[TensorRT] 지원되는 연산자 목록 (2020.04.29 기준)

 

TensorRT 에서 지원되는 Operators 는 다음과 같다. (2020.04.29 기준)

 

 

Caffe

 

• BatchNormalization
• BNLL
• Clip 
• Concatenation
• Convolution
• Crop
• Deconvolution
• Dropout
• ElementWise
• ELU
• InnerProduct
• Input
• LeakyReLU
• LRN
• Permute
• Pooling
• Power
• Reduction
• ReLU, TanH, and Sigmoid
• Reshape
• SoftMax
• Scale

Clip : When using the Clip operation, Caffe users must serialize their layers using ditcaffe.pb.h instead of caffe.pb.h in order to import the layer into TensorRT.

 

 

 

 

 

 

TensorFlow

 

• Add, Sub, Mul, Div, Minimum and Maximum
• ArgMax
• ArgMin
• AvgPool
• BiasAdd
• Clip
• ConcatV2
• Const
• Conv2D
• ConvTranspose2D
• DepthwiseConv2dNative
• Elu
• ExpandDims
• FusedBatchNorm
• Identity
• LeakyReLU
• MaxPool
• Mean
• Negative, Abs, Sqrt, Recip, Rsqrt, Pow, Exp and Log
• Pad is supported if followed by one of these TensorFlow layers: Conv2D, DepthwiseConv2dNative, MaxPool, and AvgPool.
• Placeholder
• ReLU, TanH, and Sigmoid
• Relu6
• Reshape
• Sin, Cos, Tan, Asin, Acos, Atan, Sinh, Cosh, Asinh, Acosh, Atanh, Ceil and Floor
• Selu
• Slice
• SoftMax
• Note: If the input to a TensorFlow SoftMax op is not NHWC, TensorFlow will automatically insert a transpose layer with a non-constant permutation, causing the UFF converter to fail. It is therefore advisable to manually transpose SoftMax inputs to NHWC using a constant permutation.
• Softplus
• Softsign
• Transpose

 

 

 

 

ONNX

• Abs
• Acos
• Acosh
• And
• Asin
• Asinh
• Atan
• Atanh
• Add
• ArgMax
• ArgMin
• AveragePool
• BatchNormalization
• Cast
• Ceil
• Clip
• Concat
• Constant
• ConstantOfShape
• Conv
• ConvTranspose
• Cos
• Cosh
• DepthToSpace
• DequantizeLinear
• Div
• Dropout
• Elu
• Equal
• Erf
• Exp
• Expand
• Flatten
• Floor
• Gather
• Gemm
• GlobalAveragePool
• GlobalMaxPool
• Greater
• GRU
• HardSigmoid
• Identity
• ImageScaler
• InstanceNormalization
• LRN
• LeakyRelU
• Less
• Log
• LogSoftmax
• Loop
• LRN
• LSTM
• MatMul
• Max
• MaxPool
• Mean
• Min
• Mul
• Neg
• Not
• Or
• Pad
• ParametricSoftplus
• Pow
• PRelu
• QuantizeLinear
• RandomUniform
• RandomUniformLike
• Range
• Reciprocal
• ReduceL1
• ReduceL2
• ReduceLogSum
• ReduceLogSumExp
• ReduceMax
• ReduceMean
• ReduceMin
• ReduceProd
• ReduceSum
• ReduceSumSquare
• Relu
• Reshape
• Resize
• RNN
• ScaledTanh
• Scan
• Selu
• Shape
• Sigmoid
• Sin
• Sinh
• Size
• Slice
• Softmax
• Softplus
• Softsign
• SpaceToDepth
• Split
• Sqrt
• Squeeze
• Sub
• Sum
• Tan
• Tanh
• ThresholdedRelu
• Tile
• TopK
• Transpose
• Unsqueeze
• Upsample
• Where

 

 

참고자료 : https://docs.nvidia.com/deeplearning/sdk/tensorrt-support-matrix/index.html#fntarg_12

TensorRT Support Matrix :: NVIDIA Deep Learning SDK Documentation