/// Copyright (c) 2021, 2023 ARM Limited and Contributors. All rights reserved.
///
/// SPDX-License-Identifier: MIT
///
namespace armnn
{
/**
@page operator_list Arm NN Operators
@tableofcontents
@section S5_1_operator_list Arm NN Operators
Arm NN supports operators that are listed in below table.
Arm NN supports a wide list of data-types.
The main data-types that the Machine Learning functions support are the following:
Operator
| Description
| Equivalent Android NNAPI Operator
| Backends
| Data Layouts
| Data Types
|
AbsLayer
| Layer to perform absolute operation.
|
| CpuRef
|
|
|
---|
BFLOAT16
| FLOAT16
| FLOAT32
| QASYMMS8
| QASYMMU8
| QSYMMS16
| SIGNED32
|
|
CpuAcc
|
|
|
GpuAcc
|
|
|
ActivationLayer
| Layer to simulate an activation layer with the specified activation function.
|
- ANEURALNETWORKS_ABS
- ANEURALNETWORKS_ELU
- ANEURALNETWORKS_HARD_SWISH
- ANEURALNETWORKS_LOGISTIC
- ANEURALNETWORKS_PRELU
- ANEURALNETWORKS_RELU
- ANEURALNETWORKS_RELU1
- ANEURALNETWORKS_RELU6
- ANEURALNETWORKS_SQRT
- ANEURALNETWORKS_TANH
| CpuRef
|
|
|
---|
BFLOAT16
| FLOAT16
| FLOAT32
| QASYMMS8
| QASYMMU8
| QSYMMS16
|
|
CpuAcc
|
|
|
---|
QASYMMU8
| QASYMMS8
| QSYMMS16
| FLOAT16
| FLOAT32
|
|
GpuAcc
|
|
|
---|
QASYMMU8
| QASYMMS8
| QSYMMS16
| FLOAT16
| FLOAT32
|
|
AdditionLayer
| Layer to add 2 tensors.
|
| CpuRef
|
|
|
---|
BFLOAT16
| FLOAT16
| FLOAT32
| QASYMMS8
| QASYMMU8
| QSYMMS16
| SIGNED32
|
|
CpuAcc
|
|
|
---|
QASYMMU8
| QASYMMS8
| QSYMMS16
| SIGNED32
| FLOAT16
| FLOAT32
|
|
GpuAcc
|
|
|
---|
QASYMMU8
| QASYMMS8
| QSYMMS16
| SIGNED32
| FLOAT16
| FLOAT32
|
|
ArgMinMaxLayer
| Layer to calculate the index of the minimum or maximum values in a tensor
based on an axis.
|
- ANEURALNETWORKS_ARGMAX
- ANEURALNETWORKS_ARGMIN
| CpuRef
|
|
|
---|
BFLOAT16
| FLOAT16
| FLOAT32
| QASYMMS8
| QASYMMU8
| QSYMMS16
| SIGNED32
| SIGNED64
|
|
CpuAcc
|
|
|
---|
QASYMMU8
| QASYMMS8
| SIGNED32
| FLOAT16
| FLOAT32
|
|
GpuAcc
|
|
|
---|
QASYMMU8
| QASYMMS8
| SIGNED32
| FLOAT16
| FLOAT32
|
|
BatchMatMulLayer
| Layer to perform batch matrix multiplication.
|
| CpuRef
|
|
|
---|
BFLOAT16
| FLOAT16
| FLOAT32
| QASYMMS8
| QASYMMU8
| QSYMMS16
|
|
CpuAcc
|
|
|
GpuAcc
|
|
|
BatchNormalizationLayer
| Layer to perform batch normalization.
|
| CpuRef
|
|
|
---|
BFLOAT16
| FLOAT16
| FLOAT32
| QASYMMS8
| QASYMMU8
| QSYMMS16
|
|
CpuAcc
|
|
|
GpuAcc
|
|
|
BatchToSpaceNdLayer
| Layer to perform a batch to space transformation.
|
- ANEURALNETWORKS_BATCH_TO_SPACE_ND
| CpuRef
|
|
|
---|
BFLOAT16
| FLOAT16
| FLOAT32
| QASYMMS8
| QASYMMU8
| QSYMMS16
|
|
CpuAcc
|
|
|
GpuAcc
|
|
|
CastLayer
| Layer to cast a tensor to a type.
|
| CpuRef
|
|
|
---|
BFLOAT16
| FLOAT16
| FLOAT32
| QSYMMS8
| QASYMMS8
| QASYMMU8
| QSYMMS16
| SIGNED32
|
|
CpuAcc
|
|
|
---|
QASYMMS8
| QASYMMU8
| FLOAT16
| SIGNED32
| FLOAT32
|
|
GpuAcc
|
|
|
---|
QASYMMS8
| QASYMMU8
| SIGNED32
| FLOAT16
| FLOAT32
|
|
ChannelShuffleLayer
| Layer to reorganize the channels of a tensor.
|
- ANEURALNETWORKS_CHANNEL_SHUFFLE
| CpuRef
|
|
|
---|
FLOAT16
| FLOAT32
| QSYMMS8
| QASYMMS8
| QASYMMU8
|
|
CpuAcc
|
|
|
---|
QASYMMS8
| QASYMMU8
| FLOAT16
| FLOAT32
|
|
GpuAcc
|
|
|
---|
QASYMMS8
| QASYMMU8
| FLOAT16
| FLOAT32
|
|
ComparisonLayer
| Layer to compare 2 tensors.
|
- ANEURALNETWORKS_EQUAL
- ANEURALNETWORKS_GREATER
- ANEURALNETWORKS_GREATER_EQUAL
- ANEURALNETWORKS_LESS
- ANEURALNETWORKS_LESS_EQUAL
- ANEURALNETWORKS_NOT_EQUAL
| CpuRef
|
|
|
---|
BFLOAT16
| FLOAT16
| FLOAT32
| BOOLEAN
| QASYMMS8
| QASYMMU8
| QSYMMS16
| SIGNED32
|
|
CpuAcc
|
|
|
GpuAcc
|
|
|
ConcatLayer
| Layer to concatenate tensors along a given axis.
|
- ANEURALNETWORKS_CONCATENATION
| CpuRef
|
|
|
---|
BFLOAT16
| FLOAT16
| FLOAT32
| QASYMMS8
| QASYMMU8
| QSYMMS16
|
|
CpuAcc
|
|
|
---|
QASYMMU8
| QASYMMS8
| FLOAT16
| FLOAT32
|
|
GpuAcc
|
|
|
---|
QASYMMU8
| QASYMMS8
| FLOAT16
| FLOAT32
|
|
ConstantLayer
| Layer to provide a constant tensor.
|
| CpuRef
|
|
|
---|
BFLOAT16
| FLOAT16
| FLOAT32
| QASYMMS8
| QASYMMU8
| QSYMMS8
| QSYMMS16
| SIGNED32
|
|
CpuAcc
|
|
|
GpuAcc
|
|
|
ConvertFp16ToFp32Layer
| Layer to convert Float16 tensor to Float32 tensor.
|
| CpuRef
|
|
|
CpuAcc
|
|
|
GpuAcc
|
|
|
ConvertFp32ToFp16Layer
| Layer to convert Float32 tensor to Float16 tensor.
|
| CpuRef
|
|
|
CpuAcc
|
|
|
GpuAcc
|
|
|
Convolution2dLayer
| Layer to compute a convolution operation.
|
- ANEURALNETWORKS_CONV_2D
- ANEURALNETWORKS_GROUPED_CONV_2D
| CpuRef
|
|
|
---|
BFLOAT16
| FLOAT16
| FLOAT32
| QASYMMS8
| QASYMMU8
| QSYMMS16
|
|
CpuAcc
|
|
|
---|
SIGNED32
| FLOAT16
| FLOAT32
| QASYMMU8
| QASYMMS8
| QUANTIZEDSYMM8PERAXIS
|
|
GpuAcc
|
|
|
---|
SIGNED32
| FLOAT16
| FLOAT32
| QASYMMU8
| QASYMMS8
| QUANTIZEDSYMM8PERAXIS
|
|
Convolution3dLayer
| Layer to compute a 3D convolution operation.
|
| CpuRef
|
|
|
---|
BFLOAT16
| FLOAT16
| FLOAT32
| QASYMMS8
| QASYMMU8
| QSYMMS8
| QSYMMS16
|
|
CpuAcc
|
|
|
GpuAcc
|
|
|
DebugLayer
| Layer to print out inter layer tensor information.
|
| CpuRef
|
|
|
---|
BFLOAT16
| FLOAT16
| FLOAT32
| QASYMMS8
| QASYMMU8
| QSYMMS8
| QSYMMS16
| SIGNED32
|
|
DepthToSpaceLayer
| Layer to perform Depth to Space transformation.
|
- ANEURALNETWORKS_DEPTH_TO_SPACE
| CpuRef
|
|
|
---|
BFLOAT16
| FLOAT16
| FLOAT32
| QASYMMS8
| QASYMMU8
| QSYMMS16
|
|
CpuAcc
|
|
|
GpuAcc
|
|
|
DepthwiseConvolution2dLayer
| Layer to compute a deconvolution or transpose convolution.
|
- ANEURALNETWORKS_DEPTHWISE_CONV_2D
| CpuRef
|
|
|
---|
BFLOAT16
| FLOAT16
| FLOAT32
| QASYMMS8
| QASYMMU8
| QSYMMS8
| QSYMMS16
|
|
CpuAcc
|
|
|
---|
FLOAT16
| FLOAT32
| SIGNED32
| QASYMMU8
| QASYMMS8
| QUANTIZEDSYMM8PERAXIS
|
|
GpuAcc
|
|
|
---|
FLOAT16
| FLOAT32
| SIGNED32
| QASYMMU8
| QASYMMS8
| QUANTIZEDSYMM8PERAXIS
|
|
DequantizeLayer
| Layer to dequantize the values in a tensor.
|
- ANEURALNETWORKS_DEQUANTIZE
| CpuRef
|
|
|
---|
QASYMMS8
| QASYMMU8
| QSYMMS8
| QSYMMS16
|
|
CpuAcc
|
|
|
---|
FLOAT16
| FLOAT32
| QASYMMU8
| QASYMMS8
| QUANTIZEDSYMM8PERAXIS
| QSYMMS8
| QSYMMS16
|
|
GpuAcc
|
|
|
---|
FLOAT16
| FLOAT32
| QASYMMU8
| QASYMMS8
| QUANTIZEDSYMM8PERAXIS
| QSYMMS8
| QSYMMS16
|
|
DetectionPostProcessLayer
| Layer to generate the detection output based on center size encoded boxes, class prediction and anchors by doing non maximum suppression (NMS).
|
- ANEURALNETWORKS_DETECTION_POSTPROCESSING
| CpuRef
|
|
|
---|
BFLOAT16
| FLOAT16
| FLOAT32
| QASYMMS8
| QASYMMU8
| QSYMMS16
|
|
CpuAcc
|
|
|
---|
QASYMMU8
| QASYMMS8
| FLOAT32
|
|
DivisionLayer
| Layer to divide 2 tensors.
|
| CpuRef
|
|
|
---|
BFLOAT16
| FLOAT16
| FLOAT32
| QASYMMS8
| QASYMMU8
| QSYMMS16
| SIGNED32
|
|
CpuAcc
|
|
|
GpuAcc
|
|
|
ElementwiseBaseLayer
| Layer to perform Add - Div - Max - Min - Mul operations.
|
- ANEURALNETWORKS_ADD
- ANEURALNETWORKS_DIV
- ANEURALNETWORKS_MAXIMUM
- ANEURALNETWORKS_MINIMUM
- ANEURALNETWORKS_MUL
| CpuRef
|
|
|
---|
BFLOAT16
| FLOAT16
| FLOAT32
| QASYMMS8
| QASYMMU8
| QSYMMS16
| SIGNED32
|
|
CpuAcc
|
|
|
---|
QASYMMU8
| QASYMMS8
| QSYMMS16
| SIGNED32
| FLOAT16
| FLOAT32
|
|
GpuAcc
|
|
|
---|
QASYMMU8
| QASYMMS8
| QSYMMS16
| SIGNED32
| FLOAT16
| FLOAT32
|
|
ElementwiseBinaryLayer
| Layer to perform Power and Square Difference operations.
|
| CpuRef
|
|
|
---|
FLOAT16
| FLOAT32
| QASYMMS8
| QASYMMU8
| QSYMMS16
| SIGNED32
|
|
CpuAcc
|
|
|
GpuAcc
|
|
|
ElementwiseUnaryLayer
| Layer to perform Rsqrt - Exp - Neg - Log - Abs - Sin - Sqrt - Ceil operations.
|
- ANEURALNETWORKS_ABS
- ANEURALNETWORKS_EXP
- ANEURALNETWORKS_LOG
- ANEURALNETWORKS_NEG
- ANEURALNETWORKS_RSQRT
- ANEURALNETWORKS_SIN
- ANEURALNETWORKS_SQRT
| CpuRef
|
|
|
---|
BFLOAT16
| FLOAT16
| FLOAT32
| QASYMMS8
| QASYMMU8
| QSYMMS16
|
|
CpuAcc
|
|
|
GpuAcc
|
|
|
FakeQuantizationLayer
| Layer to quantize float values and dequantize afterwards. The current implementation does not dequantize the values.
|
| CpuRef
|
|
|
FillLayer
| Layer to set the values of a tensor with a given value.
|
| CpuRef
|
|
|
CpuAcc
|
|
|
GpuAcc
|
|
|
FloorLayer
| Layer to round the value to the lowest whole number.
|
| CpuRef
|
|
|
CpuAcc
|
|
|
GpuAcc
|
|
|
FullyConnectedLayer
| Layer to perform a fully connected / dense operation.
|
- ANEURALNETWORKS_FULLY_CONNECTED
| CpuRef
|
|
|
---|
BFLOAT16
| FLOAT16
| FLOAT32
| QASYMMS8
| QASYMMU8
| QSYMMS16
|
|
CpuAcc
|
|
|
---|
SIGNED32
| FLOAT16
| FLOAT32
| QASYMMU8
| QASYMMS8
|
|
GpuAcc
|
|
|
---|
SIGNED32
| FLOAT16
| FLOAT32
| QASYMMU8
| QASYMMS8
|
|
GatherLayer
| Layer to perform the gather operation along the chosen axis.
|
| CpuRef
|
|
|
---|
BFLOAT16
| FLOAT16
| FLOAT32
| QASYMMS8
| QASYMMU8
| QSYMMS16
| SIGNED32
|
|
CpuAcc
|
|
|
GpuAcc
|
|
|
GatherNdLayer
| Layer to perform the gatherNd operation.
|
| CpuRef
|
|
|
---|
BFLOAT16
| FLOAT16
| FLOAT32
| QASYMMS8
| QASYMMU8
| QSYMMS16
| SIGNED32
|
|
CpuAcc
|
|
|
---|
BFLOAT16
| FLOAT16
| FLOAT32
| QASYMMS8
| QASYMMU8
| QSYMMS16
| SIGNED32
|
|
GpuAcc
|
|
|
---|
BFLOAT16
| FLOAT16
| FLOAT32
| QASYMMS8
| QASYMMU8
| QSYMMS16
| SIGNED32
|
|
InputLayer
| Special layer used to provide input data to the computational network.
|
| All
|
|
|
InstanceNormalizationLayer
| Layer to perform an instance normalization on a given axis.
|
- ANEURALNETWORKS_INSTANCE_NORMALIZATION
| CpuRef
|
|
|
CpuAcc
|
|
|
GpuAcc
|
|
|
L2NormalizationLayer
| Layer to perform an L2 normalization on a given axis.
|
- ANEURALNETWORKS_L2_NORMALIZATION
| CpuRef
|
|
|
---|
BFLOAT16
| FLOAT16
| FLOAT32
| QASYMMS8
| QASYMMU8
| QSYMMS16
|
|
CpuAcc
|
|
|
GpuAcc
|
|
|
LogSoftmaxLayer
| Layer to perform the log softmax activations given logits.
|
| CpuRef
|
|
|
CpuAcc
|
|
|
---|
QASYMMU8
| QASYMMS8
| FLOAT16
| FLOAT32
|
|
GpuAcc
|
|
|
---|
QASYMMU8
| QASYMMS8
| FLOAT16
| FLOAT32
|
|
LogicalBinaryLayer
| Layer to perform Logical AND - Logical NOT - Logical OR operations.
|
- ANEURALNETWORKS_LOGICAL_AND
- ANEURALNETWORKS_LOGICAL_NOT
- ANEURALNETWORKS_LOGICAL_OR
| CpuRef
|
|
|
CpuAcc
|
|
|
GpuAcc
|
|
|
LstmLayer
| Layer to perform a single time step in a Long Short-Term Memory (LSTM) operation.
|
| CpuRef
|
|
|
---|
BFLOAT16
| FLOAT16
| QSYMMS16
|
|
CpuAcc
|
|
|
GpuAcc
|
|
|
MapLayer
| Layer to perform map operation on tensor.
|
| CpuRef
|
|
|
CpuAcc
|
|
|
GpuAcc
|
|
|
MaximumLayer
| Layer to perform an elementwise maximum of two tensors.
|
| CpuRef
|
|
|
---|
BFLOAT16
| FLOAT16
| FLOAT32
| QASYMMS8
| QASYMMU8
| QSYMMS16
| SIGNED32
|
|
CpuAcc
|
|
|
---|
QASYMMU8
| QASYMMS8
| FLOAT16
| FLOAT32
| SIGNED32
|
|
GpuAcc
|
|
|
---|
QASYMMU8
| QASYMMS8
| QSYMMS16
| FLOAT16
| FLOAT32
| SIGNED32
|
|
MeanLayer
| Layer to perform reduce mean operation.
|
| CpuRef
|
|
|
---|
BFLOAT16
| FLOAT16
| FLOAT32
| QASYMMS8
| QASYMMU8
| QSYMMS16
|
|
CpuAcc
|
|
|
---|
QASYMMU8
| QASYMMS8
| FLOAT16
| FLOAT32
|
|
GpuAcc
|
|
|
---|
QASYMMU8
| QASYMMS8
| FLOAT16
| FLOAT32
|
|
MemCopyLayer
| Layer to perform memory copy operation.
|
| CpuRef
|
|
|
---|
BFLOAT16
| FLOAT16
| FLOAT32
| QASYMMS8
| QASYMMU8
| QSYMMS16
| BOOLEAN
|
|
CpuAcc
|
|
|
GpuAcc
|
|
|
MemImportLayer
| Layer to perform memory import operation.
|
| CpuRef
|
|
|
CpuAcc
|
|
|
GpuAcc
|
|
|
MergeLayer
| Layer to concatenate tensors along a given axis.
|
- ANEURALNETWORKS_CONCATENATION
| CpuRef
|
|
|
---|
BFLOAT16
| FLOAT16
| FLOAT32
| QASYMMS8
| QASYMMU8
| QSYMMS16
|
|
CpuAcc
|
|
|
---|
QASYMMU8
| QASYMMS8
| FLOAT16
| FLOAT32
|
|
GpuAcc
|
|
|
---|
QASYMMU8
| QASYMMS8
| FLOAT16
| FLOAT32
|
|
MinimumLayer
| Layer to perform an elementwise minimum of two tensors.
|
| CpuRef
|
|
|
---|
BFLOAT16
| FLOAT16
| FLOAT32
| QASYMMS8
| QASYMMU8
| QSYMMS16
| SIGNED32
|
|
CpuAcc
|
|
|
---|
QASYMMU8
| QASYMMS8
| QSYMMS16
| FLOAT16
| FLOAT32
|
|
GpuAcc
|
|
|
---|
QASYMMU8
| QASYMMS8
| QSYMMS16
| FLOAT16
| FLOAT32
| SIGNED32
|
|
MultiplicationLayer
| Layer to perform an elementwise multiplication of two tensors.
|
| CpuRef
|
|
|
---|
BFLOAT16
| FLOAT16
| FLOAT32
| QASYMMS8
| QASYMMU8
| QSYMMS16
| SIGNED32
|
|
CpuAcc
|
|
|
---|
QASYMMU8
| QASYMMS8
| QSYMMS16
| SIGNED32
| FLOAT16
| FLOAT32
|
|
GpuAcc
|
|
|
---|
QASYMMU8
| QASYMMS8
| QSYMMS16
| SIGNED32
| FLOAT16
| FLOAT32
| SIGNED32
|
|
NormalizationLayer
| Layer to compute normalization operation.
|
- ANEURALNETWORKS_LOCAL_RESPONSE_NORMALIZATION
| CpuRef
|
|
|
---|
BFLOAT16
| FLOAT16
| FLOAT32
| QASYMMS8
| QASYMMU8
| QSYMMS16
|
|
CpuAcc
|
|
|
GpuAcc
|
|
|
OutputLayer
| A special layer providing access to a user supplied buffer into which the output of a network can be written.
|
| All
|
|
|
PadLayer
| Layer to pad a tensor.
|
- ANEURALNETWORKS_PAD
- ANEURALNETWORKS_PAD_V2
| CpuRef
|
|
|
---|
BFLOAT16
| FLOAT16
| FLOAT32
| QASYMMS8
| QASYMMU8
| QSYMMS16
|
|
CpuAcc
|
|
|
GpuAcc
|
|
|
PermuteLayer
| Layer to transpose an ND tensor.
|
- ANEURALNETWORKS_TRANSPOSE
| CpuRef
|
|
|
---|
BFLOAT16
| FLOAT16
| FLOAT32
| QASYMMS8
| QASYMMU8
| QSYMMS16
|
|
CpuAcc
|
|
|
GpuAcc
|
|
|
Pooling2dLayer
| Layer to perform 2D pooling with the specified pooling operation.
|
- ANEURALNETWORKS_AVERAGE_POOL_2D
- ANEURALNETWORKS_L2_POOL_2D
- ANEURALNETWORKS_MAX_POOL_2D
| CpuRef
|
|
|
---|
BFLOAT16
| FLOAT16
| FLOAT32
| QASYMMS8
| QASYMMU8
| QSYMMS16
|
|
CpuAcc
|
|
|
---|
QASYMMU8
| QASYMMS8
| FLOAT16
| FLOAT32
|
|
GpuAcc
|
|
|
---|
QASYMMU8
| QASYMMS8
| FLOAT16
| FLOAT32
|
|
Pooling3dLayer
| Layer to perform 3D pooling with the specified pooling operation.
|
- ANEURALNETWORKS_AVERAGE_POOL_3D
- ANEURALNETWORKS_L2_POOL_3D
- ANEURALNETWORKS_MAX_POOL_3D
| CpuRef
|
|
|
---|
BFLOAT16
| FLOAT16
| FLOAT32
| QASYMMS8
| QASYMMU8
| QSYMMS16
|
|
CpuAcc
|
|
|
GpuAcc
|
|
|
PreCompiledLayer
| Opaque layer provided by a backend which provides an executable representation of a subgraph from the original network.
|
| N/A
| N/A
| N/A
|
PreluLayer
| Layer to compute the activation layer with the PRELU activation function.
|
| CpuRef
|
|
|
---|
BFLOAT16
| FLOAT16
| FLOAT32
| QASYMMS8
| QASYMMU8
| QSYMMS16
|
|
CpuAcc
|
|
|
---|
QASYMMU8
| QASYMMS8
| FLOAT16
| FLOAT32
|
|
GpuAcc
|
|
|
---|
QASYMMU8
| QASYMMS8
| FLOAT16
| FLOAT32
|
|
QLstmLayer
| Layer to perform quantized LSTM (Long Short-Term Memory) operation.
|
- ANEURALNETWORKS_QUANTIZED_LSTM
- ANEURALNETWORKS_QUANTIZED_16BIT_LSTM
| CpuRef
|
|
|
CpuAcc
|
|
|
---|
QASYMMS8
| QASYMMU8
| SIGNED32
| QSYMMS16
|
|
GpuAcc
|
|
|
---|
QASYMMS8
| QASYMMU8
| SIGNED32
| QSYMMS16
|
|
QuantizeLayer
| Layer to perform quantization operation.
|
| CpuRef
|
|
|
---|
BFLOAT16
| FLOAT16
| FLOAT32
| QASYMMS8
| QASYMMU8
| QSYMMS8
| QSYMMS16
|
|
CpuAcc
|
|
|
---|
QASYMMU8
| QASYMMS8
| QASYMM16
| FLOAT16
| FLOAT32
|
|
GpuAcc
|
|
|
---|
QASYMMU8
| QASYMMS8
| QASYMM16
| FLOAT16
| FLOAT32
|
|
QuantizedLstmLayer
| Layer to perform quantized LSTM (Long Short-Term Memory) operation.
|
- ANEURALNETWORKS_QUANTIZED_LSTM
- ANEURALNETWORKS_QUANTIZED_16BIT_LSTM
| CpuRef
|
|
|
CpuAcc
|
|
|
---|
SIGNED32
| QASYMMU8
| QSYMMS16
|
|
GpuAcc
|
|
|
---|
SIGNED32
| QASYMMU8
| QSYMMS16
|
|
RankLayer
| Layer to perform a rank operation.
|
| CpuRef
|
|
|
CpuAcc
|
|
|
GpuAcc
|
|
|
ReduceLayer
| Layer to perform reduce with the following operations - ARG_IDX_MAX: Index of the max value - ARG_IDX_MIN: Index of the min value - MEAN_SUM: Mean of sum - PROD: Product - SUM_SQUARE: Sum of squares - SUM: Sum - MIN: Min - MAX: Max
|
- ANEURALNETWORKS_REDUCE_MAX
- ANEURALNETWORKS_REDUCE_MIN
- ANEURALNETWORKS_REDUCE_SUM
- ANEURALNETWORKS_REDUCE_PROD
| CpuRef
|
|
|
---|
BFLOAT16
| FLOAT16
| FLOAT32
| QASYMMS8
| QASYMMU8
| QSYMMS16
| SIGNED32
|
|
CpuAcc
|
|
|
---|
QASYMMU8
| QASYMMS8
| FLOAT16
| FLOAT32
| SIGNED32
|
|
GpuAcc
|
|
|
---|
QASYMMU8
| QASYMMS8
| FLOAT16
| FLOAT32
| SIGNED32
|
|
ReshapeLayer
| Layer to reshape a tensor.
|
- ANEURALNETWORKS_RESHAPE
- ANEURALNETWORKS_SQUEEZE
- ANEURALNETWORKS_EXPAND_DIMS
| CpuRef
|
|
|
---|
BFLOAT16
| FLOAT16
| FLOAT32
| QASYMMS8
| QASYMMU8
| QSYMMS16
| SIGNED32
| BOOLEAN
|
|
CpuAcc
|
|
|
GpuAcc
|
|
|
ResizeLayer
| Layer to perform resize of a tensor using one of the interpolation methods: - Bilinear - Nearest Neighbor.
|
- ANEURALNETWORKS_RESIZE_BILINEAR
- ANEURALNETWORKS_RESIZE_NEAREST_NEIGHBOR
| CpuRef
|
|
|
---|
BFLOAT16
| FLOAT16
| FLOAT32
| QASYMMS8
| QASYMMU8
| QSYMMS16
|
|
CpuAcc
|
|
|
---|
QASYMMU8
| QASYMMS8
| FLOAT16
| FLOAT32
|
|
ReverseV2Layer
| Layer to perform ReverseV2 operations.
|
| CpuRef
|
|
|
---|
BFLOAT16
| FLOAT16
| FLOAT32
| QASYMMS8
| QASYMMU8
| QSYMMS16
|
|
CpuAcc
|
|
|
GpuAcc
|
|
|
GpuAcc
|
|
|
---|
QASYMMU8
| QASYMMS8
| FLOAT16
| FLOAT32
|
|
RsqrtLayer
| Layer to perform Rsqrt operation.
|
| CpuRef
|
|
|
---|
BFLOAT16
| FLOAT16
| FLOAT32
| QASYMMS8
| QASYMMU8
| QSYMMS16
| SIGNED32
|
|
CpuAcc
|
|
|
GpuAcc
|
|
|
ShapeLayer
| Layer to return the shape of the input tensor.
|
| CpuRef
|
|
|
CpuAcc
|
|
|
GpuAcc
|
|
|
SliceLayer
| Layer to perform tensor slicing.
|
| CpuRef
|
|
|
---|
BFLOAT16
| FLOAT32
| QASYMMS8
| QASYMMU8
| QSYMMS16
|
|
CpuAcc
|
|
|
GpuAcc
|
|
|
SoftmaxLayer
| Layer to perform softmax, log-softmax operation over the specified axis.
|
- ANEURALNETWORKS_LOG_SOFTMAX
- ANEURALNETWORKS_SOFTMAX
| CpuRef
|
|
|
---|
BFLOAT16
| FLOAT16
| FLOAT32
| QASYMMS8
| QASYMMU8
| QSYMMS8
| QSYMMS16
|
|
CpuAcc
|
|
|
---|
QASYMMU8
| QASYMMS8
| FLOAT16
| FLOAT32
|
|
GpuAcc
|
|
|
---|
QASYMMU8
| QASYMMS8
| FLOAT16
| FLOAT32
|
|
SpaceToBatchNdLayer
| Layer to divide spatial dimensions of the tensor into a grid of blocks and interleaves these blocks with the batch dimension.
|
- ANEURALNETWORKS_SPACE_TO_BATCH_ND
| CpuRef
|
|
|
---|
BFLOAT16
| FLOAT16
| FLOAT32
| QASYMMS8
| QASYMMU8
| QSYMMS16
|
|
CpuAcc
|
|
|
GpuAcc
|
|
|
SpaceToDepthLayer
| Layer to rearrange blocks of spatial data into depth.
|
- ANEURALNETWORKS_SPACE_TO_DEPTH
| CpuRef
|
|
|
---|
BFLOAT16
| FLOAT16
| FLOAT32
| QASYMMS8
| QASYMMU8
| QSYMMS16
|
|
CpuAcc
|
|
|
GpuAcc
|
|
|
SplitterLayer
| Layer to split a tensor along a given axis.
|
| CpuRef
|
|
|
---|
BFLOAT16
| FLOAT16
| FLOAT32
| QASYMMS8
| QASYMMU8
| QSYMMS16
|
|
CpuAcc
|
|
|
GpuAcc
|
|
|
StackLayer
| Layer to stack tensors along an axis.
|
| CpuRef
|
|
|
---|
BFLOAT16
| FLOAT16
| FLOAT32
| QASYMMS8
| QASYMMU8
| QSYMMS16
|
|
CpuAcc
|
|
|
GpuAcc
|
|
|
StandInLayer
| A layer to represent "unknown" or "unsupported" operations in the input graph. It has a configurable number of input and output slots and an optional name.
|
| N/A
| N/A
| N/A
|
StridedSliceLayer
| Layer to extract a strided slice of a tensor.
|
- ANEURALNETWORKS_STRIDED_SLICE
| CpuRef
|
|
|
---|
BFLOAT16
| FLOAT32
| QASYMMS8
| QASYMMU8
| QSYMMS16
|
|
CpuAcc
|
|
|
GpuAcc
|
|
|
SubtractionLayer
| Layer to perform an elementwise subtract of 2 tensors.
|
| CpuRef
|
|
|
---|
BFLOAT16
| FLOAT16
| FLOAT32
| QASYMMS8
| QASYMMU8
| QSYMMS16
| SIGNED32
|
|
CpuAcc
|
|
|
---|
QASYMMU8
| QASYMMS8
| QSYMMS16
| SIGNED32
| FLOAT16
| FLOAT32
|
|
GpuAcc
|
|
|
---|
QASYMMU8
| QASYMMS8
| QSYMMS16
| SIGNED32
| FLOAT16
| FLOAT32
|
|
TransposeConvolution2dLayer
| Layer to perform 2D transpose convolution (deconvolution) operation.
|
- ANEURALNETWORKS_TRANSPOSE_CONV_2D
| CpuRef
|
|
|
---|
BFLOAT16
| FLOAT16
| FLOAT32
| QASYMMS8
| QASYMMU8
| QSYMMS8
| QSYMMS16
|
|
CpuAcc
|
|
|
---|
SIGNED32
| FLOAT16
| FLOAT32
| QASYMMU8
| QASYMMS8
| QUANTIZEDSYMM8PERAXIS
|
|
GpuAcc
|
|
|
---|
SIGNED32
| FLOAT16
| FLOAT32
| QASYMMU8
| QASYMMS8
| QUANTIZEDSYMM8PERAXIS
|
|
TransposeLayer
| Layer to transpose a tensor.
|
- ANEURALNETWORKS_TRANSPOSE
| CpuRef
|
|
|
---|
BFLOAT16
| FLOAT16
| FLOAT32
| QASYMMS8
| QASYMMU8
| QSYMMS16
|
|
CpuAcc
|
|
|
GpuAcc
|
|
|
UnidirectionalSquenceLstmLayer
| Layer to perform unidirectional sequence LSTM operation.
|
- ANEURALNETWORKS_UNIDIRECTIONAL_SEQUENCE_LSTM
| CpuRef
|
|
Weight Types
|
---|
FLOAT32
| QASYMMS8
|
|
CpuAcc
|
|
|
GpuAcc
|
|
|
UnmapLayer
| Layer to perform unmap operation on tensor.
|
| CpuRef
|
|
|
CpuAcc
|
|
|
GpuAcc
|
|
|
*/
} // namespace