conv¶

dragon.nn.conv(
  inputs,
  kernel_shape=(3, 3),
  strides=1,
  pads=0,
  dilations=1,
  group=1,
  padding='VALID',
  data_format='NCHW',
  **kwargs
)[source]¶

Apply the n-dimension convolution.

• If data_format is 'NCHW', excepts input shape \((N, C_{\text{in}}, D1, D2, ...)\), weight shape \((C_{\text{out}}, C_{\text{in}}, D1_{\text{k}}, D2_{\text{k}}, ...)\), and output shape is \((N, C_{\text{out}}, D1_{\text{out}}, D2_{\text{out}}, ...)\).
• If data_format is 'NHWC', excepts input shape \((N, D1, D2, ..., C_{\text{in}})\), weight shape \((C_{\text{out}}, D1_{\text{k}}, D2_{\text{k}}, ..., C_{\text{in}})\), and output shape is \((N, D1_{\text{out}}, D2_{\text{out}}, ..., C_{\text{out}})\).
• If padding is 'VALID', pads controls the explicit padding size. Otherwise, size are computed automatically use the given method.

Examples:

for i in range(3):
    ndim = i + 1
    x = dragon.ones((1, 2) + (2,) * ndim)
    w = dragon.ones((3, 2) + (1,) * ndim)
    y = dragon.nn.conv([x, w], kernel_shape=(1,) * ndim)
    assert y.shape == (1, 3) + (2,) * ndim

Parameters:

• inputs (Sequence[dragon.Tensor]) – The tensor x, weight and bias.
• kernel_shape (Sequence[int], optional, default=(3, 3)) – The shape of convolution window.
• strides (Union[int, Sequence[int]], optional, default=1) – The stride of convolution window.
• pads (Union[int, Sequence[int]], optional, default=0) – The zero padding size.
• dilations (Union[int, Sequence[int]], optional, default=1) – The rate of dilated convolution.
• group (int, optional, default=1) – The number of groups to split channels into.
• padding (str, optional, default='VALID') – 'VALID', 'SAME', 'SAME_UPPER' or 'SAME_LOWER'.
• data_format (str, optional, default='NCHW') – 'NCHW' or 'NHWC'.

Returns:

dragon.Tensor – The output tensor.