtype: "float32"
        shape: [2, 3, 4, 4]
        range: [ -1, 1]
      kernel_size:
        value: [3, 3]
      in_channels: 3
      out_channels: 1

Conv2DTranspose:
  desc: "2维反卷积"
  paddle:
    api_name: "paddle.nn.Conv2DTranspose"
    params:
      data:
        random: true
        type: "Tensor"
        dtype: "float32"
        shape: [2, 3, 4, 4]
        range: [ -1, 1]
      kernel_size:
        value: [2, 2]
      in_channels: 3
      out_channels: 1
      stride: 2
      padding: 1
      output_padding: 1
      groups: 1
      dilation: 1

Conv3D:
  desc: "3维卷积"
  paddle:
    api_name: "paddle.nn.Conv3D"
    params:
      data:
        random: true
        type: "Tensor"
        dtype: "float32"
        shape: [2, 3, 4, 4, 4]
        range: [ -1, 1]
      kernel_size:
        value: [3, 3, 3]
      in_channels: 3
      out_channels: 1

Conv3D5:
  desc: "3维卷积第5个case"
  paddle:
    api_name: "paddle.nn.Conv3D"
    params:
      data:
        random: true
        type: "Tensor"
        dtype: "float32"
        shape: [2, 3, 4, 4, 4]
        range: [ -1, 1]
      kernel_size:
        value: [3, 3, 3]
      in_channels: 3
      out_channels: 6
      stride: 2
      padding: 0
      groups: 3

Conv3DTranspose:
  desc: "3维反卷积"
  paddle:
    api_name: "paddle.nn.Conv3DTranspose"
    params:
      data:
        random: true
        type: "Tensor"
        dtype: "float32"
        shape: [2, 3, 4, 4, 4]
        range: [ -1, 1]
      kernel_size:
        value: [2, 2, 2]
      in_channels: 3
      out_channels: 1
      stride: 2
      padding: 1
      output_padding: 1
      groups: 1
      dilation: 1

AdaptiveAvgPool1D:
  desc: "1维自适应池化"
  paddle:
    api_name: "paddle.nn.AdaptiveAvgPool1D"
    params:
      data:
        random: true
        type: "Tensor"
        dtype: "float32"
        shape: [2, 3, 8]
        range: [ -1, 1]
      output_size: 4

AdaptiveAvgPool2D:
  desc: "2维自适应池化"
  paddle:
    api_name: "paddle.nn.AdaptiveAvgPool2D"
    params:
      data:
        random: true
        type: "Tensor"
        dtype: "float32"
        shape: [2, 3, 8, 8]
        range: [ -1, 1]
      output_size: 4
      data_format: 'NCHW'

AdaptiveAvgPool3D:
  desc: "3维自适应池化"
  paddle:
    api_name: "paddle.nn.AdaptiveAvgPool3D"
    params:
      data:
        random: true
        type: "Tensor"
        dtype: "float32"
        shape: [2, 3, 8, 8, 8]
        range: [ -1, 1]
      output_size: 4
      data_format: 'NCDHW'

AdaptiveMaxPool1D:
  desc: "1维自适应最大值池化"
  paddle:
    api_name: "paddle.nn.AdaptiveMaxPool1D"
    params:
      data:
        random: true
        type: "Tensor"
        dtype: "float32"
        shape: [2, 3, 8]
        range: [ -1, 1]
      output_size: 4
      return_mask: False

AdaptiveMaxPool1D1:
  desc: "1维自适应最大值池化，返回mask为True"
  paddle:
    api_name: "paddle.nn.AdaptiveMaxPool1D"
    params:
      data:
        random: true
        type: "Tensor"
        dtype: "float32"
        shape: [2, 3, 8]
        range: [ -1, 1]
      output_size: 4
      return_mask: True

AdaptiveMaxPool2D:
  desc: "2维自适应最大值池化"
  paddle:
    api_name: "paddle.nn.AdaptiveMaxPool2D"
    params:
      data:
        random: true
        type: "Tensor"
        dtype: "float32"
        shape: [2, 3, 8, 8]
        range: [ -1, 1]
      output_size: 4
      return_mask: False

AdaptiveMaxPool3D:
  desc: "3维自适应最大值池化"
  paddle:
    api_name: "paddle.nn.AdaptiveMaxPool3D"
    params:
      data:
        random: true
        type: "Tensor"
        dtype: "float32"
        shape: [2, 3, 8, 8, 8]
        range: [ -1, 1]
      output_size: 4
      return_mask: False

AvgPool1D:
  desc: "1维平均池化"
  paddle:
    api_name: "paddle.nn.AvgPool1D"
    params:
      data:
        random: true
        type: "Tensor"
        dtype: "float32"
        shape: [2, 3, 8]
        range: [ -1, 1]
      kernel_size: 2
      stride: 2
      padding: 0

AvgPool2D:
  desc: "2维平均池化"
  paddle:
    api_name: "paddle.nn.AvgPool2D"
    params:
      data:
        random: true
        type: "Tensor"
        dtype: "float32"
        shape: [2, 3, 8, 8]
        range: [ -1, 1]
      kernel_size: 2
      stride: 2
      padding: 0

AvgPool3D:
  desc: "3维平均池化"
  paddle:
    api_name: "paddle.nn.AvgPool3D"
    params:
      data:
        random: true
        type: "Tensor"
        dtype: "float32"
        shape: [2, 3, 8, 8, 8]
        range: [ -1, 1]
      kernel_size: 2
      stride: 2
      padding: 0

MaxPool1D:
  desc: "1维最大池化"
  paddle:
    api_name: "paddle.nn.MaxPool1D"
    params:
      data:
        random: true
        type: "Tensor"
        dtype: "float32"
        shape: [2, 3, 8]
        range: [ -1, 1]
      kernel_size: 2
      stride: 2
      padding: 0

MaxPool2D:
  desc: "2维最大池化"
  paddle:
    api_name: "paddle.nn.MaxPool2D"
    params:
      data:
        random: true
        type: "Tensor"
        dtype: "float32"
        shape: [2, 3, 8, 8]
        range: [ -1, 1]
      kernel_size: 2
      stride: 2
      padding: 0

MaxPool3D:
  desc: "3维最大池化"
  paddle:
    api_name: "paddle.nn.MaxPool3D"
    params:
      data:
        random: true
        type: "Tensor"
        dtype: "float32"
        shape: [2, 3, 8, 8, 8]
        range: [ -1, 1]
      kernel_size: 2
      stride: 2
      padding: 0

MaxUnPool2D:
  desc: "2维最大逆池化"
  paddle:
    api_name: "paddle.nn.MaxUnPool2D"
    params:
      data:
        random: true
        type: "Tensor"
        dtype: "float32"
        shape: [2, 3, 5, 5]
        range: [ -1, 1]
      indices:
        random: true
        type: "Tensor"
        dtype: "int32"
        shape: [2, 3, 5, 5]
        range: [ 0, 20]
      kernel_size: 2
      stride: 2
      padding: 0

Pad1D:
  desc: "1维pad填充"
  paddle:
    api_name: "paddle.nn.Pad1D"
    params:
      data:
        random: true
        type: "Tensor"
        dtype: "float32"
        shape: [2, 3, 8]
        range: [ -1, 1]
      padding: [1, 2]
      mode: "constant"
      value: 0
      data_format: 'NCL'

Pad2D:
  desc: "2维pad填充"
  paddle:
    api_name: "paddle.nn.Pad2D"
    params:
      data:
        random: true
        type: "Tensor"
        dtype: "float32"
        shape: [2, 3, 8, 8]
        range: [ -1, 1]
      padding: [1, 4, 3, 2]
      mode: "constant"
      value: 3.0
      data_format: 'NCHW'

Pad3D:
  desc: "3维pad填充"
  paddle:
    api_name: "paddle.nn.Pad3D"
    params:
      data:
        random: true
        type: "Tensor"
        dtype: "float32"
        shape: [2, 3, 8, 8, 8]
        range: [ -1, 1]
      padding: [1, 4, 3, 2, 1, 1]
      mode: "constant"
      value: 2.1
      data_format: 'NCDHW'

CELU:
  desc: "CELU激活层"
  paddle:
    api_name: "paddle.nn.CELU"
    params:
      data:
        random: true
        type: "Tensor"
        dtype: "float32"
        shape: [2, 3, 4, 4, 4]
        range: [ -1, 1]
      alpha: 0.56

ELU:
  desc: "ELU激活层"
  paddle:
    api_name: "paddle.nn.ELU"
    params:
      data:
        random: true
        type: "Tensor"
        dtype: "float32"
        shape: [2, 3, 4, 4, 4]
        range: [ -1, 1]
      alpha: 0.56

GELU:
  desc: "GELU激活层"
  paddle:
    api_name: "paddle.nn.GELU"
    params:
      data:
        random: true
        type: "Tensor"
        dtype: "float32"
        shape: [2, 3, 4, 4, 4]
        range: [ -1, 1]
      approximate: True

Hardshrink:
  desc: "Hardshrink激活层"
  paddle:
    api_name: "paddle.nn.Hardshrink"
    params:
      data:
        random: true
        type: "Tensor"
        dtype: "float32"
        shape: [2, 3, 4, 4, 4]
        range: [ -1, 1]
      threshold: 0.4

Hardsigmoid:
  desc: "Hardsigmoid激活层"
  paddle:
    api_name: "paddle.nn.Hardsigmoid"
    params:
      data:
        random: true
        type: "Tensor"
        dtype: "float32"
        shape: [2, 3, 4, 4, 4]
        range: [ -1, 1]

Hardswish:
  desc: "Hardswish激活层"
  paddle:
    api_name: "paddle.nn.Hardswish"
    params:
      data:
        random: true
        type: "Tensor"
        dtype: "float32"
        shape: [2, 3, 4, 4, 4]
        range: [ -1, 1]

Hardtanh:
  desc: "Hardtanh激活层"
  paddle:
    api_name: "paddle.nn.Hardtanh"
    params:
      data:
        random: true
        type: "Tensor"
        dtype: "float32"
        shape: [2, 3, 4, 4, 4]
        range: [ -3, 3]
      min: -1.5
      max: 2.0

LeakyReLU:
  desc: "LeakyReLU激活层"
  paddle:
    api_name: "paddle.nn.LeakyReLU"
    params:
      data:
        random: true
        type: "Tensor"
        dtype: "float32"
        shape: [2, 3, 4, 4, 4]
        range: [ -1, 1]
      negative_slope: 0.015

LogSigmoid:
  desc: "LogSigmoid激活层"
  paddle:
    api_name: "paddle.nn.LogSigmoid"
    params:
      data:
        random: true
        type: "Tensor"
        dtype: "float32"
        shape: [2, 3, 4, 4, 4]
        range: [ -1, 1]

LogSoftmax:
  desc: "LogSoftmax激活层"
  paddle:
    api_name: "paddle.nn.LogSoftmax"
    params:
      data:
        random: true
        type: "Tensor"
        dtype: "float32"
        shape: [2, 3, 4, 4, 4]
        range: [ -1, 1]
      axis: 1

Maxout:
  desc: "Maxout激活层"
  paddle:
    api_name: "paddle.nn.Maxout"
    params:
      data:
        random: true
        type: "Tensor"
        dtype: "float32"
        shape: [2, 6, 4, 4]
        range: [ -1, 1]
      groups: 2
      axis: 1

#PReLU:
#  desc: "PReLU激活层"
#  paddle:
#    api_name: "paddle.nn.PReLU"
#    params:
#      data:
#        random: true
#        type: "Tensor"
#        dtype: "float32"
#        shape: [2, 6, 4, 4]
#        range: [ -1, 1]
#      num_parameters: 0.5
#      init: 0.35
#      weight_attr: None
#      data_format: "NCHW"

ReLU:
  desc: "ReLU激活层"
  paddle:
    api_name: "paddle.nn.ReLU"
    params:
      data:
        random: true
        type: "Tensor"
        dtype: "float32"
        shape: [2, 3, 4, 4]
        range: [ -1, 1]

ReLU6:
  desc: "ReLU激活层"
  paddle:
    api_name: "paddle.nn.ReLU6"
    params:
      data:
        random: true
        type: "Tensor"
        dtype: "float32"
        shape: [2, 3, 4, 4]
        range: [ -1, 10]

SELU:
  desc: "SELU激活层"
  paddle:
    api_name: "paddle.nn.SELU"
    params:
      data:
        random: true
        type: "Tensor"
        dtype: "float32"
        shape: [2, 3, 4, 4]
        range: [ -1, 1]
      scale: 1.05173
      alpha: 1.67325

Sigmoid:
  desc: "Sigmoid激活层"
  paddle:
    api_name: "paddle.nn.Sigmoid"
    params:
      data:
        random: true
        type: "Tensor"
        dtype: "float32"
        shape: [2, 3, 4, 4]
        range: [ -1, 1]

Silu:
  desc: "Silu激活层"
  paddle:
    api_name: "paddle.nn.Silu"
    params:
      data:
        random: true
        type: "Tensor"
        dtype: "float32"
        shape: [2, 3, 4, 4]
        range: [ -1, 1]

Softmax:
  desc: "Softmax激活层"
  paddle:
    api_name: "paddle.nn.Softmax"
    params:
      data:
        random: true
        type: "Tensor"
        dtype: "float32"
        shape: [2, 3, 4, 4]
        range: [ -1, 1]
      axis: 1

Softplus:
  desc: "Softplus激活层"
  paddle:
    api_name: "paddle.nn.Softplus"
    params:
      data:
        random: true
        type: "Tensor"
        dtype: "float32"
        shape: [2, 3, 4, 4]
        range: [ -1, 1]
      beta: 1
      threshold: 20

Softshrink:
  desc: "Softshrink激活层"
  paddle:
    api_name: "paddle.nn.Softshrink"
    params:
      data:
        random: true
        type: "Tensor"
        dtype: "float32"
        shape: [2, 3, 4, 4]
        range: [ -1, 1]
      threshold: 0.2

Softsign:
  desc: "Softsign激活层"
  paddle:
    api_name: "paddle.nn.Softsign"
    params:
      data:
        random: true
        type: "Tensor"
        dtype: "float32"
        shape: [2, 3, 4, 4]
        range: [ -1, 1]

Swish:
  desc: "Swish激活层"
  paddle:
    api_name: "paddle.nn.Swish"
    params:
      data:
        random: true
        type: "Tensor"
        dtype: "float32"
        shape: [2, 3, 4, 4]
        range: [ -4, 4]

Tanh:
  desc: "Tanh激活层"
  paddle:
    api_name: "paddle.nn.Tanh"
    params:
      data:
        random: true
        type: "Tensor"
        dtype: "float32"
        shape: [2, 3, 4, 4]
        range: [ -1, 1]

Tanhshrink:
  desc: "Tanhshrink激活层"
  paddle:
    api_name: "paddle.nn.Tanhshrink"
    params:
      data:
        random: true
        type: "Tensor"
        dtype: "float32"
        shape: [2, 3, 4, 4]
        range: [ -1, 1]

ThresholdedReLU:
  desc: "ThresholdedReLU激活层"
  paddle:
    api_name: "paddle.nn.ThresholdedReLU"
    params:
      data:
        random: true
        type: "Tensor"
        dtype: "float32"
        shape: [2, 3, 4, 4]
        range: [ -1, 2]
      threshold: 1.25

BatchNorm:
  desc: "BN批归一化"
  paddle:
    api_name: "paddle.nn.BatchNorm"
    params:
      data:
        random: true
        type: "Tensor"
        dtype: "float32"
        shape: [2, 3, 2, 2]
        range: [ -1, 1]
      num_channels: 3
      is_test: False
      momentum: 0.9
      epsilon: 0.00001
      dtype: 'float32'
      data_layout: 'NCHW'
      in_place: False
      do_model_average_for_mean_and_var: False
      use_global_stats: False
      trainable_statistics: False

BatchNorm1D:
  desc: "1维BN批归一化"
  paddle:
    api_name: "paddle.nn.BatchNorm1D"
    params:
      data:
        random: true
        type: "Tensor"
        dtype: "float32"
        shape: [2, 3, 4]
        range: [ -1, 1]
      num_features: 3
      momentum: 0.9
      epsilon: 0.00001
      data_format: 'NCL'

BatchNorm2D:
  desc: "2维BN批归一化"
  paddle:
    api_name: "paddle.nn.BatchNorm2D"
    params:
      data:
        random: true
        type: "Tensor"
        dtype: "float32"
        shape: [2, 3, 4, 4]
        range: [ -1, 1]
      num_features: 3
      momentum: 0.9
      epsilon: 0.00001
      data_format: 'NCHW'

BatchNorm3D:
  desc: "3维BN批归一化"
  paddle:
    api_name: "paddle.nn.BatchNorm3D"
    params:
      data:
        random: true
        type: "Tensor"
        dtype: "float32"
        shape: [2, 3, 4, 4, 4]
        range: [ -1, 1]
      num_features: 3
      momentum: 0.9
      epsilon: 0.00001
      data_format: 'NCDHW'

GroupNorm:
  desc: "分组归一化"
  paddle:
    api_name: "paddle.nn.GroupNorm"
    params:
      data:
        random: true
        type: "Tensor"
        dtype: "float32"
        shape: [2, 6, 2, 2]
        range: [ -1, 1]
      num_groups: 6
      num_channels: 6
      epsilon: 0.00001
      data_format: 'NCHW'

InstanceNorm1D:
  desc: "1维实例归一化"
  paddle:
    api_name: "paddle.nn.InstanceNorm1D"
    params:
      data:
        random: true
        type: "Tensor"
        dtype: "float32"
        shape: [2, 2, 3]
        range: [ -1, 1]
      num_features: 2
      epsilon: 0.00001
      momentum: 0.9
      data_format: 'NCL'

InstanceNorm2D:
  desc: "2维实例归一化"
  paddle:
    api_name: "paddle.nn.InstanceNorm2D"
    params:
      data:
        random: true
        type: "Tensor"
        dtype: "float32"
        shape: [2, 2, 3, 3]
        range: [ -1, 1]
      num_features: 2
      epsilon: 0.00001
      momentum: 0.9
      data_format: 'NCHW'

InstanceNorm3D:
  desc: "3维实例归一化"
  paddle:
    api_name: "paddle.nn.InstanceNorm3D"
    params:
      data:
        random: true
        type: "Tensor"
        dtype: "float32"
        shape: [2, 2, 3, 3, 3]
        range: [ -1, 1]
      num_features: 2
      epsilon: 0.00001
      momentum: 0.9
      data_format: 'NCDHW'

LayerNorm:
  desc: "层归一化"
  paddle:
    api_name: "paddle.nn.LayerNorm"
    params:
      data:
        random: true
        type: "Tensor"
        dtype: "float32"
        shape: [2, 2, 2, 3]
        range: [ -1, 1]
      normalized_shape: [2, 2, 3]
      epsilon: 0.00001

LocalResponseNorm:
  desc: "层归一化"
  paddle:
    api_name: "paddle.nn.LocalResponseNorm"
    params:
      data:
        random: true
        type: "Tensor"
        dtype: "float32"
        shape: [3, 3, 8, 8]
        range: [ -1, 1]
      size: 5
      alpha: 0.0001
      beta: 0.665
      k: 0.9
      data_format: 'NCHW'

SpectralNorm:
  desc: "谱归一化"
  paddle:
    api_name: "paddle.nn.SpectralNorm"
    params:
      data:
        random: true
        type: "Tensor"
        dtype: "float32"
        shape: [2, 3, 8, 8]
        range: [ -1, 1]
      weight_shape: [2, 3, 8, 8]
      dim: 1
      power_iters: 1
      eps: 0.000000000001
      dtype: "float32"

#SyncBatchNorm:
#  desc: "跨卡GPU同步的批归一化"
#  paddle:
#    api_name: "paddle.nn.SyncBatchNorm"
#    params:
#      data:
#        random: true
#        type: "Tensor"
#        dtype: "float32"
#        shape: [2, 3, 8, 8]
#        range: [ -1, 1]
#      num_features: 3
#      epsilon: 0.00001
#      momentum: 0.9
#      data_format: 'NCHW'

#BiRNN:
#  desc: "谱归一化"
#  paddle:
#    api_name: "paddle.nn.BiRNN"
#    params:
#      data:
#        random: true
#        type: "Tensor"
#        dtype: "float32"
#        shape: [2, 3, 8, 8]
#        range: [ -1, 1]

GRU:
  desc: "门控循环单元网络"
  paddle:
    api_name: "paddle.nn.GRU"
    params:
      data0:
        random: true
        type: "Tensor"
        dtype: "float32"
        shape: [4, 23, 16]
        range: [ -1, 1]
      data1:
        random: true
        type: "Tensor"
        dtype: "float32"
        shape: [2, 4, 32]
        range: [ -1, 1]
      input_size: 16
      hidden_size: 32
      num_layers: 2
      direction: 'forward'
      dropout: 0.
      time_major: False

GRUCell:
  desc: "门控循环单元Cell"
  paddle:
    api_name: "paddle.nn.GRUCell"
    params:
      data0:
        random: true
        type: "Tensor"
        dtype: "float32"
        shape: [4, 16]
        range: [ -1, 1]
      data1:
        random: true
        type: "Tensor"
        dtype: "float32"
        shape: [4, 32]
        range: [ -1, 1]
      input_size: 16
      hidden_size: 32