def conv2d_transpose(
data,
weight,
input_zero_point,
kernel_zero_point,
input_scale,
kernel_scale,
strides=(1, 1),
padding=(0, 0),
dilation=(1, 1),
groups=1,
channels=None,
kernel_size=None,
data_layout="NCHW",
kernel_layout="IOHW",
out_layout="",
output_padding=(0, 0),
out_dtype="int32",
):
"""This operator deconvolves quantized data with quantized kernel. The scale of
the output quantized tensor is the product of the kernel_scale and
input_scale of the input quantized tensors. The zero point of the output
quantized tensor is 0. By default, the dtype of output is int32. Please also
refer to Requantize operator to understand how to scale back the int32
output to (u)int8.
Parameters
----------
data : tvm.relay.Expr
The input data to the operator.
weight : tvm.relay.Expr
The weight expressions.
input_zero_point: tvm.relay.Expr
The zero point of the data distribution.
kernel_zero_point: tvm.relay.Expr
The zero point of the quantized_kernel distribution.
input_scale: tvm.relay.Expr
The scale for the input tensor. The scale for the input tensor is
stored purely for convenience here. See more commentary below.
kernel_scale: tvm.relay.Expr
The scale for the weight tensor. The scale for the weight tensor is
stored for access to this during relay. This information is not
needed in the pass pipeline after qnn.conv2d_transpose is lowered to the
sequence of steps as in nn.conv2d_transpose. See also input_scale in Requantize.
strides : Tuple[int], optional
The strides of convolution.
padding : Tuple[int], optional
The padding of convolution.
dilation : Tuple[int], optional
Specifies the dilation rate to be used for dilated convolution.
channels : int, optional
Number of output channels of this convolution.
kernel_size : tuple of int, optional
The spatial dimensions of the convolution kernel.
groups : int, optional
Number of groups for grouped convolution.
data_layout : str, optional
Layout of the input.
kernel_layout : str, optional
Layout of the weight.
out_layout : Optional[str]
Layout of the output, by default, out_layout is the same as data_layout
output_padding : Tuple[int], optional
Used to identify the padding within the output shape
(only used in training, where transpose_conv represents the gradient of a convolution )
out_dtype : str, optional
Specifies the output data type for mixed precision conv2d.
Returns
-------
result : tvm.relay.Expr
The computed result.
"""
# convert 2-way padding to 4-way padding
padding = get_pad_tuple2d(padding)
return _make.conv2d_transpose(
data,
weight,
input_zero_point,
kernel_zero_point,
input_scale,
kernel_scale,
strides,
padding,
dilation,
groups,
channels,
kernel_size,
data_layout,
kernel_layout,
out_layout,
output_padding,
out_dtype,
)
def conv2d(
data,
kernel,
input_zero_point,
kernel_zero_point,
input_scale,
kernel_scale,
kernel_size,
channels,
strides=(1, 1),
padding=(0, 0),
dilation=(1, 1),
groups=1,
data_layout="NCHW",
kernel_layout="OIHW",
out_layout="",
out_dtype="int32",
):
r"""Quantized 2D convolution.
This operator convolves quantized data with quantized kernel.
If doing Per-channel quantization, qnn expects the kernel_zero_scale
and optionally the kernel_zero_point will be 1-D vectors instead of scalars.
The scale of the output quantized tensor is the product of the kernel_scale and
input_scale of the input quantized tensors. The zero point of the output
quantized tensor is 0. By default, the dtype of output is int32. Please also
refer to Requantize operator to understand how to scale back the int32
output to (u)int8.
Parameters
----------
data : tvm.relay.Expr
The input data to the operator.
kernel : tvm.relay.Expr
The kernel expressions.
input_zero_point: tvm.relay.Expr
The zero point of the data distribution.
kernel_zero_point: tvm.relay.Expr
The zero point of the quantized_kernel distribution.
input_scale: tvm.relay.Expr
The scale for the input tensor. The scale for the input tensor is
stored purely for convenience here. See more commentary below.
kernel_scale: tvm.relay.Expr
The scale for the weight tensor. The scale for the weight tensor is
stored for access to this during relay. This information is not
needed in the pass pipeline after qnn.conv2d is lowered to the
sequence of steps as in nn.conv2d. See also input_scale in Requantize.
kernel_size : tuple of int
The spatial width and height of the convolution kernel.
channels : int
Number of output channels of this convolution.
strides : tuple of int, optional
The strides of convolution.
padding : tuple of int, optional
The padding of convolution on both sides of inputs before convolution.
dilation : tuple of int, optional
Specifies the dilation rate to be used for dilated convolution.
groups : int, optional
Number of groups for grouped convolution.
data_layout : str, optional
Layout of the input.
kernel_layout : str, optional
Layout of the kernel.
out_layout : str, optional
Layout of the output, by default, out_layout is the same as data_layout
out_dtype : str, optional
Specifies the output data type for mixed precision conv2d.
Returns
-------
result : tvm.relay.Expr
The computed result.
"""
# TODO enforce 4-way padding in topi/nn/conv2d after #4644 merged
# convert 2-way padding to 4-way padding
padding = get_pad_tuple2d(padding)
return _make.conv2d(
data,
kernel,
input_zero_point,
kernel_zero_point,
input_scale,
kernel_scale,
strides,
padding,
dilation,
groups,
channels,
kernel_size,
data_layout,
kernel_layout,
out_layout,
out_dtype,
)