ConvTranspose#

Versioned name: ConvTranspose-1

Category: ConvTranspose

Short description: Applies a transposed convolution operator over an input. It is also known as a Fractionally-strided convolution, Up convolution or a Deconvolution (although it is not an actual deconvolution operation).

Detailed description:

ConvTranspose spaces out and adds the input with zero paddings based on pads_begin and pads_end and applies the filter over the data to project the input to a higher-dimensional space. The computation of ConvTranspose is the same as computing the gradients of input of Convolution operation.

If the pads parameter is provided, the shape of the output is calculated via the following equation (similar to onnx definition):

output_shape[i] = stride[i] * (input_shape[i] - 1) + output_padding[i] + ((filter_shape[i] - 1) * dilations[i] + 1) - pads_begin[i] - pads_end[i]

If output_shape is explicitly specified in which case pads_begin and pads_end are ignored, pads values are automatically generated using these equations:

total_padding[i] = stride[i] * (input_shape[i] - 1) + output_padding[i] + ((filter_shape[i] - 1) * dilations[i] + 1) - output_shape[i]
if (auto_pad == SAME)
  total_padding[i] = stride[i] * (input_shape[i] - 1) + ((filter_shape[i] - 1) * dilations[i] + 1) - input_shape[i] * stride[i]
                   = (filter_shape[i] - 1) * dilations[i] + 1 - stride[i]
if (auto_pad == SAME_UPPER):
  pads_begin[i] = total_padding[i] / 2
  pads_end[i] = total_padding[i] - pads_begin[i]
else:
  pads_end[i] = total_padding[i] / 2
  pads_begin[i] = total_padding[i] - pads_end[i]
if (auto_pad == VALID)
  pads_begin[i] = pads_end[i] = 0

Attributes

  • strides

    • Description: strides controls the stride along each spatial axis.

    • Range of values: positive s64 values.

    • Type: s64[]

    • Required: yes

  • pads_begin

    • Description: pads_begin controls the amount of implicit zero padding of each spatial axis.

    • Range of values: non-negative s64 values.

    • Type: s64[]

    • Required: yes

    • Note: the attribute is ignored when auto_pad attribute is specified.

  • pads_end

    • Description: pads_end controls the amount of implicit zero padding of each spatial axis.

    • Range of values: non-negative s64 values.

    • Type: s64[]

    • Required: yes

    • Note: the attribute is ignored when auto_pad attribute is specified.

  • dilations

    • Description: dilations controls the spacing between the kernel points.

    • Range of values: positive s64 values.

    • Type: s64[]

    • Required: yes

  • auto_pad

    • Description: auto_pad describes how the padding is calculated.

      • none (not specified): use explicit padding values from pads_begin and pads_end.

      • same_upper (same_lower) the input is padded to match the output size. In case of odd padding value an extra padding is added at the end (at the beginning).

      • valid - do not use padding.

    • Type: string

    • Default value: none

    • Required: no

    • Note: pads_begin and pads_end attributes are ignored when auto_pad is specified.

  • output_padding

    • Description: output_padding adds additional amount of padding per each spatial axis in the output tensor. It unlocks more elements in the output allowing them to be computed. Elements are added at the higher coordinate indices for the spatial dimensions. Number of elements in output_padding list matches the number of spatial dimensions in data and output tensors.

    • Range of values: non-negative s64 values.

    • Type: s64[]

    • Default value: all zeros

    • Required: no

  • groups

    • Description: groups denotes the number of groups input channels and output channels are divided into. In_channels and out_channels must both be divisible by groups.

    • Range of values: a positive s64 value.

    • Type: s64

    • Default value: 1

    • Required: no

  • data_format

    • Description: data_format denotes the data format of the input and output data.

    • Range of values: NXC or NCX (X means HW for 2D ConvTranspose, DHW for 3D ConvTranspose)

    • Type: string

    • Default value: NXC

    • Required: no

  • filter_format

    • Description: filter_format denotes the data format of the filter.

    • Range of values: XIO or OIX (X means HW for 2D ConvTranspose, DHW for 3D ConvTranspose)

    • Type: string

    • Default value: XIO

    • Required: no

Inputs:

  • 1: input - input tensor. The format is specified by data_format. Required.

    • Type: T

  • 2: filter - filter tensor. The format is specified by filter_format attribute. The shape of filter is ::math::(out_channels / groups, in_channels, spatial_shape) for OIX format or ::math::(spatial_shape, in_channels, out_channels / groups) for XIO format. ::math::in_channels and ::math::out_channels must both be divisible by groups. Required.

    • Type: T

  • 3: bias - a 1-D tensor adds to channel dimension of output. Optional.

    • Type: T

Outputs:

  • 1: output - the output tensor of the same rank as the input tensor.

    • Type: T

Types:

  • T: f32, f16, bf16.

  • Note: Inputs and outputs have the same data type denoted by T. For example, if input is f32 tensor, then all other tensors have f32 data type.