ARM plugin operation set specification

This specification document based on opset7 operation set supported in OpenVINO. Examples of operation instances are expressed as IR V10 xml snippets. Such IR is generated by the Model Optimizer. The semantics match corresponding nGraph operation classes declared in namespace opset7.

NOTE: ARM plugin doesn't support operations with dynamic shape.

Table of Contents

Layers	ARM plugin
Abs	Supported
Acos	Supported**
Acosh	Supported**
AdaptiveAvgPool	Supported**
AdaptiveMaxPool	Supported**
Add	Supported
Asin	Supported**
Asinh	Supported**
Assign	Not Supported
Atan	Supported**
Atanh	Supported**
AvgPool	Supported***
BatchNormInference	Supported
BatchToSpace	Supported*
BinaryConvolution	Not Supported
Broadcast	Supported
Bucketize	Supported**
CTCGreedyDecoder	Supported**
CTCGreedyDecoderSeqLen	Supported**
CTCLoss	Supported**
Ceiling	Supported
Clamp	Supported
Concat	Supported
Constant	Supported
Convert	Supported*
ConvertLike	Supported*
Convolution	Supported***
ConvolutionBackpropData	Supported**
Cos	Supported**
Cosh	Supported**
CumSum	Supported**
DeformableConvolution	Not Supported
DeformablePSROIPooling	Not Supported
DepthToSpace	Supported*
DetectionOutput	Supported**
DFT	Supported**
Divide	Supported
Elu	Supported
EmbeddingBagOffsetsSum	Supported**
EmbeddingBagPackedSum	Supported**
EmbeddingSegmentsSum	Supported**
Equal	Supported*
Erf	Supported**
Exp	Supported
ExperimentalDetectronDetectionOutput_6	Not Supported
ExperimentalDetectronGenerateProposalsSingleImage_6	Not Supported
ExperimentalDetectronPriorGridGenerator_6	Not Supported
ExperimentalDetectronROIFeatureExtractor_6	Not Supported
ExperimentalDetectronTopKROIs_6	Not Supported
ExtractImagePatches	Not Supported
FakeQuantize	Not Supported
Floor	Supported
FloorMod	Supported**
Gather	Supported*
GatherElements	Supported**
GatherND_5	Supported**
GatherTree	Supported**
Gelu	Supported**
Greater	Supported
GreaterEqual	Supported
GRN	Supported
GroupConvolution	Supported***
GroupConvolutionBackpropData	Supported**
GRUCell	Supported
GRUSequence	Supported**
HardSigmoid	Supported**
HSigmoid	Supported**
HSwish	Supported
IDFT	Supported**
Interpolate	Supported*
I420toBGR	Supported**
I420toRGB	Supported**
Less	Supported*
LessEqual	Supported*
Log	Supported
LogicalAnd	Supported
LogicalNot	Supported
LogicalOr	Supported
LogicalXor	Supported
LogSoftmax	Supported
Loop	Not Supported
LRN	Supported*
LSTMCell	Supported
LSTMSequence	Supported**
MatMul	Supported
MaxPool-1	Supported*
MaxPool-8	Supported**
Maximum	Supported
Minimum	Supported
Mish	Supported
Mod	Supported*
MVN	Supported*
Multiply	Supported
Negative	Supported
NonMaxSuppression	Supported**
NonZero	Not Supported
NormalizeL2	Supported*
NotEqual	Supported*
NV12toBGR	Supported**
NV12toRGB	Supported**
OneHot	Supported**
Pad	Supported*
Parameter	Supported
Power	Supported
PReLU	Supported
PriorBoxClustered	Supported
PriorBox	Supported
Proposal	Supported**
PSROIPooling	Supported**
Range	Not Supported
ReLU	Supported
ReadValue	Not Supported
ReduceL1	Supported
ReduceL2	Supported
ReduceLogicalAnd	Supported**
ReduceLogicalOr	Supported**
ReduceMax	Supported
ReduceMean	Supported
ReduceMin	Supported
ReduceProd	Supported
ReduceSum	Supported
RegionYolo	Supported**
ReorgYolo	Supported
Reshape	Supported
Result	Supported
ReverseSequence	Supported**
RNNCell	Supported
RNNSequence	Supported**
ROIAlign	Supported**
ROIPooling	Supported**
Roll	Not Supported
Round	Supported*
ScatterElementsUpdate	Supported**
ScatterNDUpdate	Supported**
ScatterUpdate	Supported**
Select	Supported
Selu	Supported**
ShapeOf	Not Supported
ShuffleChannels	Supported
Sigmoid	Supported
Sign	Supported
Sin	Supported
Sinh	Supported**
SoftMax	Supported
SoftPlus	Supported
SpaceToBatch	Supported*
SpaceToDepth	Supported*
Split	Supported
Sqrt	Supported
SquaredDifference	Supported
Squeeze	Supported
StridedSlice	Supported*
Subtract	Supported
Swish	Supported
Tan	Supported**
Tanh	Supported
TensorIterator	Supported
Tile	Supported
TopK	Supported**
Transpose	Supported*
Unsqueeze	Supported
VariadicSplit	Supported

*- support is limited to the specific parameters. Refer to "Known Layers Limitation" section.

**- support is implemented via ngraph::reference.

***- supports only 4D tensors.

Known Layers Limitations

• 'BatchToSpace' layer is supported 4D tensors only and constant nodes: block_shape with N = 1 and C= 1, crops_begin with zero values and crops_end with zero values.

• 'Convert' layer is supported via arm_compute library with configuration: src data type -> dst data type

  • U8 -> U16, S16, S32
  • U16 -> U8, U32
  • S16 -> U8, S32
  • F16 -> F32

  for another cases layer is implemented via ngraph::reference.

• 'ConvertLike' layer is supported configuration like 'Convert'.

• 'DepthToSpace' layer is supported 4D tensors only and for BLOCKS_FIRST of 'mode' attribute.

• 'Equal' doesn't support broadcast for inputs.

• 'Gather' layer is supported constant scalar or 1D indices axes only. Layer is supported as via arm_compute library for non negative indices and via reference implementation otherwise.

• 'Interpolate' layer is supported 4D tensors and zero value of 'pads_begin' and 'pads_end' attributes only.

• 'Less' doesn't support broadcast for inputs.

• 'LessEqual' doesn't support broadcast for inputs.

• 'LRN' layer is supported axes = {1} or axes = {2, 3} only.

• 'MaxPool-1' layer is supported via arm_compute library for 4D input tensor and via reference implementation for another cases.

• 'Mod' layer is supported for FP32 only.

• 'MVN' layer is supported via arm_compute library for 2D inputs and false value of 'normalize_variance' and false value of 'across_channels', for another cases layer is implemented via ngraph::reference.

• 'Normalize' layer is supported via arm_compute library with MAX value of 'eps_mode' and axes = {2 | 3}, and for ADD value of 'eps_mode' layer uses 'DecomposeNormalizeL2Add', for another cases layer is implemented via ngraph::reference.

• 'NotEqual'doesn't support broadcast for inputs.

• 'Pad' layer works with 'pad_mode' = {REFLECT | CONSTANT | SYMMETRIC} parameters only.

• 'Round' layer is supported via arm_compute library with RoundMode::HALF_AWAY_FROM_ZERO value of mode, for another cases layer is implemented via ngraph::reference.

• 'SpaceToBatch' layer is supported 4D tensors only and constant nodes: shapes, pads_begin or pads_end with zero paddings for batch or channels and one values shapes for batch and channels.

• 'SpaceToDepth' layer is supported 4D tensors only and for BLOCKS_FIRST of 'mode' attribute.

• 'StridedSlice' layer is supported via arm_compute library for tensors with dims < 5 and zero values of ellipsis_mask or zero values of new_axis_mask and shrink_axis_mask, for another cases layer is implemented via ngraph::reference.

• 'Transpose'layer is supported via arm_compute library for tensors with dims < 5, for another cases layer is implemented via ngraph::reference.