UFF Operators¶

All shapes include batch dimension, unless otherwise specified.

Input¶

An input to the network. Expects a CHW shape.

Supported Datatypes¶

float32, float16, int32, int8

Identity¶

Identity layer.

Inputs¶

Input0 [Tensor or Constant]: Input0 to the identity. Must be float32.

Supported Datatypes¶

float32

Const¶

A constant in the network. Should not include batch dimension.

Supported Datatypes¶

float32, float16, int32, int8

Conv¶

A convolution operation.

Inputs¶

Input0 [Tensor]: The input to the convolution. Must be 4 dimensional. Automatically transposed to NCHW.
Kernel [Constant]: The kernel weights for the convolution. Must be 4 dimensional. Automatically transposed to NCHW.

Attributes¶

dilation [List[int]]: The HW dilations.
strides [List[int]]: The HW strides.
padding [List[int]]: The HW padding. Asymmetric padding is unsupported.

Supported Datatypes¶

float32, float16, int32, int8

ConvTranspose¶

A transposed convolution, also known as deconvolution.

Inputs¶

Input0 [Tensor]: The input to the transposed convolution. Must be 4 dimensional. Automatically transposed to NCHW.
Kernel [Constant]: The kernel weights for the transposed convolution. Must be 4 dimensional. Automatically transposed to NCHW.
Shape [Constant]: The HW dimensions of the output.

Attributes¶

strides [List[int]]: The HW strides.
padding [List[int]]: The HW padding. Asymmetric padding is unsupported.

Supported Datatypes¶

float32, float16, int32, int8

Pool¶

A pooling layer.

Inputs¶

Input0 [Tensor]: The input to the pooling layer. Must be 4 dimensional.

Attributes¶

func [Enum[max, avg]]: The type of pooling to apply.
kernel [List[int]]: The HW shape of the kernel.
strides [List[int]]: The HW strides.
padding [List[int]]: The HW padding.

Supported Datatypes¶

float32, float16, int32, int8

FullyConnected¶

A fully connected layer.

Inputs¶

Input0 [Tensor]: The input to the fully connected layer. Must be at least 4 dimensional. Automatically transposed to -NC-.
Weights [Constant]: The weights for the fully connected layer. Must be 3 dimensional. Automatically transposed to CHW, where C is the number of output channels.

Supported Datatypes¶

float32, float16, int32, int8

LRN¶

An LRN layer.

Inputs¶

Input0 [Tensor]: The input to the LRN. Must be at least 4 dimensional.

Attributes¶

window_size [int]: The window size.
alpha [double]: The LRN alpha value.
beta [double]: The LRN beta value.
k [double]: The LRN k value.

Supported Datatypes¶

float32, float16, int32, int8

Binary¶

A binary layer.

Inputs¶

Input0 [Tensor or Constant]: The first input to the binary layer.
Input1 [Tensor or Constant]: The second input to the binary layer.

If either input is a constant, then at least one of the inputs must be 4 dimensional.

Attributes¶

func [Enum[min, max, mul, sub, div, add, pow]]: The type of operation to perform.

Supported Datatypes¶

float32, float16, int32, int8

Unary¶

A unary layer.

Inputs¶

Input0 [Tensor or Constant]: The input to the unary layer.

The output of a unary layer with a Constant input is treated as a Constant, and therefore will not work with layers expecting a Tensor input.

Attributes¶

func [Enum[neg, exp, log, abs, sqrt, rsqrt, square, sin, cos, tan, sinh, cosh, asin, acos, atan, asinh, acosh, atanh, ceil, floor]]: The type of operation to perform.

Supported Datatypes¶

float32, float16, int32, int8

Reshape¶

A reshape layer. NOTE: this layer destroys order information. Therefore, subsequent layers will cease to automatically transpose their inputs to the correct format.

Inputs¶

Input0 [Tensor or Constant]: The input to the reshape layer.
Shape [Constant]: The desired shape. If the shape has fewer than 3 non-batch dimensions, 1s are inserted in the least significant dimensions. For example, if the shape specified is [1, 300, 5], it will be treated as [1, 300, 5, 1] instead. - -1 specifies that the dimension should be automatically deduced - this can only be used at most once in any given shape. - 0 specifies that the dimension should be copied from the input.

The output of a reshape layer with a Constant input is treated as a Constant, and therefore will not work with layers expecting a Tensor input.

Supported Datatypes¶

float32, float16, int32, int8

ExpandDims¶

An unsqueeze layer. NOTE: this layer destroys order information. Therefore, subsequent layers will cease to automatically transpose their inputs to the correct format.

Inputs¶

Input0 [Tensor]: The input to unsqueeze.

If the resulting output has fewer than 3 non-batch dimensions, it is unsqueezed further with additional 1s inserted in the least significant dimensions. For example, with an input of shape [1, 300], and axis of 1, the shape of the output will be [1, 300, 1, 1] rather than [1, 300, 1].

Attributes¶

axis [int]: The axis on which to unsqueeze, excluding batch dimension. For example, unsqueezing an NCHW tensor with an axis of 0 will result in a N1CHW tensor.

Supported Datatypes¶

float32, float16, int32, int8

ArgMax¶

An argmax layer.

Inputs¶

Input0 [Tensor]: The input to the argmax layer.

Attributes¶

axis [int]: The axis on which to perform the argmax, with 0 corresponding to the batch dimension. The specified dimension is removed. For example, performing argmax on an input of shape [1, 300, 150], with an axis of 1 would result in an output of shape [1, 150]. NOTE: argmax on the batch dimension is not supported.

Supported Datatypes¶

float32, float16, int32, int8

ArgMin¶

An argmin layer.

Inputs¶

Input0 [Tensor]: The input to the argmin layer.

Attributes¶

axis [int]: The axis on which to perform the argmin, with 0 corresponding to the batch dimension. The specified dimension is removed. For example, performing argmin on an input of shape [1, 300, 150], with an axis of 1 would result in an output of shape [1, 150]. NOTE: argmin on the batch dimension is not supported.

Supported Datatypes¶

float32, float16, int32, int8

Transpose¶

A transpose layer. A no-op in the network, this layer only modifies the UFF parser’s internal order information. Therefore, when followed by any layer that destroys order information, the transpose will not be performed.

Inputs¶

Input0 [Tensor]: The input to the transpose layer.

Attributes¶

permutation [int]: The permutation to perform. Must be 4 dimensional.

Supported Datatypes¶

float32, float16, int32, int8

Reduce¶

A reduce layer. NOTE: this layer destroys order information. Therefore, subsequent layers will cease to automatically transpose their inputs to the correct format.

Inputs¶

Input0 [Tensor or Constant]: The input to the reduce layer.

The output of a reduce layer with a Constant input is treated as a Constant, and therefore will not work with layers expecting a Tensor input.

Attributes¶

func [Enum[sum, prod, max, min, mean]]: The reduction operation to perform.
axes [List[int]]: The axes on which to reduce, with 0 corresponding to the batch dimension. Reduction on the batch dimension is unsupported.
keepdims [bool]: Whether to keep the dimensions which were reduced. NOTE: The UFF parser ignored this value, and always keeps dimensions.

Supported Datatypes¶

float32, float16, int32, int8

Concat¶

A concatenation layer.

Inputs¶

Inputs (variadic) [List[Tensor]]: The tensors to concatenate. All inputs are transposed to the same format as the first input. Inputs must be at least 4 dimensional.

Attributes¶

axis [int]: The axis on which to perform the concatenation, with 0 corresponding to the batch dimension. Concatenating on the batch dimension is unsupported.

Supported Datatypes¶

float32, float16, int32, int8

MarkOutput¶

The output of the network.

Inputs¶

Inputs (variadic) [List[Tensor]]: The inputs to this layer. Automatically transposed to the same order as the outputs of the original TensorFlow network.

Supported Datatypes¶

float32, float16, int32, int8

Activation¶

An activation layer.

Inputs¶

Input0 [Tensor]: The input to the activation.

Attributes¶

func [Enum[relu, relu6, sigmoid, tanh, elu, selu, softsign, softplus]]: The operation to perform.

Supported Datatypes¶

float32, float16, int32, int8

Softmax¶

A softmax layer.

Inputs¶

Input0 [Tensor]: The input to the softmax.

Attributes¶

axis [int]: The axis on which to perform the reduction. NOTE: This value is ignored by the UFF parser.

Supported Datatypes¶

float32, float16, int32, int8

BatchNorm¶

A batchnorm layer.

Inputs¶

Input0 [Tensor]: The input to the batchnorm. Must be 4 dimensional.
Gamma [Constant]: The gamma values.
Beta [Constant]: The beta values.
Mean [Constant]: The mean values.
Variance [Constant]: The variance values.

Attributes¶

epsilon [double]: The epsilon value.

Supported Datatypes¶

float32, float16, int32, int8

Shape¶

A shape layer. Returns the shape of its input. NOTE: the output of this layer is a Constant, and therefore will not work with layers expecting a Tensor input.

Inputs¶

Input0 [Tensor]: The input to the shape layer.

Supported Datatypes¶

float32, float16, int32, int8

StridedSlice¶

A strided slice layer.

Inputs¶

Input0 [Tensor or Constant]: The input to the strided slice.
Begin [Constant]: The indices at which to begin slicing.
End [Constant]: The indices at which to end slicing.
Strides [Constant]: Strides to use when slicing.

Attributes¶

begin_mask [int]: See TensorFlow stridedSlice documentation.
end_mask [int]: See TensorFlow stridedSlice documentation.
shrink_axis_mask [int]: See TensorFlow stridedSlice documentation. This value is ignored unless the input is a constant.

Supported Datatypes¶

float32, float16, int32, int8

Stack¶

A stack layer. NOTE: the output of this layer is a Constant, and therefore will not work with layers expecting a Tensor input.

Inputs¶

Inputs (variadic) [List[Constant]]: The inputs to the stack layer.

Attributes¶

axis [int]: The axis on which to stack. NOTE: this value is ignored by the UFF parser.

Supported Datatypes¶

float32, float16, int32, int8

Squeeze¶

Not implemented

Flatten¶

A flatten layer. A no-op in the UFF parser.

Inputs¶

Input0 [Tensor]: The tensor to flatten.

Supported Datatypes¶

float32, float16, int32, int8

Pad¶

A padding layer.

Inputs¶

Input0 [Tensor or Constant]: The input to pad. The input is automatically transposed if padding is applied to non-HW dimensions.
Padding [Constant]: The padding to apply. Padding is supported on 2 dimensions at most.

The output of a padding layer with a Constant input is treated as a Constant, and therefore will not work with layers expecting a Tensor input.

Supported Datatypes¶

float32, float16, int32, int8

Gather¶

A gather layer.

Inputs¶

Input0 [Tensor or Constant]: The input to the gather layer. If the input is constant, it is assumed to be of shape NC11.
Indices [Tensor or Constant]: The indices to gather along. These are assumed to be on the first non-batch dimension.

Supported Datatypes¶

float32, float16, int32, int8

GatherV2¶

A gatherV2 layer.

Inputs¶

Input0 [Tensor or Constant]: The input to the gather layer. If the input is constant, it is assumed to be of shape NC11.
Indices [Tensor or Constant]: The indices to gather along.

Attributes¶

axis [int]: The axis along which to gather, excluding batch dimension.

Supported Datatypes¶

float32, float16, int32, int8