Skip to Content
DocumentationFunctionalSpacial Opsconv3d_op

conv3d_op

View the code on GitHub 

Structs

Struct: Conv3d

Namespace for 3D convolution operations.

Fields

Methods

compute_shape(mut curr: ArrayShape, args: List[ArrayShape])
Computes the shape of an array after a 3-dimensional convolution operation.
Args

  • curr: ArrayShape The ArrayShape to store the result of the computation.

  • args: List[ArrayShape] The input ArrayShape, and the convolution parameters encoded in an ArrayShape.

__call__(mut curr: Array, args: List[Array])
more details
Args

  • curr: Array

  • args: List[Array]

vjp(primals: List[Array], grad: Array, out: Array) -> List[Array]
more details
Args

  • primals: List[Array]

  • grad: Array

  • out: Array

Returns
  • List[Array]
jvp(primals: List[Array], tangents: List[Array]) -> Array
more details
Args

  • primals: List[Array]

  • tangents: List[Array]

Returns
  • Array
fwd(arg0: Array, kernel: Array, bias: Array, stride: Tuple[Int, Int, Int] = Tuple(VariadicPack(<store_to_mem({1}), store_to_mem({1}), store_to_mem({1})>, True)), padding: Tuple[Int, Int, Int] = Tuple(VariadicPack(<store_to_mem({0}), store_to_mem({0}), store_to_mem({0})>, True)), dilation: Tuple[Int, Int, Int] = Tuple(VariadicPack(<store_to_mem({1}), store_to_mem({1}), store_to_mem({1})>, True)), groups: Int = 1) -> Array
more details
Args

  • arg0: Array

  • kernel: Array

  • bias: Array

  • stride: Tuple[Int, Int, Int] (default: Tuple(VariadicPack(<store_to_mem({1}), store_to_mem({1}), store_to_mem({1})>, True)))

  • padding: Tuple[Int, Int, Int] (default: Tuple(VariadicPack(<store_to_mem({0}), store_to_mem({0}), store_to_mem({0})>, True)))

  • dilation: Tuple[Int, Int, Int] (default: Tuple(VariadicPack(<store_to_mem({1}), store_to_mem({1}), store_to_mem({1})>, True)))

  • groups: Int (default: 1)

Returns
  • Array

Functions

conv3d

conv3d(arg0: Array, kernel: Array, bias: Array, stride: Tuple[Int, Int, Int] = Tuple(VariadicPack(<store_to_mem({1}), store_to_mem({1}), store_to_mem({1})>, True)), padding: Tuple[Int, Int, Int] = Tuple(VariadicPack(<store_to_mem({0}), store_to_mem({0}), store_to_mem({0})>, True)), dilation: Tuple[Int, Int, Int] = Tuple(VariadicPack(<store_to_mem({1}), store_to_mem({1}), store_to_mem({1})>, True)), groups: Int = 1) -> Array
Applies a 3D convolution operation over an input array.
Args

  • arg0: Array The input array.

  • kernel: Array The convolution kernel.

  • bias: Array The bias tensor.

  • stride: Tuple[Int, Int, Int] (default: Tuple(VariadicPack(<store_to_mem({1}), store_to_mem({1}), store_to_mem({1})>, True))) The stride of the convolution operation. Defaults to (1, 1, 1).

  • padding: Tuple[Int, Int, Int] (default: Tuple(VariadicPack(<store_to_mem({0}), store_to_mem({0}), store_to_mem({0})>, True))) The padding to apply to the input. Defaults to (0, 0, 0).

  • dilation: Tuple[Int, Int, Int] (default: Tuple(VariadicPack(<store_to_mem({1}), store_to_mem({1}), store_to_mem({1})>, True))) The dilation to apply to the input. Defaults to (1, 1, 1).

  • groups: Int (default: 1) The number of groups to split the input and output channels into. Defaults to 1.

Returns
  • Array - Array: The output array.
Last updated on
conv2d_opconv_transpose1d