6750 lines
138 KiB
Plaintext
6750 lines
138 KiB
Plaintext
mappings {
|
|
frameworkName: "onnx"
|
|
opName: "add"
|
|
inputFrameworkOpName: "Add"
|
|
rule {
|
|
ruleName: "ndarraymapping"
|
|
functionName: "ndarraymapping"
|
|
inputTensorName: "A"
|
|
inputTensorName: "B"
|
|
outputTensorName: "input"
|
|
outputTensorName: "y"
|
|
inputToOutput {
|
|
key: "input"
|
|
value: "A"
|
|
}
|
|
inputToOutput {
|
|
key: "y"
|
|
value: "B"
|
|
}
|
|
ruleType: "tensor"
|
|
inputFrameworkOpName: "Add"
|
|
}
|
|
rule {
|
|
ruleName: "argdescriptorconstant"
|
|
functionName: "argdescriptorconstant"
|
|
inputBooleanName: "inPlace"
|
|
ruleType: "attribute"
|
|
transformerArgs {
|
|
key: "value"
|
|
transformerArgs {
|
|
name: "inPlace"
|
|
argType: BOOL
|
|
}
|
|
}
|
|
inputFrameworkOpName: "Add"
|
|
}
|
|
}
|
|
mappings {
|
|
frameworkName: "onnx"
|
|
opName: "tan"
|
|
inputFrameworkOpName: "Tan"
|
|
rule {
|
|
ruleName: "ndarraymapping"
|
|
functionName: "ndarraymapping"
|
|
inputTensorName: "input"
|
|
outputTensorName: "input"
|
|
inputToOutput {
|
|
key: "input"
|
|
value: "input"
|
|
}
|
|
ruleType: "tensor"
|
|
inputFrameworkOpName: "Tan"
|
|
}
|
|
rule {
|
|
ruleName: "argdescriptorconstant"
|
|
functionName: "argdescriptorconstant"
|
|
inputBooleanName: "inPlace"
|
|
ruleType: "attribute"
|
|
transformerArgs {
|
|
key: "value"
|
|
transformerArgs {
|
|
name: "inPlace"
|
|
argType: BOOL
|
|
}
|
|
}
|
|
inputFrameworkOpName: "Tan"
|
|
}
|
|
}
|
|
mappings {
|
|
frameworkName: "onnx"
|
|
opName: "or"
|
|
inputFrameworkOpName: "Or"
|
|
rule {
|
|
ruleName: "ndarraymapping"
|
|
functionName: "ndarraymapping"
|
|
inputTensorName: "A"
|
|
outputTensorName: "input"
|
|
inputToOutput {
|
|
key: "input"
|
|
value: "A"
|
|
}
|
|
ruleType: "tensor"
|
|
inputFrameworkOpName: "Or"
|
|
}
|
|
rule {
|
|
ruleName: "argdescriptorconstant"
|
|
functionName: "argdescriptorconstant"
|
|
inputBooleanName: "inPlace"
|
|
ruleType: "attribute"
|
|
transformerArgs {
|
|
key: "value"
|
|
transformerArgs {
|
|
name: "inPlace"
|
|
argType: BOOL
|
|
}
|
|
}
|
|
inputFrameworkOpName: "Or"
|
|
}
|
|
rule {
|
|
ruleName: "argdescriptorconstant"
|
|
functionName: "argdescriptorconstant"
|
|
inputFloatName: "comparable"
|
|
ruleType: "attribute"
|
|
transformerArgs {
|
|
key: "value"
|
|
transformerArgs {
|
|
name: "comparable"
|
|
argType: DOUBLE
|
|
}
|
|
}
|
|
inputFrameworkOpName: "Or"
|
|
}
|
|
}
|
|
mappings {
|
|
frameworkName: "onnx"
|
|
opName: "reduce_max"
|
|
inputFrameworkOpName: "ReduceMax"
|
|
rule {
|
|
ruleName: "ndarraymapping"
|
|
functionName: "ndarraymapping"
|
|
inputTensorName: "data"
|
|
outputTensorName: "input"
|
|
inputToOutput {
|
|
key: "input"
|
|
value: "data"
|
|
}
|
|
ruleType: "tensor"
|
|
inputFrameworkOpName: "ReduceMax"
|
|
}
|
|
rule {
|
|
ruleName: "invertbooleannumber"
|
|
functionName: "invertbooleannumber"
|
|
inputIntName: "keepdims"
|
|
outputBooleanName: "keepDims"
|
|
inputToOutput {
|
|
key: "keepDims"
|
|
value: "keepdims"
|
|
}
|
|
ruleType: "attribute"
|
|
inputFrameworkOpName: "ReduceMax"
|
|
}
|
|
rule {
|
|
ruleName: "listnumbertolistnumber"
|
|
functionName: "listnumbertolistnumber"
|
|
outputIntName: "dimensions"
|
|
inputToOutput {
|
|
key: "dimensions"
|
|
value: "axes"
|
|
}
|
|
ruleType: "attribute"
|
|
inputFrameworkOpName: "ReduceMax"
|
|
}
|
|
}
|
|
mappings {
|
|
frameworkName: "onnx"
|
|
opName: "maxpool2d"
|
|
inputFrameworkOpName: "MaxPool"
|
|
rule {
|
|
ruleName: "ndarraymapping"
|
|
functionName: "ndarraymapping"
|
|
inputTensorName: "X"
|
|
outputTensorName: "input"
|
|
inputToOutput {
|
|
key: "input"
|
|
value: "X"
|
|
}
|
|
ruleType: "tensor"
|
|
inputFrameworkOpName: "MaxPool"
|
|
}
|
|
rule {
|
|
ruleName: "argdescriptorconstant"
|
|
functionName: "argdescriptorconstant"
|
|
inputIntName: "isNCHW"
|
|
ruleType: "attribute"
|
|
transformerArgs {
|
|
key: "value"
|
|
transformerArgs {
|
|
name: "isNCHW"
|
|
argType: INT64
|
|
argIndex: 10
|
|
}
|
|
}
|
|
inputFrameworkOpName: "MaxPool"
|
|
}
|
|
rule {
|
|
ruleName: "argdescriptorconstant"
|
|
functionName: "argdescriptorconstant"
|
|
inputIntName: "extraParam0"
|
|
ruleType: "attribute"
|
|
transformerArgs {
|
|
key: "value"
|
|
transformerArgs {
|
|
name: "extraParam0"
|
|
argType: INT64
|
|
argIndex: 9
|
|
}
|
|
}
|
|
inputFrameworkOpName: "MaxPool"
|
|
}
|
|
rule {
|
|
ruleName: "argdescriptorconstant"
|
|
functionName: "argdescriptorconstant"
|
|
inputIntName: "isSameMode"
|
|
ruleType: "attribute"
|
|
transformerArgs {
|
|
key: "value"
|
|
transformerArgs {
|
|
name: "isSameMode"
|
|
argType: INT64
|
|
argIndex: 8
|
|
}
|
|
}
|
|
inputFrameworkOpName: "MaxPool"
|
|
}
|
|
rule {
|
|
ruleName: "listattributevaluelookuptoindex"
|
|
functionName: "listattributevaluelookuptoindex"
|
|
inputIntName: "dH"
|
|
outputIntName: "dH"
|
|
inputFloatName: "dilations"
|
|
inputToOutput {
|
|
key: "dH"
|
|
value: "dilations"
|
|
}
|
|
ruleType: "attribute"
|
|
transformerArgs {
|
|
key: "dH"
|
|
transformerArgs {
|
|
name: "dilations"
|
|
argIndex: 6
|
|
}
|
|
transformerArgs {
|
|
name: "dH"
|
|
int64Value: 1
|
|
argType: INT64
|
|
argIndex: 6
|
|
}
|
|
}
|
|
transformerArgs {
|
|
key: "dH"
|
|
transformerArgs {
|
|
name: "dilations"
|
|
argIndex: 6
|
|
}
|
|
transformerArgs {
|
|
name: "dH"
|
|
int64Value: 1
|
|
argType: INT64
|
|
argIndex: 6
|
|
}
|
|
}
|
|
inputFrameworkOpName: "MaxPool"
|
|
}
|
|
rule {
|
|
ruleName: "listattributevaluelookuptoindex"
|
|
functionName: "listattributevaluelookuptoindex"
|
|
inputIntName: "dW"
|
|
outputIntName: "dW"
|
|
inputFloatName: "dilations"
|
|
inputToOutput {
|
|
key: "dW"
|
|
value: "dilations"
|
|
}
|
|
ruleType: "attribute"
|
|
transformerArgs {
|
|
key: "dW"
|
|
transformerArgs {
|
|
name: "dilations"
|
|
int64Value: 1
|
|
argIndex: 7
|
|
}
|
|
transformerArgs {
|
|
name: "dW"
|
|
int64Value: 1
|
|
argType: INT64
|
|
argIndex: 7
|
|
}
|
|
}
|
|
transformerArgs {
|
|
key: "dW"
|
|
transformerArgs {
|
|
name: "dilations"
|
|
int64Value: 1
|
|
argIndex: 7
|
|
}
|
|
transformerArgs {
|
|
name: "dW"
|
|
int64Value: 1
|
|
argType: INT64
|
|
argIndex: 7
|
|
}
|
|
}
|
|
inputFrameworkOpName: "MaxPool"
|
|
}
|
|
rule {
|
|
ruleName: "listattributevaluelookuptoindex"
|
|
functionName: "listattributevaluelookuptoindex"
|
|
inputIntName: "pads"
|
|
outputIntName: "pH"
|
|
inputFloatName: "pads"
|
|
inputToOutput {
|
|
key: "pH"
|
|
value: "pads"
|
|
}
|
|
ruleType: "attribute"
|
|
transformerArgs {
|
|
key: "pH"
|
|
transformerArgs {
|
|
name: "pads"
|
|
argIndex: 4
|
|
}
|
|
transformerArgs {
|
|
name: "pads"
|
|
argType: INT64
|
|
argIndex: 4
|
|
}
|
|
}
|
|
transformerArgs {
|
|
key: "pH"
|
|
transformerArgs {
|
|
name: "pads"
|
|
argIndex: 4
|
|
}
|
|
transformerArgs {
|
|
name: "pads"
|
|
argType: INT64
|
|
argIndex: 4
|
|
}
|
|
}
|
|
inputFrameworkOpName: "MaxPool"
|
|
}
|
|
rule {
|
|
ruleName: "listattributevaluelookuptoindex"
|
|
functionName: "listattributevaluelookuptoindex"
|
|
inputIntName: "pads"
|
|
outputIntName: "pW"
|
|
inputFloatName: "pads"
|
|
inputToOutput {
|
|
key: "pW"
|
|
value: "pads"
|
|
}
|
|
ruleType: "attribute"
|
|
transformerArgs {
|
|
key: "pW"
|
|
transformerArgs {
|
|
name: "pads"
|
|
int64Value: 1
|
|
argIndex: 5
|
|
}
|
|
transformerArgs {
|
|
name: "pads"
|
|
argType: INT64
|
|
argIndex: 5
|
|
}
|
|
}
|
|
transformerArgs {
|
|
key: "pW"
|
|
transformerArgs {
|
|
name: "pads"
|
|
int64Value: 1
|
|
argIndex: 5
|
|
}
|
|
transformerArgs {
|
|
name: "pads"
|
|
argType: INT64
|
|
argIndex: 5
|
|
}
|
|
}
|
|
inputFrameworkOpName: "MaxPool"
|
|
}
|
|
rule {
|
|
ruleName: "listattributevaluelookuptoindex"
|
|
functionName: "listattributevaluelookuptoindex"
|
|
inputIntName: "sH"
|
|
outputIntName: "sH"
|
|
inputFloatName: "strides"
|
|
inputToOutput {
|
|
key: "sH"
|
|
value: "strides"
|
|
}
|
|
ruleType: "attribute"
|
|
transformerArgs {
|
|
key: "sH"
|
|
transformerArgs {
|
|
name: "strides"
|
|
argIndex: 2
|
|
}
|
|
transformerArgs {
|
|
name: "sH"
|
|
int64Value: 1
|
|
argType: INT64
|
|
argIndex: 6
|
|
}
|
|
}
|
|
transformerArgs {
|
|
key: "sH"
|
|
transformerArgs {
|
|
name: "strides"
|
|
argIndex: 2
|
|
}
|
|
transformerArgs {
|
|
name: "sH"
|
|
int64Value: 1
|
|
argType: INT64
|
|
argIndex: 6
|
|
}
|
|
}
|
|
inputFrameworkOpName: "MaxPool"
|
|
}
|
|
rule {
|
|
ruleName: "listattributevaluelookuptoindex"
|
|
functionName: "listattributevaluelookuptoindex"
|
|
inputIntName: "sW"
|
|
outputIntName: "sW"
|
|
inputFloatName: "strides"
|
|
inputToOutput {
|
|
key: "sW"
|
|
value: "strides"
|
|
}
|
|
ruleType: "attribute"
|
|
transformerArgs {
|
|
key: "sW"
|
|
transformerArgs {
|
|
name: "strides"
|
|
int64Value: 1
|
|
argIndex: 3
|
|
}
|
|
transformerArgs {
|
|
name: "sW"
|
|
int64Value: 1
|
|
argType: INT64
|
|
argIndex: 7
|
|
}
|
|
}
|
|
transformerArgs {
|
|
key: "sW"
|
|
transformerArgs {
|
|
name: "strides"
|
|
int64Value: 1
|
|
argIndex: 3
|
|
}
|
|
transformerArgs {
|
|
name: "sW"
|
|
int64Value: 1
|
|
argType: INT64
|
|
argIndex: 7
|
|
}
|
|
}
|
|
inputFrameworkOpName: "MaxPool"
|
|
}
|
|
rule {
|
|
ruleName: "listattributevaluelookuptoindex"
|
|
functionName: "listattributevaluelookuptoindex"
|
|
outputIntName: "kH"
|
|
inputFloatName: "kernel_shape"
|
|
inputToOutput {
|
|
key: "kH"
|
|
value: "kernel_shape"
|
|
}
|
|
ruleType: "attribute"
|
|
transformerArgs {
|
|
key: "kH"
|
|
transformerArgs {
|
|
name: "kernel_shape"
|
|
}
|
|
}
|
|
inputFrameworkOpName: "MaxPool"
|
|
}
|
|
rule {
|
|
ruleName: "listattributevaluelookuptoindex"
|
|
functionName: "listattributevaluelookuptoindex"
|
|
outputIntName: "kW"
|
|
inputFloatName: "kernel_shape"
|
|
inputToOutput {
|
|
key: "kW"
|
|
value: "kernel_shape"
|
|
}
|
|
ruleType: "attribute"
|
|
transformerArgs {
|
|
key: "kW"
|
|
transformerArgs {
|
|
name: "kernel_shape"
|
|
int64Value: 1
|
|
argIndex: 1
|
|
}
|
|
}
|
|
inputFrameworkOpName: "MaxPool"
|
|
}
|
|
}
|
|
mappings {
|
|
frameworkName: "onnx"
|
|
opName: "size"
|
|
inputFrameworkOpName: "Size"
|
|
rule {
|
|
ruleName: "ndarraymapping"
|
|
functionName: "ndarraymapping"
|
|
inputTensorName: "data"
|
|
outputTensorName: "input"
|
|
inputToOutput {
|
|
key: "input"
|
|
value: "data"
|
|
}
|
|
ruleType: "tensor"
|
|
inputFrameworkOpName: "Size"
|
|
}
|
|
}
|
|
mappings {
|
|
frameworkName: "onnx"
|
|
opName: "lrn"
|
|
inputFrameworkOpName: "LRN"
|
|
rule {
|
|
ruleName: "ndarraymapping"
|
|
functionName: "ndarraymapping"
|
|
inputTensorName: "X"
|
|
outputTensorName: "input"
|
|
inputToOutput {
|
|
key: "input"
|
|
value: "X"
|
|
}
|
|
ruleType: "tensor"
|
|
inputFrameworkOpName: "LRN"
|
|
}
|
|
rule {
|
|
ruleName: "valuemapping"
|
|
functionName: "valuemapping"
|
|
inputIntName: "size"
|
|
outputIntName: "depth"
|
|
inputFloatName: "alpha"
|
|
inputFloatName: "beta"
|
|
inputFloatName: "bias"
|
|
outputDoubleName: "alpha"
|
|
outputDoubleName: "beta"
|
|
outputDoubleName: "bias"
|
|
inputToOutput {
|
|
key: "alpha"
|
|
value: "alpha"
|
|
}
|
|
inputToOutput {
|
|
key: "beta"
|
|
value: "beta"
|
|
}
|
|
inputToOutput {
|
|
key: "bias"
|
|
value: "bias"
|
|
}
|
|
inputToOutput {
|
|
key: "depth"
|
|
value: "size"
|
|
}
|
|
ruleType: "attribute"
|
|
inputFrameworkOpName: "LRN"
|
|
}
|
|
rule {
|
|
ruleName: "argdescriptorconstant"
|
|
functionName: "argdescriptorconstant"
|
|
inputBooleanName: "inPlace"
|
|
ruleType: "attribute"
|
|
transformerArgs {
|
|
key: "value"
|
|
transformerArgs {
|
|
name: "inPlace"
|
|
argType: BOOL
|
|
}
|
|
}
|
|
inputFrameworkOpName: "LRN"
|
|
}
|
|
}
|
|
mappings {
|
|
frameworkName: "onnx"
|
|
opName: "isinf"
|
|
inputFrameworkOpName: "IsInf"
|
|
rule {
|
|
ruleName: "ndarraymapping"
|
|
functionName: "ndarraymapping"
|
|
inputTensorName: "X"
|
|
outputTensorName: "input"
|
|
inputToOutput {
|
|
key: "input"
|
|
value: "X"
|
|
}
|
|
ruleType: "tensor"
|
|
inputFrameworkOpName: "IsInf"
|
|
}
|
|
rule {
|
|
ruleName: "argdescriptorconstant"
|
|
functionName: "argdescriptorconstant"
|
|
inputBooleanName: "inPlace"
|
|
ruleType: "attribute"
|
|
transformerArgs {
|
|
key: "value"
|
|
transformerArgs {
|
|
name: "inPlace"
|
|
argType: BOOL
|
|
}
|
|
}
|
|
inputFrameworkOpName: "IsInf"
|
|
}
|
|
}
|
|
mappings {
|
|
frameworkName: "onnx"
|
|
opName: "batchnorm"
|
|
inputFrameworkOpName: "BatchNormalization"
|
|
rule {
|
|
ruleName: "ndarraymapping"
|
|
functionName: "ndarraymapping"
|
|
inputTensorName: "X"
|
|
inputTensorName: "mean"
|
|
inputTensorName: "var"
|
|
inputTensorName: "scale"
|
|
outputTensorName: "input"
|
|
outputTensorName: "mean"
|
|
outputTensorName: "variance"
|
|
outputTensorName: "gamma"
|
|
inputToOutput {
|
|
key: "input"
|
|
value: "X"
|
|
}
|
|
inputToOutput {
|
|
key: "mean"
|
|
value: "mean"
|
|
}
|
|
inputToOutput {
|
|
key: "variance"
|
|
value: "var"
|
|
}
|
|
inputToOutput {
|
|
key: "gamma"
|
|
value: "scale"
|
|
}
|
|
ruleType: "tensor"
|
|
inputFrameworkOpName: "BatchNormalization"
|
|
}
|
|
rule {
|
|
ruleName: "valuemapping"
|
|
functionName: "valuemapping"
|
|
inputFloatName: "epsilon"
|
|
outputDoubleName: "epsilon"
|
|
inputToOutput {
|
|
key: "epsilon"
|
|
value: "epsilon"
|
|
}
|
|
ruleType: "attribute"
|
|
inputFrameworkOpName: "BatchNormalization"
|
|
}
|
|
rule {
|
|
ruleName: "argdescriptorconstant"
|
|
functionName: "argdescriptorconstant"
|
|
inputBooleanName: "inPlace"
|
|
ruleType: "attribute"
|
|
transformerArgs {
|
|
key: "value"
|
|
transformerArgs {
|
|
name: "inPlace"
|
|
argType: BOOL
|
|
}
|
|
}
|
|
inputFrameworkOpName: "BatchNormalization"
|
|
}
|
|
rule {
|
|
ruleName: "argdescriptorconstant"
|
|
functionName: "argdescriptorconstant"
|
|
inputBooleanName: "applyGamma"
|
|
ruleType: "attribute"
|
|
transformerArgs {
|
|
key: "value"
|
|
transformerArgs {
|
|
name: "applyGamma"
|
|
boolValue: true
|
|
argType: BOOL
|
|
argIndex: 1
|
|
}
|
|
}
|
|
inputFrameworkOpName: "BatchNormalization"
|
|
}
|
|
rule {
|
|
ruleName: "argdescriptorconstant"
|
|
functionName: "argdescriptorconstant"
|
|
inputBooleanName: "applyBeta"
|
|
ruleType: "attribute"
|
|
transformerArgs {
|
|
key: "value"
|
|
transformerArgs {
|
|
name: "applyBeta"
|
|
boolValue: true
|
|
argType: BOOL
|
|
argIndex: 2
|
|
}
|
|
}
|
|
inputFrameworkOpName: "BatchNormalization"
|
|
}
|
|
rule {
|
|
ruleName: "argdescriptorconstant"
|
|
functionName: "argdescriptorconstant"
|
|
inputIntName: "applyScale"
|
|
ruleType: "attribute"
|
|
transformerArgs {
|
|
key: "value"
|
|
transformerArgs {
|
|
name: "applyScale"
|
|
int64Value: 1
|
|
argType: INT64
|
|
}
|
|
}
|
|
inputFrameworkOpName: "BatchNormalization"
|
|
}
|
|
rule {
|
|
ruleName: "argdescriptorconstant"
|
|
functionName: "argdescriptorconstant"
|
|
inputIntName: "applyOffset"
|
|
ruleType: "attribute"
|
|
transformerArgs {
|
|
key: "value"
|
|
transformerArgs {
|
|
name: "applyOffset"
|
|
int64Value: 1
|
|
argType: INT64
|
|
argIndex: 1
|
|
}
|
|
}
|
|
inputFrameworkOpName: "BatchNormalization"
|
|
}
|
|
}
|
|
mappings {
|
|
frameworkName: "onnx"
|
|
opName: "elu"
|
|
inputFrameworkOpName: "Elu"
|
|
rule {
|
|
ruleName: "ndarraymapping"
|
|
functionName: "ndarraymapping"
|
|
inputTensorName: "X"
|
|
outputTensorName: "input"
|
|
inputToOutput {
|
|
key: "input"
|
|
value: "X"
|
|
}
|
|
ruleType: "tensor"
|
|
inputFrameworkOpName: "Elu"
|
|
}
|
|
rule {
|
|
ruleName: "valuemapping"
|
|
functionName: "valuemapping"
|
|
inputFloatName: "alpha"
|
|
outputDoubleName: "alpha"
|
|
inputToOutput {
|
|
key: "alpha"
|
|
value: "alpha"
|
|
}
|
|
ruleType: "attribute"
|
|
inputFrameworkOpName: "Elu"
|
|
}
|
|
}
|
|
mappings {
|
|
frameworkName: "onnx"
|
|
opName: "concat"
|
|
inputFrameworkOpName: "Concat"
|
|
rule {
|
|
ruleName: "ndarraymapping"
|
|
functionName: "ndarraymapping"
|
|
inputTensorName: "inputs"
|
|
outputTensorName: "input"
|
|
inputToOutput {
|
|
key: "input"
|
|
value: "inputs"
|
|
}
|
|
ruleType: "tensor"
|
|
inputFrameworkOpName: "Concat"
|
|
}
|
|
rule {
|
|
ruleName: "valuemapping"
|
|
functionName: "valuemapping"
|
|
inputIntName: "axis"
|
|
outputIntName: "concatDimension"
|
|
inputToOutput {
|
|
key: "concatDimension"
|
|
value: "axis"
|
|
}
|
|
ruleType: "attribute"
|
|
inputFrameworkOpName: "Concat"
|
|
}
|
|
rule {
|
|
ruleName: "argdescriptorconstant"
|
|
functionName: "argdescriptorconstant"
|
|
inputBooleanName: "isDynamicAxis"
|
|
ruleType: "attribute"
|
|
transformerArgs {
|
|
key: "value"
|
|
transformerArgs {
|
|
name: "isDynamicAxis"
|
|
argType: BOOL
|
|
}
|
|
}
|
|
inputFrameworkOpName: "Concat"
|
|
}
|
|
}
|
|
mappings {
|
|
frameworkName: "onnx"
|
|
opName: "top_k"
|
|
inputFrameworkOpName: "TopK"
|
|
rule {
|
|
ruleName: "ndarraymapping"
|
|
functionName: "ndarraymapping"
|
|
inputTensorName: "X"
|
|
outputTensorName: "input"
|
|
inputToOutput {
|
|
key: "input"
|
|
value: "X"
|
|
}
|
|
ruleType: "tensor"
|
|
inputFrameworkOpName: "TopK"
|
|
}
|
|
rule {
|
|
ruleName: "invertbooleannumber"
|
|
functionName: "invertbooleannumber"
|
|
inputIntName: "sorted"
|
|
outputBooleanName: "needSort"
|
|
inputToOutput {
|
|
key: "needSort"
|
|
value: "sorted"
|
|
}
|
|
ruleType: "attribute"
|
|
inputFrameworkOpName: "TopK"
|
|
}
|
|
rule {
|
|
ruleName: "ndarrayinputtonumericalattribute"
|
|
functionName: "ndarrayinputtonumericalattribute"
|
|
outputIntName: "k"
|
|
inputToOutput {
|
|
key: "k"
|
|
value: "K"
|
|
}
|
|
ruleType: "attribute"
|
|
inputFrameworkOpName: "TopK"
|
|
}
|
|
}
|
|
mappings {
|
|
frameworkName: "onnx"
|
|
opName: "equals"
|
|
inputFrameworkOpName: "Equal"
|
|
rule {
|
|
ruleName: "ndarraymapping"
|
|
functionName: "ndarraymapping"
|
|
inputTensorName: "A"
|
|
inputTensorName: "B"
|
|
outputTensorName: "input"
|
|
outputTensorName: "y"
|
|
inputToOutput {
|
|
key: "input"
|
|
value: "A"
|
|
}
|
|
inputToOutput {
|
|
key: "y"
|
|
value: "B"
|
|
}
|
|
ruleType: "tensor"
|
|
inputFrameworkOpName: "Equal"
|
|
}
|
|
rule {
|
|
ruleName: "argdescriptorconstant"
|
|
functionName: "argdescriptorconstant"
|
|
inputBooleanName: "inPlace"
|
|
ruleType: "attribute"
|
|
transformerArgs {
|
|
key: "value"
|
|
transformerArgs {
|
|
name: "inPlace"
|
|
argType: BOOL
|
|
}
|
|
}
|
|
inputFrameworkOpName: "Equal"
|
|
}
|
|
}
|
|
mappings {
|
|
frameworkName: "onnx"
|
|
opName: "matmul"
|
|
inputFrameworkOpName: "MatMul"
|
|
rule {
|
|
ruleName: "ndarraymapping"
|
|
functionName: "ndarraymapping"
|
|
inputTensorName: "A"
|
|
inputTensorName: "B"
|
|
outputTensorName: "input"
|
|
outputTensorName: "y"
|
|
inputToOutput {
|
|
key: "input"
|
|
value: "A"
|
|
}
|
|
inputToOutput {
|
|
key: "y"
|
|
value: "B"
|
|
}
|
|
ruleType: "tensor"
|
|
inputFrameworkOpName: "MatMul"
|
|
}
|
|
rule {
|
|
ruleName: "argdescriptorconstant"
|
|
functionName: "argdescriptorconstant"
|
|
inputBooleanName: "transposeX"
|
|
ruleType: "attribute"
|
|
transformerArgs {
|
|
key: "value"
|
|
transformerArgs {
|
|
name: "transposeX"
|
|
argType: BOOL
|
|
}
|
|
}
|
|
inputFrameworkOpName: "MatMul"
|
|
}
|
|
rule {
|
|
ruleName: "argdescriptorconstant"
|
|
functionName: "argdescriptorconstant"
|
|
inputBooleanName: "transposeY"
|
|
ruleType: "attribute"
|
|
transformerArgs {
|
|
key: "value"
|
|
transformerArgs {
|
|
name: "transposeY"
|
|
argType: BOOL
|
|
argIndex: 1
|
|
}
|
|
}
|
|
inputFrameworkOpName: "MatMul"
|
|
}
|
|
rule {
|
|
ruleName: "argdescriptorconstant"
|
|
functionName: "argdescriptorconstant"
|
|
inputBooleanName: "transposeZ"
|
|
ruleType: "attribute"
|
|
transformerArgs {
|
|
key: "value"
|
|
transformerArgs {
|
|
name: "transposeZ"
|
|
argType: BOOL
|
|
argIndex: 2
|
|
}
|
|
}
|
|
inputFrameworkOpName: "MatMul"
|
|
}
|
|
rule {
|
|
ruleName: "argdescriptorconstant"
|
|
functionName: "argdescriptorconstant"
|
|
inputFloatName: "alpha"
|
|
ruleType: "attribute"
|
|
transformerArgs {
|
|
key: "value"
|
|
transformerArgs {
|
|
name: "alpha"
|
|
argType: DOUBLE
|
|
}
|
|
}
|
|
inputFrameworkOpName: "MatMul"
|
|
}
|
|
rule {
|
|
ruleName: "argdescriptorconstant"
|
|
functionName: "argdescriptorconstant"
|
|
inputFloatName: "beta"
|
|
ruleType: "attribute"
|
|
transformerArgs {
|
|
key: "value"
|
|
transformerArgs {
|
|
name: "beta"
|
|
doubleValue: 1.0
|
|
argType: DOUBLE
|
|
argIndex: 1
|
|
}
|
|
}
|
|
inputFrameworkOpName: "MatMul"
|
|
}
|
|
}
|
|
mappings {
|
|
frameworkName: "onnx"
|
|
opName: "reduce_min"
|
|
inputFrameworkOpName: "ReduceMin"
|
|
rule {
|
|
ruleName: "ndarraymapping"
|
|
functionName: "ndarraymapping"
|
|
inputTensorName: "data"
|
|
outputTensorName: "input"
|
|
inputToOutput {
|
|
key: "input"
|
|
value: "data"
|
|
}
|
|
ruleType: "tensor"
|
|
inputFrameworkOpName: "ReduceMin"
|
|
}
|
|
rule {
|
|
ruleName: "invertbooleannumber"
|
|
functionName: "invertbooleannumber"
|
|
inputIntName: "keepdims"
|
|
outputBooleanName: "keepDims"
|
|
inputToOutput {
|
|
key: "keepDims"
|
|
value: "keepdims"
|
|
}
|
|
ruleType: "attribute"
|
|
inputFrameworkOpName: "ReduceMin"
|
|
}
|
|
rule {
|
|
ruleName: "listnumbertolistnumber"
|
|
functionName: "listnumbertolistnumber"
|
|
outputIntName: "dimensions"
|
|
inputToOutput {
|
|
key: "dimensions"
|
|
value: "axes"
|
|
}
|
|
ruleType: "attribute"
|
|
inputFrameworkOpName: "ReduceMin"
|
|
}
|
|
}
|
|
mappings {
|
|
frameworkName: "onnx"
|
|
opName: "sinh"
|
|
inputFrameworkOpName: "Sinh"
|
|
rule {
|
|
ruleName: "ndarraymapping"
|
|
functionName: "ndarraymapping"
|
|
inputTensorName: "input"
|
|
outputTensorName: "input"
|
|
inputToOutput {
|
|
key: "input"
|
|
value: "input"
|
|
}
|
|
ruleType: "tensor"
|
|
inputFrameworkOpName: "Sinh"
|
|
}
|
|
rule {
|
|
ruleName: "argdescriptorconstant"
|
|
functionName: "argdescriptorconstant"
|
|
inputBooleanName: "inPlace"
|
|
ruleType: "attribute"
|
|
transformerArgs {
|
|
key: "value"
|
|
transformerArgs {
|
|
name: "inPlace"
|
|
argType: BOOL
|
|
}
|
|
}
|
|
inputFrameworkOpName: "Sinh"
|
|
}
|
|
}
|
|
mappings {
|
|
frameworkName: "onnx"
|
|
opName: "asinh"
|
|
inputFrameworkOpName: "Asinh"
|
|
rule {
|
|
ruleName: "ndarraymapping"
|
|
functionName: "ndarraymapping"
|
|
inputTensorName: "input"
|
|
outputTensorName: "input"
|
|
inputToOutput {
|
|
key: "input"
|
|
value: "input"
|
|
}
|
|
ruleType: "tensor"
|
|
inputFrameworkOpName: "Asinh"
|
|
}
|
|
rule {
|
|
ruleName: "argdescriptorconstant"
|
|
functionName: "argdescriptorconstant"
|
|
inputBooleanName: "inPlace"
|
|
ruleType: "attribute"
|
|
transformerArgs {
|
|
key: "value"
|
|
transformerArgs {
|
|
name: "inPlace"
|
|
argType: BOOL
|
|
}
|
|
}
|
|
inputFrameworkOpName: "Asinh"
|
|
}
|
|
}
|
|
mappings {
|
|
frameworkName: "onnx"
|
|
opName: "gather_nd"
|
|
inputFrameworkOpName: "GatherND"
|
|
rule {
|
|
ruleName: "ndarraymapping"
|
|
functionName: "ndarraymapping"
|
|
inputTensorName: "indices"
|
|
inputTensorName: "data"
|
|
outputTensorName: "indices"
|
|
outputTensorName: "input"
|
|
inputToOutput {
|
|
key: "indices"
|
|
value: "indices"
|
|
}
|
|
inputToOutput {
|
|
key: "input"
|
|
value: "data"
|
|
}
|
|
ruleType: "tensor"
|
|
inputFrameworkOpName: "GatherND"
|
|
}
|
|
rule {
|
|
ruleName: "argdescriptorconstant"
|
|
functionName: "argdescriptorconstant"
|
|
inputBooleanName: "checkIndices"
|
|
ruleType: "attribute"
|
|
transformerArgs {
|
|
key: "value"
|
|
transformerArgs {
|
|
name: "checkIndices"
|
|
boolValue: true
|
|
argType: BOOL
|
|
}
|
|
}
|
|
inputFrameworkOpName: "GatherND"
|
|
}
|
|
}
|
|
mappings {
|
|
frameworkName: "onnx"
|
|
opName: "squeeze"
|
|
inputFrameworkOpName: "Squeeze"
|
|
rule {
|
|
ruleName: "ndarraymapping"
|
|
functionName: "ndarraymapping"
|
|
inputTensorName: "data"
|
|
outputTensorName: "input"
|
|
inputToOutput {
|
|
key: "input"
|
|
value: "data"
|
|
}
|
|
ruleType: "tensor"
|
|
inputFrameworkOpName: "Squeeze"
|
|
}
|
|
rule {
|
|
ruleName: "convertinputnumberlisttondarray"
|
|
functionName: "convertinputnumberlisttondarray"
|
|
inputToOutput {
|
|
key: "a"
|
|
value: "axes"
|
|
}
|
|
ruleType: "attribute"
|
|
inputFrameworkOpName: "Squeeze"
|
|
}
|
|
rule {
|
|
ruleName: "listnumbertolistnumber"
|
|
functionName: "listnumbertolistnumber"
|
|
outputIntName: "_a"
|
|
inputToOutput {
|
|
key: "_a"
|
|
value: "axes"
|
|
}
|
|
ruleType: "attribute"
|
|
inputFrameworkOpName: "Squeeze"
|
|
}
|
|
}
|
|
mappings {
|
|
frameworkName: "onnx"
|
|
opName: "identity"
|
|
inputFrameworkOpName: "Identity"
|
|
rule {
|
|
ruleName: "ndarraymapping"
|
|
functionName: "ndarraymapping"
|
|
inputTensorName: "input"
|
|
outputTensorName: "input"
|
|
inputToOutput {
|
|
key: "input"
|
|
value: "input"
|
|
}
|
|
ruleType: "tensor"
|
|
inputFrameworkOpName: "Identity"
|
|
}
|
|
rule {
|
|
ruleName: "argdescriptorconstant"
|
|
functionName: "argdescriptorconstant"
|
|
inputBooleanName: "inPlace"
|
|
ruleType: "attribute"
|
|
transformerArgs {
|
|
key: "value"
|
|
transformerArgs {
|
|
name: "inPlace"
|
|
argType: BOOL
|
|
}
|
|
}
|
|
inputFrameworkOpName: "Identity"
|
|
}
|
|
}
|
|
mappings {
|
|
frameworkName: "onnx"
|
|
opName: "less"
|
|
inputFrameworkOpName: "Less"
|
|
rule {
|
|
ruleName: "ndarraymapping"
|
|
functionName: "ndarraymapping"
|
|
inputTensorName: "A"
|
|
inputTensorName: "B"
|
|
outputTensorName: "input"
|
|
outputTensorName: "y"
|
|
inputToOutput {
|
|
key: "input"
|
|
value: "A"
|
|
}
|
|
inputToOutput {
|
|
key: "y"
|
|
value: "B"
|
|
}
|
|
ruleType: "tensor"
|
|
inputFrameworkOpName: "Less"
|
|
}
|
|
rule {
|
|
ruleName: "argdescriptorconstant"
|
|
functionName: "argdescriptorconstant"
|
|
inputBooleanName: "inPlace"
|
|
ruleType: "attribute"
|
|
transformerArgs {
|
|
key: "value"
|
|
transformerArgs {
|
|
name: "inPlace"
|
|
argType: BOOL
|
|
}
|
|
}
|
|
inputFrameworkOpName: "Less"
|
|
}
|
|
}
|
|
mappings {
|
|
frameworkName: "onnx"
|
|
opName: "softplus"
|
|
inputFrameworkOpName: "Softplus"
|
|
rule {
|
|
ruleName: "ndarraymapping"
|
|
functionName: "ndarraymapping"
|
|
inputTensorName: "X"
|
|
outputTensorName: "input"
|
|
inputToOutput {
|
|
key: "input"
|
|
value: "X"
|
|
}
|
|
ruleType: "tensor"
|
|
inputFrameworkOpName: "Softplus"
|
|
}
|
|
rule {
|
|
ruleName: "argdescriptorconstant"
|
|
functionName: "argdescriptorconstant"
|
|
inputBooleanName: "inPlace"
|
|
ruleType: "attribute"
|
|
transformerArgs {
|
|
key: "value"
|
|
transformerArgs {
|
|
name: "inPlace"
|
|
argType: BOOL
|
|
}
|
|
}
|
|
inputFrameworkOpName: "Softplus"
|
|
}
|
|
}
|
|
mappings {
|
|
frameworkName: "onnx"
|
|
opName: "reduce_sum"
|
|
inputFrameworkOpName: "ReduceSum"
|
|
rule {
|
|
ruleName: "ndarraymapping"
|
|
functionName: "ndarraymapping"
|
|
inputTensorName: "data"
|
|
outputTensorName: "input"
|
|
inputToOutput {
|
|
key: "input"
|
|
value: "data"
|
|
}
|
|
ruleType: "tensor"
|
|
inputFrameworkOpName: "ReduceSum"
|
|
}
|
|
rule {
|
|
ruleName: "invertbooleannumber"
|
|
functionName: "invertbooleannumber"
|
|
inputIntName: "keepdims"
|
|
outputBooleanName: "keepDims"
|
|
inputToOutput {
|
|
key: "keepDims"
|
|
value: "keepdims"
|
|
}
|
|
ruleType: "attribute"
|
|
inputFrameworkOpName: "ReduceSum"
|
|
}
|
|
rule {
|
|
ruleName: "listnumbertolistnumber"
|
|
functionName: "listnumbertolistnumber"
|
|
outputIntName: "dimensions"
|
|
inputToOutput {
|
|
key: "dimensions"
|
|
value: "axes"
|
|
}
|
|
ruleType: "attribute"
|
|
inputFrameworkOpName: "ReduceSum"
|
|
}
|
|
}
|
|
mappings {
|
|
frameworkName: "onnx"
|
|
opName: "tanh"
|
|
inputFrameworkOpName: "Tanh"
|
|
rule {
|
|
ruleName: "ndarraymapping"
|
|
functionName: "ndarraymapping"
|
|
inputTensorName: "input"
|
|
outputTensorName: "input"
|
|
inputToOutput {
|
|
key: "input"
|
|
value: "input"
|
|
}
|
|
ruleType: "tensor"
|
|
inputFrameworkOpName: "Tanh"
|
|
}
|
|
rule {
|
|
ruleName: "argdescriptorconstant"
|
|
functionName: "argdescriptorconstant"
|
|
inputBooleanName: "inPlace"
|
|
ruleType: "attribute"
|
|
transformerArgs {
|
|
key: "value"
|
|
transformerArgs {
|
|
name: "inPlace"
|
|
argType: BOOL
|
|
}
|
|
}
|
|
inputFrameworkOpName: "Tanh"
|
|
}
|
|
}
|
|
mappings {
|
|
frameworkName: "onnx"
|
|
opName: "subtract"
|
|
inputFrameworkOpName: "Sub"
|
|
rule {
|
|
ruleName: "ndarraymapping"
|
|
functionName: "ndarraymapping"
|
|
inputTensorName: "A"
|
|
inputTensorName: "B"
|
|
outputTensorName: "input"
|
|
outputTensorName: "y"
|
|
inputToOutput {
|
|
key: "input"
|
|
value: "A"
|
|
}
|
|
inputToOutput {
|
|
key: "y"
|
|
value: "B"
|
|
}
|
|
ruleType: "tensor"
|
|
inputFrameworkOpName: "Sub"
|
|
}
|
|
rule {
|
|
ruleName: "argdescriptorconstant"
|
|
functionName: "argdescriptorconstant"
|
|
inputBooleanName: "inPlace"
|
|
ruleType: "attribute"
|
|
transformerArgs {
|
|
key: "value"
|
|
transformerArgs {
|
|
name: "inPlace"
|
|
argType: BOOL
|
|
}
|
|
}
|
|
inputFrameworkOpName: "Sub"
|
|
}
|
|
}
|
|
mappings {
|
|
frameworkName: "onnx"
|
|
opName: "reduce_prod"
|
|
inputFrameworkOpName: "ReduceProd"
|
|
rule {
|
|
ruleName: "ndarraymapping"
|
|
functionName: "ndarraymapping"
|
|
inputTensorName: "data"
|
|
outputTensorName: "input"
|
|
inputToOutput {
|
|
key: "input"
|
|
value: "data"
|
|
}
|
|
ruleType: "tensor"
|
|
inputFrameworkOpName: "ReduceProd"
|
|
}
|
|
rule {
|
|
ruleName: "invertbooleannumber"
|
|
functionName: "invertbooleannumber"
|
|
inputIntName: "keepdims"
|
|
outputBooleanName: "keepDims"
|
|
inputToOutput {
|
|
key: "keepDims"
|
|
value: "keepdims"
|
|
}
|
|
ruleType: "attribute"
|
|
inputFrameworkOpName: "ReduceProd"
|
|
}
|
|
rule {
|
|
ruleName: "listnumbertolistnumber"
|
|
functionName: "listnumbertolistnumber"
|
|
outputIntName: "dimensions"
|
|
inputToOutput {
|
|
key: "dimensions"
|
|
value: "axes"
|
|
}
|
|
ruleType: "attribute"
|
|
inputFrameworkOpName: "ReduceProd"
|
|
}
|
|
}
|
|
mappings {
|
|
frameworkName: "onnx"
|
|
opName: "multiply"
|
|
inputFrameworkOpName: "Mul"
|
|
rule {
|
|
ruleName: "ndarraymapping"
|
|
functionName: "ndarraymapping"
|
|
inputTensorName: "A"
|
|
inputTensorName: "B"
|
|
outputTensorName: "input"
|
|
outputTensorName: "y"
|
|
inputToOutput {
|
|
key: "input"
|
|
value: "A"
|
|
}
|
|
inputToOutput {
|
|
key: "y"
|
|
value: "B"
|
|
}
|
|
ruleType: "tensor"
|
|
inputFrameworkOpName: "Mul"
|
|
}
|
|
rule {
|
|
ruleName: "argdescriptorconstant"
|
|
functionName: "argdescriptorconstant"
|
|
inputBooleanName: "inPlace"
|
|
ruleType: "attribute"
|
|
transformerArgs {
|
|
key: "value"
|
|
transformerArgs {
|
|
name: "inPlace"
|
|
argType: BOOL
|
|
}
|
|
}
|
|
inputFrameworkOpName: "Mul"
|
|
}
|
|
}
|
|
mappings {
|
|
frameworkName: "onnx"
|
|
opName: "log"
|
|
inputFrameworkOpName: "Log"
|
|
rule {
|
|
ruleName: "ndarraymapping"
|
|
functionName: "ndarraymapping"
|
|
inputTensorName: "input"
|
|
outputTensorName: "input"
|
|
inputToOutput {
|
|
key: "input"
|
|
value: "input"
|
|
}
|
|
ruleType: "tensor"
|
|
inputFrameworkOpName: "Log"
|
|
}
|
|
rule {
|
|
ruleName: "argdescriptorconstant"
|
|
functionName: "argdescriptorconstant"
|
|
inputBooleanName: "inPlace"
|
|
ruleType: "attribute"
|
|
transformerArgs {
|
|
key: "value"
|
|
transformerArgs {
|
|
name: "inPlace"
|
|
argType: BOOL
|
|
}
|
|
}
|
|
inputFrameworkOpName: "Log"
|
|
}
|
|
}
|
|
mappings {
|
|
frameworkName: "onnx"
|
|
opName: "flatten_2d"
|
|
inputFrameworkOpName: "Flatten"
|
|
rule {
|
|
ruleName: "ndarraymapping"
|
|
functionName: "ndarraymapping"
|
|
inputTensorName: "input"
|
|
outputTensorName: "input"
|
|
inputToOutput {
|
|
key: "input"
|
|
value: "input"
|
|
}
|
|
ruleType: "tensor"
|
|
inputFrameworkOpName: "Flatten"
|
|
}
|
|
rule {
|
|
ruleName: "valuemapping"
|
|
functionName: "valuemapping"
|
|
inputIntName: "axis"
|
|
outputIntName: "flattenDimension"
|
|
inputToOutput {
|
|
key: "flattenDimension"
|
|
value: "axis"
|
|
}
|
|
ruleType: "attribute"
|
|
inputFrameworkOpName: "Flatten"
|
|
}
|
|
}
|
|
mappings {
|
|
frameworkName: "onnx"
|
|
opName: "range"
|
|
inputFrameworkOpName: "Range"
|
|
rule {
|
|
ruleName: "ndarraymapping"
|
|
functionName: "ndarraymapping"
|
|
inputTensorName: "start"
|
|
inputTensorName: "limit"
|
|
inputTensorName: "delta"
|
|
outputTensorName: "from"
|
|
outputTensorName: "to"
|
|
outputTensorName: "step"
|
|
inputToOutput {
|
|
key: "from"
|
|
value: "start"
|
|
}
|
|
inputToOutput {
|
|
key: "to"
|
|
value: "limit"
|
|
}
|
|
inputToOutput {
|
|
key: "step"
|
|
value: "delta"
|
|
}
|
|
ruleType: "tensor"
|
|
inputFrameworkOpName: "Range"
|
|
}
|
|
rule {
|
|
ruleName: "ndarrayinputtonumericalattribute"
|
|
functionName: "ndarrayinputtonumericalattribute"
|
|
outputIntName: "from"
|
|
inputToOutput {
|
|
key: "from"
|
|
value: "start"
|
|
}
|
|
ruleType: "attribute"
|
|
inputFrameworkOpName: "Range"
|
|
}
|
|
rule {
|
|
ruleName: "ndarrayinputtonumericalattribute"
|
|
functionName: "ndarrayinputtonumericalattribute"
|
|
outputIntName: "to"
|
|
inputToOutput {
|
|
key: "to"
|
|
value: "limit"
|
|
}
|
|
ruleType: "attribute"
|
|
inputFrameworkOpName: "Range"
|
|
}
|
|
rule {
|
|
ruleName: "ndarrayinputtonumericalattribute"
|
|
functionName: "ndarrayinputtonumericalattribute"
|
|
outputIntName: "step"
|
|
inputToOutput {
|
|
key: "step"
|
|
value: "delta"
|
|
}
|
|
ruleType: "attribute"
|
|
inputFrameworkOpName: "Range"
|
|
}
|
|
}
|
|
mappings {
|
|
frameworkName: "onnx"
|
|
opName: "transpose"
|
|
inputFrameworkOpName: "Transpose"
|
|
rule {
|
|
ruleName: "ndarraymapping"
|
|
functionName: "ndarraymapping"
|
|
inputTensorName: "data"
|
|
outputTensorName: "input"
|
|
inputToOutput {
|
|
key: "input"
|
|
value: "data"
|
|
}
|
|
ruleType: "tensor"
|
|
inputFrameworkOpName: "Transpose"
|
|
}
|
|
rule {
|
|
ruleName: "listnumbertondarray"
|
|
functionName: "listnumbertondarray"
|
|
inputToOutput {
|
|
key: "permuteDims"
|
|
value: "perm"
|
|
}
|
|
ruleType: "attribute"
|
|
inputFrameworkOpName: "Transpose"
|
|
}
|
|
}
|
|
mappings {
|
|
frameworkName: "onnx"
|
|
opName: "gather"
|
|
inputFrameworkOpName: "Gather"
|
|
rule {
|
|
ruleName: "ndarraymapping"
|
|
functionName: "ndarraymapping"
|
|
inputTensorName: "indices"
|
|
inputTensorName: "data"
|
|
outputTensorName: "indices"
|
|
outputTensorName: "input"
|
|
inputToOutput {
|
|
key: "indices"
|
|
value: "indices"
|
|
}
|
|
inputToOutput {
|
|
key: "input"
|
|
value: "data"
|
|
}
|
|
ruleType: "tensor"
|
|
inputFrameworkOpName: "Gather"
|
|
}
|
|
rule {
|
|
ruleName: "valuemapping"
|
|
functionName: "valuemapping"
|
|
inputIntName: "axis"
|
|
outputIntName: "dimensions"
|
|
inputToOutput {
|
|
key: "dimensions"
|
|
value: "axis"
|
|
}
|
|
ruleType: "attribute"
|
|
inputFrameworkOpName: "Gather"
|
|
}
|
|
rule {
|
|
ruleName: "argdescriptorconstant"
|
|
functionName: "argdescriptorconstant"
|
|
inputBooleanName: "inPlace"
|
|
ruleType: "attribute"
|
|
transformerArgs {
|
|
key: "value"
|
|
transformerArgs {
|
|
name: "inPlace"
|
|
argType: BOOL
|
|
}
|
|
}
|
|
inputFrameworkOpName: "Gather"
|
|
}
|
|
}
|
|
mappings {
|
|
frameworkName: "onnx"
|
|
opName: "argmax"
|
|
inputFrameworkOpName: "ArgMax"
|
|
rule {
|
|
ruleName: "ndarraymapping"
|
|
functionName: "ndarraymapping"
|
|
inputTensorName: "data"
|
|
outputTensorName: "input"
|
|
inputToOutput {
|
|
key: "input"
|
|
value: "data"
|
|
}
|
|
ruleType: "tensor"
|
|
inputFrameworkOpName: "ArgMax"
|
|
}
|
|
rule {
|
|
ruleName: "invertbooleannumber"
|
|
functionName: "invertbooleannumber"
|
|
inputIntName: "keepdims"
|
|
outputBooleanName: "keepDims"
|
|
inputToOutput {
|
|
key: "keepDims"
|
|
value: "keepdims"
|
|
}
|
|
ruleType: "attribute"
|
|
inputFrameworkOpName: "ArgMax"
|
|
}
|
|
rule {
|
|
ruleName: "valuemapping"
|
|
functionName: "valuemapping"
|
|
inputIntName: "axis"
|
|
outputIntName: "dimensions"
|
|
inputToOutput {
|
|
key: "dimensions"
|
|
value: "axis"
|
|
}
|
|
ruleType: "attribute"
|
|
inputFrameworkOpName: "ArgMax"
|
|
}
|
|
}
|
|
mappings {
|
|
frameworkName: "onnx"
|
|
opName: "not"
|
|
inputFrameworkOpName: "Not"
|
|
rule {
|
|
ruleName: "ndarraymapping"
|
|
functionName: "ndarraymapping"
|
|
inputTensorName: "X"
|
|
outputTensorName: "input"
|
|
inputToOutput {
|
|
key: "input"
|
|
value: "X"
|
|
}
|
|
ruleType: "tensor"
|
|
inputFrameworkOpName: "Not"
|
|
}
|
|
rule {
|
|
ruleName: "argdescriptorconstant"
|
|
functionName: "argdescriptorconstant"
|
|
inputFloatName: "comparable"
|
|
ruleType: "attribute"
|
|
transformerArgs {
|
|
key: "value"
|
|
transformerArgs {
|
|
name: "comparable"
|
|
argType: DOUBLE
|
|
}
|
|
}
|
|
inputFrameworkOpName: "Not"
|
|
}
|
|
}
|
|
mappings {
|
|
frameworkName: "onnx"
|
|
opName: "reduce_mean"
|
|
inputFrameworkOpName: "ReduceMean"
|
|
rule {
|
|
ruleName: "ndarraymapping"
|
|
functionName: "ndarraymapping"
|
|
inputTensorName: "data"
|
|
outputTensorName: "input"
|
|
inputToOutput {
|
|
key: "input"
|
|
value: "data"
|
|
}
|
|
ruleType: "tensor"
|
|
inputFrameworkOpName: "ReduceMean"
|
|
}
|
|
rule {
|
|
ruleName: "invertbooleannumber"
|
|
functionName: "invertbooleannumber"
|
|
inputIntName: "keepdims"
|
|
outputBooleanName: "keepDims"
|
|
inputToOutput {
|
|
key: "keepDims"
|
|
value: "keepdims"
|
|
}
|
|
ruleType: "attribute"
|
|
inputFrameworkOpName: "ReduceMean"
|
|
}
|
|
rule {
|
|
ruleName: "listnumbertolistnumber"
|
|
functionName: "listnumbertolistnumber"
|
|
outputIntName: "dimensions"
|
|
inputToOutput {
|
|
key: "dimensions"
|
|
value: "axes"
|
|
}
|
|
ruleType: "attribute"
|
|
inputFrameworkOpName: "ReduceMean"
|
|
}
|
|
}
|
|
mappings {
|
|
frameworkName: "onnx"
|
|
opName: "reshape"
|
|
inputFrameworkOpName: "Reshape"
|
|
rule {
|
|
ruleName: "ndarraymapping"
|
|
functionName: "ndarraymapping"
|
|
inputTensorName: "data"
|
|
inputTensorName: "shape"
|
|
outputTensorName: "input"
|
|
outputTensorName: "shape"
|
|
inputToOutput {
|
|
key: "input"
|
|
value: "data"
|
|
}
|
|
inputToOutput {
|
|
key: "shape"
|
|
value: "shape"
|
|
}
|
|
ruleType: "tensor"
|
|
inputFrameworkOpName: "Reshape"
|
|
}
|
|
rule {
|
|
ruleName: "ndarraytointattributevalue"
|
|
functionName: "ndarraytointattributevalue"
|
|
outputIntName: "shapeArr"
|
|
inputToOutput {
|
|
key: "shapeArr"
|
|
value: "shape"
|
|
}
|
|
ruleType: "attribute"
|
|
inputFrameworkOpName: "Reshape"
|
|
}
|
|
}
|
|
mappings {
|
|
frameworkName: "onnx"
|
|
opName: "randomuniform"
|
|
inputFrameworkOpName: "RandomUniform"
|
|
rule {
|
|
ruleName: "valuemapping"
|
|
functionName: "valuemapping"
|
|
inputFloatName: "low"
|
|
inputFloatName: "high"
|
|
outputDoubleName: "min"
|
|
outputDoubleName: "max"
|
|
inputToOutput {
|
|
key: "min"
|
|
value: "low"
|
|
}
|
|
inputToOutput {
|
|
key: "max"
|
|
value: "high"
|
|
}
|
|
ruleType: "attribute"
|
|
inputFrameworkOpName: "RandomUniform"
|
|
}
|
|
rule {
|
|
ruleName: "listnumbertondarray"
|
|
functionName: "listnumbertondarray"
|
|
inputToOutput {
|
|
key: "shape"
|
|
value: "shape"
|
|
}
|
|
ruleType: "attribute"
|
|
inputFrameworkOpName: "RandomUniform"
|
|
}
|
|
}
|
|
mappings {
|
|
frameworkName: "onnx"
|
|
opName: "boolean_and"
|
|
inputFrameworkOpName: "And"
|
|
rule {
|
|
ruleName: "ndarraymapping"
|
|
functionName: "ndarraymapping"
|
|
inputTensorName: "A"
|
|
inputTensorName: "B"
|
|
outputTensorName: "input"
|
|
outputTensorName: "y"
|
|
inputToOutput {
|
|
key: "input"
|
|
value: "A"
|
|
}
|
|
inputToOutput {
|
|
key: "y"
|
|
value: "B"
|
|
}
|
|
ruleType: "tensor"
|
|
inputFrameworkOpName: "And"
|
|
}
|
|
}
|
|
mappings {
|
|
frameworkName: "onnx"
|
|
opName: "softmax"
|
|
inputFrameworkOpName: "Softmax"
|
|
rule {
|
|
ruleName: "ndarraymapping"
|
|
functionName: "ndarraymapping"
|
|
inputTensorName: "input"
|
|
outputTensorName: "input"
|
|
inputToOutput {
|
|
key: "input"
|
|
value: "input"
|
|
}
|
|
ruleType: "tensor"
|
|
inputFrameworkOpName: "Softmax"
|
|
}
|
|
rule {
|
|
ruleName: "valuemapping"
|
|
functionName: "valuemapping"
|
|
inputIntName: "axis"
|
|
outputIntName: "dimension"
|
|
inputToOutput {
|
|
key: "dimension"
|
|
value: "axis"
|
|
}
|
|
ruleType: "attribute"
|
|
inputFrameworkOpName: "Softmax"
|
|
}
|
|
rule {
|
|
ruleName: "argdescriptorconstant"
|
|
functionName: "argdescriptorconstant"
|
|
inputBooleanName: "inPlace"
|
|
ruleType: "attribute"
|
|
transformerArgs {
|
|
key: "value"
|
|
transformerArgs {
|
|
name: "inPlace"
|
|
argType: BOOL
|
|
}
|
|
}
|
|
inputFrameworkOpName: "Softmax"
|
|
}
|
|
}
|
|
mappings {
|
|
frameworkName: "onnx"
|
|
opName: "leakyrelu"
|
|
inputFrameworkOpName: "LeakyRelu"
|
|
rule {
|
|
ruleName: "ndarraymapping"
|
|
functionName: "ndarraymapping"
|
|
inputTensorName: "X"
|
|
outputTensorName: "input"
|
|
inputToOutput {
|
|
key: "input"
|
|
value: "X"
|
|
}
|
|
ruleType: "tensor"
|
|
inputFrameworkOpName: "LeakyRelu"
|
|
}
|
|
rule {
|
|
ruleName: "valuemapping"
|
|
functionName: "valuemapping"
|
|
inputFloatName: "alpha"
|
|
outputDoubleName: "alpha"
|
|
inputToOutput {
|
|
key: "alpha"
|
|
value: "alpha"
|
|
}
|
|
ruleType: "attribute"
|
|
inputFrameworkOpName: "LeakyRelu"
|
|
}
|
|
}
|
|
mappings {
|
|
frameworkName: "onnx"
|
|
opName: "erf"
|
|
inputFrameworkOpName: "Erf"
|
|
rule {
|
|
ruleName: "ndarraymapping"
|
|
functionName: "ndarraymapping"
|
|
inputTensorName: "input"
|
|
outputTensorName: "input"
|
|
inputToOutput {
|
|
key: "input"
|
|
value: "input"
|
|
}
|
|
ruleType: "tensor"
|
|
inputFrameworkOpName: "Erf"
|
|
}
|
|
rule {
|
|
ruleName: "argdescriptorconstant"
|
|
functionName: "argdescriptorconstant"
|
|
inputBooleanName: "inPlace"
|
|
ruleType: "attribute"
|
|
transformerArgs {
|
|
key: "value"
|
|
transformerArgs {
|
|
name: "inPlace"
|
|
argType: BOOL
|
|
}
|
|
}
|
|
inputFrameworkOpName: "Erf"
|
|
}
|
|
}
|
|
mappings {
|
|
frameworkName: "onnx"
|
|
opName: "pow_pairwise"
|
|
inputFrameworkOpName: "Pow"
|
|
rule {
|
|
ruleName: "ndarraymapping"
|
|
functionName: "ndarraymapping"
|
|
inputTensorName: "X"
|
|
inputTensorName: "Y"
|
|
outputTensorName: "input"
|
|
outputTensorName: "y"
|
|
inputToOutput {
|
|
key: "input"
|
|
value: "X"
|
|
}
|
|
inputToOutput {
|
|
key: "y"
|
|
value: "Y"
|
|
}
|
|
ruleType: "tensor"
|
|
inputFrameworkOpName: "Pow"
|
|
}
|
|
rule {
|
|
ruleName: "argdescriptorconstant"
|
|
functionName: "argdescriptorconstant"
|
|
inputBooleanName: "inPlace"
|
|
ruleType: "attribute"
|
|
transformerArgs {
|
|
key: "value"
|
|
transformerArgs {
|
|
name: "inPlace"
|
|
argType: BOOL
|
|
}
|
|
}
|
|
inputFrameworkOpName: "Pow"
|
|
}
|
|
}
|
|
mappings {
|
|
frameworkName: "onnx"
|
|
opName: "acos"
|
|
inputFrameworkOpName: "Acos"
|
|
rule {
|
|
ruleName: "ndarraymapping"
|
|
functionName: "ndarraymapping"
|
|
inputTensorName: "input"
|
|
outputTensorName: "input"
|
|
inputToOutput {
|
|
key: "input"
|
|
value: "input"
|
|
}
|
|
ruleType: "tensor"
|
|
inputFrameworkOpName: "Acos"
|
|
}
|
|
rule {
|
|
ruleName: "argdescriptorconstant"
|
|
functionName: "argdescriptorconstant"
|
|
inputBooleanName: "inPlace"
|
|
ruleType: "attribute"
|
|
transformerArgs {
|
|
key: "value"
|
|
transformerArgs {
|
|
name: "inPlace"
|
|
argType: BOOL
|
|
}
|
|
}
|
|
inputFrameworkOpName: "Acos"
|
|
}
|
|
}
|
|
mappings {
|
|
frameworkName: "onnx"
|
|
opName: "sin"
|
|
inputFrameworkOpName: "Sin"
|
|
rule {
|
|
ruleName: "ndarraymapping"
|
|
functionName: "ndarraymapping"
|
|
inputTensorName: "input"
|
|
outputTensorName: "input"
|
|
inputToOutput {
|
|
key: "input"
|
|
value: "input"
|
|
}
|
|
ruleType: "tensor"
|
|
inputFrameworkOpName: "Sin"
|
|
}
|
|
rule {
|
|
ruleName: "argdescriptorconstant"
|
|
functionName: "argdescriptorconstant"
|
|
inputBooleanName: "inPlace"
|
|
ruleType: "attribute"
|
|
transformerArgs {
|
|
key: "value"
|
|
transformerArgs {
|
|
name: "inPlace"
|
|
argType: BOOL
|
|
}
|
|
}
|
|
inputFrameworkOpName: "Sin"
|
|
}
|
|
}
|
|
mappings {
|
|
frameworkName: "onnx"
|
|
opName: "bitwise_xor"
|
|
inputFrameworkOpName: "Xor"
|
|
rule {
|
|
ruleName: "ndarraymapping"
|
|
functionName: "ndarraymapping"
|
|
inputTensorName: "A"
|
|
inputTensorName: "B"
|
|
outputTensorName: "input"
|
|
outputTensorName: "y"
|
|
inputToOutput {
|
|
key: "input"
|
|
value: "A"
|
|
}
|
|
inputToOutput {
|
|
key: "y"
|
|
value: "B"
|
|
}
|
|
ruleType: "tensor"
|
|
inputFrameworkOpName: "Xor"
|
|
}
|
|
rule {
|
|
ruleName: "argdescriptorconstant"
|
|
functionName: "argdescriptorconstant"
|
|
inputBooleanName: "inPlace"
|
|
ruleType: "attribute"
|
|
transformerArgs {
|
|
key: "value"
|
|
transformerArgs {
|
|
name: "inPlace"
|
|
argType: BOOL
|
|
}
|
|
}
|
|
inputFrameworkOpName: "Xor"
|
|
}
|
|
}
|
|
mappings {
|
|
frameworkName: "onnx"
|
|
opName: "ceil"
|
|
inputFrameworkOpName: "Ceil"
|
|
rule {
|
|
ruleName: "ndarraymapping"
|
|
functionName: "ndarraymapping"
|
|
inputTensorName: "X"
|
|
outputTensorName: "input"
|
|
inputToOutput {
|
|
key: "input"
|
|
value: "X"
|
|
}
|
|
ruleType: "tensor"
|
|
inputFrameworkOpName: "Ceil"
|
|
}
|
|
rule {
|
|
ruleName: "argdescriptorconstant"
|
|
functionName: "argdescriptorconstant"
|
|
inputBooleanName: "inPlace"
|
|
ruleType: "attribute"
|
|
transformerArgs {
|
|
key: "value"
|
|
transformerArgs {
|
|
name: "inPlace"
|
|
argType: BOOL
|
|
}
|
|
}
|
|
inputFrameworkOpName: "Ceil"
|
|
}
|
|
}
|
|
mappings {
|
|
frameworkName: "onnx"
|
|
opName: "relu"
|
|
inputFrameworkOpName: "Relu"
|
|
rule {
|
|
ruleName: "ndarraymapping"
|
|
functionName: "ndarraymapping"
|
|
inputTensorName: "X"
|
|
outputTensorName: "input"
|
|
inputToOutput {
|
|
key: "input"
|
|
value: "X"
|
|
}
|
|
ruleType: "tensor"
|
|
inputFrameworkOpName: "Relu"
|
|
}
|
|
rule {
|
|
ruleName: "argdescriptorconstant"
|
|
functionName: "argdescriptorconstant"
|
|
inputBooleanName: "inPlace"
|
|
ruleType: "attribute"
|
|
transformerArgs {
|
|
key: "value"
|
|
transformerArgs {
|
|
name: "inPlace"
|
|
argType: BOOL
|
|
}
|
|
}
|
|
inputFrameworkOpName: "Relu"
|
|
}
|
|
rule {
|
|
ruleName: "argdescriptorconstant"
|
|
functionName: "argdescriptorconstant"
|
|
inputFloatName: "cutoff"
|
|
ruleType: "attribute"
|
|
transformerArgs {
|
|
key: "value"
|
|
transformerArgs {
|
|
name: "cutoff"
|
|
argType: DOUBLE
|
|
}
|
|
}
|
|
inputFrameworkOpName: "Relu"
|
|
}
|
|
}
|
|
mappings {
|
|
frameworkName: "onnx"
|
|
opName: "split"
|
|
inputFrameworkOpName: "Split"
|
|
rule {
|
|
ruleName: "ndarraymapping"
|
|
functionName: "ndarraymapping"
|
|
inputTensorName: "input"
|
|
outputTensorName: "a"
|
|
inputToOutput {
|
|
key: "a"
|
|
value: "input"
|
|
}
|
|
ruleType: "tensor"
|
|
inputFrameworkOpName: "Split"
|
|
}
|
|
rule {
|
|
ruleName: "valuemapping"
|
|
functionName: "valuemapping"
|
|
inputIntName: "axis"
|
|
outputIntName: "dimensions"
|
|
inputToOutput {
|
|
key: "dimensions"
|
|
value: "axis"
|
|
}
|
|
ruleType: "attribute"
|
|
inputFrameworkOpName: "Split"
|
|
}
|
|
rule {
|
|
ruleName: "argdescriptorconstant"
|
|
functionName: "argdescriptorconstant"
|
|
inputIntName: "numSplit"
|
|
ruleType: "attribute"
|
|
transformerArgs {
|
|
key: "value"
|
|
transformerArgs {
|
|
name: "numSplit"
|
|
argType: INT64
|
|
}
|
|
}
|
|
inputFrameworkOpName: "Split"
|
|
}
|
|
rule {
|
|
ruleName: "listnumbertondarray"
|
|
functionName: "listnumbertondarray"
|
|
inputToOutput {
|
|
key: "b"
|
|
value: "split"
|
|
}
|
|
ruleType: "attribute"
|
|
inputFrameworkOpName: "Split"
|
|
}
|
|
}
|
|
mappings {
|
|
frameworkName: "onnx"
|
|
opName: "reduce_logsumexp"
|
|
inputFrameworkOpName: "ReduceLogSumExp"
|
|
rule {
|
|
ruleName: "ndarraymapping"
|
|
functionName: "ndarraymapping"
|
|
inputTensorName: "data"
|
|
outputTensorName: "input"
|
|
inputToOutput {
|
|
key: "input"
|
|
value: "data"
|
|
}
|
|
ruleType: "tensor"
|
|
inputFrameworkOpName: "ReduceLogSumExp"
|
|
}
|
|
rule {
|
|
ruleName: "invertbooleannumber"
|
|
functionName: "invertbooleannumber"
|
|
inputIntName: "keepdims"
|
|
outputDoubleName: "keepDims"
|
|
inputToOutput {
|
|
key: "keepDims"
|
|
value: "keepdims"
|
|
}
|
|
ruleType: "attribute"
|
|
inputFrameworkOpName: "ReduceLogSumExp"
|
|
}
|
|
rule {
|
|
ruleName: "valuemapping"
|
|
functionName: "valuemapping"
|
|
inputIntName: "keepdims"
|
|
outputDoubleName: "keepDims"
|
|
inputToOutput {
|
|
key: "keepDims"
|
|
value: "keepdims"
|
|
}
|
|
ruleType: "attribute"
|
|
inputFrameworkOpName: "ReduceLogSumExp"
|
|
}
|
|
rule {
|
|
ruleName: "listnumbertolistnumber"
|
|
functionName: "listnumbertolistnumber"
|
|
outputIntName: "dimensions"
|
|
inputToOutput {
|
|
key: "dimensions"
|
|
value: "axes"
|
|
}
|
|
ruleType: "attribute"
|
|
inputFrameworkOpName: "ReduceLogSumExp"
|
|
}
|
|
}
|
|
mappings {
|
|
frameworkName: "onnx"
|
|
opName: "matmul"
|
|
inputFrameworkOpName: "Gemm"
|
|
rule {
|
|
ruleName: "ndarraymapping"
|
|
functionName: "ndarraymapping"
|
|
inputTensorName: "A"
|
|
inputTensorName: "B"
|
|
outputTensorName: "input"
|
|
outputTensorName: "y"
|
|
inputToOutput {
|
|
key: "input"
|
|
value: "A"
|
|
}
|
|
inputToOutput {
|
|
key: "y"
|
|
value: "B"
|
|
}
|
|
ruleType: "tensor"
|
|
inputFrameworkOpName: "Gemm"
|
|
}
|
|
rule {
|
|
ruleName: "valuemapping"
|
|
functionName: "valuemapping"
|
|
inputIntName: "transA"
|
|
inputIntName: "transB"
|
|
inputFloatName: "alpha"
|
|
inputFloatName: "beta"
|
|
outputDoubleName: "alpha"
|
|
outputDoubleName: "beta"
|
|
outputBooleanName: "transposeX"
|
|
outputBooleanName: "transposeY"
|
|
inputToOutput {
|
|
key: "alpha"
|
|
value: "alpha"
|
|
}
|
|
inputToOutput {
|
|
key: "beta"
|
|
value: "beta"
|
|
}
|
|
inputToOutput {
|
|
key: "transposeX"
|
|
value: "transA"
|
|
}
|
|
inputToOutput {
|
|
key: "transposeY"
|
|
value: "transB"
|
|
}
|
|
ruleType: "attribute"
|
|
inputFrameworkOpName: "Gemm"
|
|
}
|
|
rule {
|
|
ruleName: "argdescriptorconstant"
|
|
functionName: "argdescriptorconstant"
|
|
inputBooleanName: "transZ"
|
|
ruleType: "attribute"
|
|
transformerArgs {
|
|
key: "value"
|
|
transformerArgs {
|
|
name: "transZ"
|
|
argType: BOOL
|
|
argIndex: 2
|
|
}
|
|
}
|
|
inputFrameworkOpName: "Gemm"
|
|
}
|
|
rule {
|
|
ruleName: "argdescriptorconstant"
|
|
functionName: "argdescriptorconstant"
|
|
inputBooleanName: "transposeZ"
|
|
ruleType: "attribute"
|
|
transformerArgs {
|
|
key: "value"
|
|
transformerArgs {
|
|
name: "transposeZ"
|
|
argType: BOOL
|
|
argIndex: 2
|
|
}
|
|
}
|
|
inputFrameworkOpName: "Gemm"
|
|
}
|
|
rule {
|
|
ruleName: "invertbooleannumber"
|
|
functionName: "invertbooleannumber"
|
|
inputIntName: "transA"
|
|
inputIntName: "transB"
|
|
outputIntName: "transX"
|
|
outputIntName: "transY"
|
|
inputToOutput {
|
|
key: "transX"
|
|
value: "transA"
|
|
}
|
|
inputToOutput {
|
|
key: "transY"
|
|
value: "transB"
|
|
}
|
|
ruleType: "attribute"
|
|
inputFrameworkOpName: "Gemm"
|
|
}
|
|
}
|
|
mappings {
|
|
frameworkName: "onnx"
|
|
opName: "acosh"
|
|
inputFrameworkOpName: "Acosh"
|
|
rule {
|
|
ruleName: "ndarraymapping"
|
|
functionName: "ndarraymapping"
|
|
inputTensorName: "input"
|
|
outputTensorName: "input"
|
|
inputToOutput {
|
|
key: "input"
|
|
value: "input"
|
|
}
|
|
ruleType: "tensor"
|
|
inputFrameworkOpName: "Acosh"
|
|
}
|
|
rule {
|
|
ruleName: "argdescriptorconstant"
|
|
functionName: "argdescriptorconstant"
|
|
inputBooleanName: "inPlace"
|
|
ruleType: "attribute"
|
|
transformerArgs {
|
|
key: "value"
|
|
transformerArgs {
|
|
name: "inPlace"
|
|
argType: BOOL
|
|
}
|
|
}
|
|
inputFrameworkOpName: "Acosh"
|
|
}
|
|
}
|
|
mappings {
|
|
frameworkName: "onnx"
|
|
opName: "less_equal"
|
|
inputFrameworkOpName: "LessOrEqual"
|
|
rule {
|
|
ruleName: "ndarraymapping"
|
|
functionName: "ndarraymapping"
|
|
inputTensorName: "A"
|
|
inputTensorName: "B"
|
|
outputTensorName: "input"
|
|
outputTensorName: "y"
|
|
inputToOutput {
|
|
key: "input"
|
|
value: "A"
|
|
}
|
|
inputToOutput {
|
|
key: "y"
|
|
value: "B"
|
|
}
|
|
ruleType: "tensor"
|
|
inputFrameworkOpName: "LessOrEqual"
|
|
}
|
|
rule {
|
|
ruleName: "argdescriptorconstant"
|
|
functionName: "argdescriptorconstant"
|
|
inputBooleanName: "inPlace"
|
|
ruleType: "attribute"
|
|
transformerArgs {
|
|
key: "value"
|
|
transformerArgs {
|
|
name: "inPlace"
|
|
argType: BOOL
|
|
}
|
|
}
|
|
inputFrameworkOpName: "LessOrEqual"
|
|
}
|
|
}
|
|
mappings {
|
|
frameworkName: "onnx"
|
|
opName: "cosh"
|
|
inputFrameworkOpName: "Cosh"
|
|
rule {
|
|
ruleName: "ndarraymapping"
|
|
functionName: "ndarraymapping"
|
|
inputTensorName: "input"
|
|
outputTensorName: "input"
|
|
inputToOutput {
|
|
key: "input"
|
|
value: "input"
|
|
}
|
|
ruleType: "tensor"
|
|
inputFrameworkOpName: "Cosh"
|
|
}
|
|
rule {
|
|
ruleName: "argdescriptorconstant"
|
|
functionName: "argdescriptorconstant"
|
|
inputBooleanName: "inPlace"
|
|
ruleType: "attribute"
|
|
transformerArgs {
|
|
key: "value"
|
|
transformerArgs {
|
|
name: "inPlace"
|
|
argType: BOOL
|
|
}
|
|
}
|
|
inputFrameworkOpName: "Cosh"
|
|
}
|
|
}
|
|
mappings {
|
|
frameworkName: "onnx"
|
|
opName: "non_max_suppression_v3"
|
|
inputFrameworkOpName: "NonMaxSuppression"
|
|
rule {
|
|
ruleName: "ndarraymapping"
|
|
functionName: "ndarraymapping"
|
|
inputTensorName: "boxes"
|
|
inputTensorName: "scores"
|
|
inputTensorName: "max_output_boxes_per_class"
|
|
inputTensorName: "iou_threshold"
|
|
inputTensorName: "score_threshold"
|
|
outputTensorName: "boxes"
|
|
outputTensorName: "scales"
|
|
outputTensorName: "maxOutSize"
|
|
outputTensorName: "iouThreshold"
|
|
outputTensorName: "scoreThreshold"
|
|
inputToOutput {
|
|
key: "boxes"
|
|
value: "boxes"
|
|
}
|
|
inputToOutput {
|
|
key: "scales"
|
|
value: "scores"
|
|
}
|
|
inputToOutput {
|
|
key: "maxOutSize"
|
|
value: "max_output_boxes_per_class"
|
|
}
|
|
inputToOutput {
|
|
key: "iouThreshold"
|
|
value: "iou_threshold"
|
|
}
|
|
inputToOutput {
|
|
key: "scoreThreshold"
|
|
value: "score_threshold"
|
|
}
|
|
ruleType: "tensor"
|
|
inputFrameworkOpName: "NonMaxSuppression"
|
|
}
|
|
rule {
|
|
ruleName: "ndarraytointattributevalue"
|
|
functionName: "ndarraytointattributevalue"
|
|
outputIntName: "maxOutputSize"
|
|
inputToOutput {
|
|
key: "maxOutputSize"
|
|
value: "max_output_boxes_per_class"
|
|
}
|
|
ruleType: "attribute"
|
|
inputFrameworkOpName: "NonMaxSuppression"
|
|
}
|
|
}
|
|
mappings {
|
|
frameworkName: "onnx"
|
|
opName: "log_softmax"
|
|
inputFrameworkOpName: "LogSoftmax"
|
|
rule {
|
|
ruleName: "ndarraymapping"
|
|
functionName: "ndarraymapping"
|
|
inputTensorName: "input"
|
|
outputTensorName: "input"
|
|
inputToOutput {
|
|
key: "input"
|
|
value: "input"
|
|
}
|
|
ruleType: "tensor"
|
|
inputFrameworkOpName: "LogSoftmax"
|
|
}
|
|
rule {
|
|
ruleName: "valuemapping"
|
|
functionName: "valuemapping"
|
|
inputIntName: "axis"
|
|
outputIntName: "dimension"
|
|
inputToOutput {
|
|
key: "dimension"
|
|
value: "axis"
|
|
}
|
|
ruleType: "attribute"
|
|
inputFrameworkOpName: "LogSoftmax"
|
|
}
|
|
}
|
|
mappings {
|
|
frameworkName: "onnx"
|
|
opName: "shape_of"
|
|
inputFrameworkOpName: "Shape"
|
|
rule {
|
|
ruleName: "ndarraymapping"
|
|
functionName: "ndarraymapping"
|
|
inputTensorName: "data"
|
|
outputTensorName: "input"
|
|
inputToOutput {
|
|
key: "input"
|
|
value: "data"
|
|
}
|
|
ruleType: "tensor"
|
|
inputFrameworkOpName: "Shape"
|
|
}
|
|
rule {
|
|
ruleName: "argdescriptorconstant"
|
|
functionName: "argdescriptorconstant"
|
|
inputBooleanName: "inPlace"
|
|
ruleType: "attribute"
|
|
transformerArgs {
|
|
key: "value"
|
|
transformerArgs {
|
|
name: "inPlace"
|
|
argType: BOOL
|
|
}
|
|
}
|
|
inputFrameworkOpName: "Shape"
|
|
}
|
|
}
|
|
mappings {
|
|
frameworkName: "onnx"
|
|
opName: "random_normal"
|
|
inputFrameworkOpName: "RandomNormal"
|
|
rule {
|
|
ruleName: "listnumbertondarray"
|
|
functionName: "listnumbertondarray"
|
|
inputToOutput {
|
|
key: "input"
|
|
value: "shape"
|
|
}
|
|
ruleType: "attribute"
|
|
inputFrameworkOpName: "RandomNormal"
|
|
}
|
|
}
|
|
mappings {
|
|
frameworkName: "onnx"
|
|
opName: "hard_sigmoid"
|
|
inputFrameworkOpName: "HardSigmoid"
|
|
rule {
|
|
ruleName: "ndarraymapping"
|
|
functionName: "ndarraymapping"
|
|
inputTensorName: "X"
|
|
outputTensorName: "input"
|
|
inputToOutput {
|
|
key: "input"
|
|
value: "X"
|
|
}
|
|
ruleType: "tensor"
|
|
inputFrameworkOpName: "HardSigmoid"
|
|
}
|
|
rule {
|
|
ruleName: "argdescriptorconstant"
|
|
functionName: "argdescriptorconstant"
|
|
inputBooleanName: "inPlace"
|
|
ruleType: "attribute"
|
|
transformerArgs {
|
|
key: "value"
|
|
transformerArgs {
|
|
name: "inPlace"
|
|
argType: BOOL
|
|
}
|
|
}
|
|
inputFrameworkOpName: "HardSigmoid"
|
|
}
|
|
}
|
|
mappings {
|
|
frameworkName: "onnx"
|
|
opName: "noop"
|
|
inputFrameworkOpName: "Constant"
|
|
}
|
|
mappings {
|
|
frameworkName: "onnx"
|
|
opName: "cumsum"
|
|
inputFrameworkOpName: "CumSum"
|
|
rule {
|
|
ruleName: "ndarraymapping"
|
|
functionName: "ndarraymapping"
|
|
inputTensorName: "x"
|
|
outputTensorName: "input"
|
|
inputToOutput {
|
|
key: "input"
|
|
value: "x"
|
|
}
|
|
ruleType: "tensor"
|
|
inputFrameworkOpName: "CumSum"
|
|
}
|
|
rule {
|
|
ruleName: "valuemapping"
|
|
functionName: "valuemapping"
|
|
inputIntName: "exclusive"
|
|
inputIntName: "reverse"
|
|
outputIntName: "exclusive"
|
|
outputIntName: "reverse"
|
|
inputToOutput {
|
|
key: "exclusive"
|
|
value: "exclusive"
|
|
}
|
|
inputToOutput {
|
|
key: "reverse"
|
|
value: "reverse"
|
|
}
|
|
ruleType: "attribute"
|
|
inputFrameworkOpName: "CumSum"
|
|
}
|
|
rule {
|
|
ruleName: "ndarraytointattributevalue"
|
|
functionName: "ndarraytointattributevalue"
|
|
outputIntName: "dimensions"
|
|
inputToOutput {
|
|
key: "dimensions"
|
|
value: "axis"
|
|
}
|
|
ruleType: "attribute"
|
|
inputFrameworkOpName: "CumSum"
|
|
}
|
|
}
|
|
mappings {
|
|
frameworkName: "onnx"
|
|
opName: "scatter_update"
|
|
inputFrameworkOpName: "ScatterElements"
|
|
rule {
|
|
ruleName: "ndarraymapping"
|
|
functionName: "ndarraymapping"
|
|
inputTensorName: "data"
|
|
inputTensorName: "updates"
|
|
inputTensorName: "indices"
|
|
outputTensorName: "operand"
|
|
outputTensorName: "updates"
|
|
outputTensorName: "indices"
|
|
inputToOutput {
|
|
key: "operand"
|
|
value: "data"
|
|
}
|
|
inputToOutput {
|
|
key: "updates"
|
|
value: "updates"
|
|
}
|
|
inputToOutput {
|
|
key: "indices"
|
|
value: "indices"
|
|
}
|
|
ruleType: "tensor"
|
|
inputFrameworkOpName: "ScatterElements"
|
|
}
|
|
}
|
|
mappings {
|
|
frameworkName: "onnx"
|
|
opName: "gruCell"
|
|
inputFrameworkOpName: "GRU"
|
|
rule {
|
|
ruleName: "ndarraymapping"
|
|
functionName: "ndarraymapping"
|
|
inputTensorName: "X"
|
|
inputTensorName: "R"
|
|
inputTensorName: "W"
|
|
inputTensorName: "B"
|
|
inputTensorName: "initial_h"
|
|
inputTensorName: "B"
|
|
outputTensorName: "input"
|
|
outputTensorName: "Wru"
|
|
outputTensorName: "Wc"
|
|
outputTensorName: "bc"
|
|
outputTensorName: "hLast"
|
|
outputTensorName: "bru"
|
|
inputToOutput {
|
|
key: "input"
|
|
value: "X"
|
|
}
|
|
inputToOutput {
|
|
key: "Wru"
|
|
value: "R"
|
|
}
|
|
inputToOutput {
|
|
key: "Wc"
|
|
value: "W"
|
|
}
|
|
inputToOutput {
|
|
key: "bc"
|
|
value: "B"
|
|
}
|
|
inputToOutput {
|
|
key: "hLast"
|
|
value: "initial_h"
|
|
}
|
|
inputToOutput {
|
|
key: "bru"
|
|
value: "B"
|
|
}
|
|
ruleType: "tensor"
|
|
inputFrameworkOpName: "GRU"
|
|
}
|
|
}
|
|
mappings {
|
|
frameworkName: "onnx"
|
|
opName: "reduce_norm1"
|
|
inputFrameworkOpName: "ReduceL1"
|
|
rule {
|
|
ruleName: "ndarraymapping"
|
|
functionName: "ndarraymapping"
|
|
inputTensorName: "data"
|
|
outputTensorName: "input"
|
|
inputToOutput {
|
|
key: "input"
|
|
value: "data"
|
|
}
|
|
ruleType: "tensor"
|
|
inputFrameworkOpName: "ReduceL1"
|
|
}
|
|
rule {
|
|
ruleName: "invertbooleannumber"
|
|
functionName: "invertbooleannumber"
|
|
inputIntName: "keepdims"
|
|
outputBooleanName: "keepDims"
|
|
inputToOutput {
|
|
key: "keepDims"
|
|
value: "keepdims"
|
|
}
|
|
ruleType: "attribute"
|
|
inputFrameworkOpName: "ReduceL1"
|
|
}
|
|
rule {
|
|
ruleName: "listnumbertolistnumber"
|
|
functionName: "listnumbertolistnumber"
|
|
outputIntName: "dimensions"
|
|
inputToOutput {
|
|
key: "dimensions"
|
|
value: "axes"
|
|
}
|
|
ruleType: "attribute"
|
|
inputFrameworkOpName: "ReduceL1"
|
|
}
|
|
}
|
|
mappings {
|
|
frameworkName: "onnx"
|
|
opName: "abs"
|
|
inputFrameworkOpName: "Abs"
|
|
rule {
|
|
ruleName: "ndarraymapping"
|
|
functionName: "ndarraymapping"
|
|
inputTensorName: "X"
|
|
outputTensorName: "input"
|
|
inputToOutput {
|
|
key: "input"
|
|
value: "X"
|
|
}
|
|
ruleType: "tensor"
|
|
inputFrameworkOpName: "Abs"
|
|
}
|
|
rule {
|
|
ruleName: "argdescriptorconstant"
|
|
functionName: "argdescriptorconstant"
|
|
inputBooleanName: "inPlace"
|
|
ruleType: "attribute"
|
|
transformerArgs {
|
|
key: "value"
|
|
transformerArgs {
|
|
name: "inPlace"
|
|
argType: BOOL
|
|
}
|
|
}
|
|
inputFrameworkOpName: "Abs"
|
|
}
|
|
}
|
|
mappings {
|
|
frameworkName: "onnx"
|
|
opName: "fill"
|
|
inputFrameworkOpName: "ConstantOfShape"
|
|
rule {
|
|
ruleName: "ndarraymapping"
|
|
functionName: "ndarraymapping"
|
|
inputTensorName: "input"
|
|
outputTensorName: "shape"
|
|
inputToOutput {
|
|
key: "shape"
|
|
value: "input"
|
|
}
|
|
ruleType: "tensor"
|
|
inputFrameworkOpName: "ConstantOfShape"
|
|
}
|
|
rule {
|
|
ruleName: "attributendarraytoscalarattribute"
|
|
functionName: "attributendarraytoscalarattribute"
|
|
outputDoubleName: "value"
|
|
inputTensorName: "value"
|
|
inputToOutput {
|
|
key: "value"
|
|
value: "value"
|
|
}
|
|
ruleType: "attribute"
|
|
inputFrameworkOpName: "ConstantOfShape"
|
|
}
|
|
rule {
|
|
ruleName: "argdescriptorconstant"
|
|
functionName: "argdescriptorconstant"
|
|
inputIntName: "outputDataType"
|
|
ruleType: "attribute"
|
|
transformerArgs {
|
|
key: "value"
|
|
transformerArgs {
|
|
name: "outputDataType"
|
|
argType: INT64
|
|
}
|
|
}
|
|
inputFrameworkOpName: "ConstantOfShape"
|
|
}
|
|
}
|
|
mappings {
|
|
frameworkName: "onnx"
|
|
opName: "reduce_norm2"
|
|
inputFrameworkOpName: "ReduceL2"
|
|
rule {
|
|
ruleName: "ndarraymapping"
|
|
functionName: "ndarraymapping"
|
|
inputTensorName: "data"
|
|
outputTensorName: "input"
|
|
inputToOutput {
|
|
key: "input"
|
|
value: "data"
|
|
}
|
|
ruleType: "tensor"
|
|
inputFrameworkOpName: "ReduceL2"
|
|
}
|
|
rule {
|
|
ruleName: "invertbooleannumber"
|
|
functionName: "invertbooleannumber"
|
|
inputIntName: "keepdims"
|
|
outputBooleanName: "keepDims"
|
|
inputToOutput {
|
|
key: "keepDims"
|
|
value: "keepdims"
|
|
}
|
|
ruleType: "attribute"
|
|
inputFrameworkOpName: "ReduceL2"
|
|
}
|
|
rule {
|
|
ruleName: "listnumbertolistnumber"
|
|
functionName: "listnumbertolistnumber"
|
|
outputIntName: "dimensions"
|
|
inputToOutput {
|
|
key: "dimensions"
|
|
value: "axes"
|
|
}
|
|
ruleType: "attribute"
|
|
inputFrameworkOpName: "ReduceL2"
|
|
}
|
|
}
|
|
mappings {
|
|
frameworkName: "onnx"
|
|
opName: "round"
|
|
inputFrameworkOpName: "Round"
|
|
rule {
|
|
ruleName: "ndarraymapping"
|
|
functionName: "ndarraymapping"
|
|
inputTensorName: "X"
|
|
outputTensorName: "input"
|
|
inputToOutput {
|
|
key: "input"
|
|
value: "X"
|
|
}
|
|
ruleType: "tensor"
|
|
inputFrameworkOpName: "Round"
|
|
}
|
|
rule {
|
|
ruleName: "argdescriptorconstant"
|
|
functionName: "argdescriptorconstant"
|
|
inputBooleanName: "inPlace"
|
|
ruleType: "attribute"
|
|
transformerArgs {
|
|
key: "value"
|
|
transformerArgs {
|
|
name: "inPlace"
|
|
argType: BOOL
|
|
}
|
|
}
|
|
inputFrameworkOpName: "Round"
|
|
}
|
|
}
|
|
mappings {
|
|
frameworkName: "onnx"
|
|
opName: "selu"
|
|
inputFrameworkOpName: "Selu"
|
|
rule {
|
|
ruleName: "ndarraymapping"
|
|
functionName: "ndarraymapping"
|
|
inputTensorName: "X"
|
|
outputTensorName: "input"
|
|
inputToOutput {
|
|
key: "input"
|
|
value: "X"
|
|
}
|
|
ruleType: "tensor"
|
|
inputFrameworkOpName: "Selu"
|
|
}
|
|
rule {
|
|
ruleName: "argdescriptorconstant"
|
|
functionName: "argdescriptorconstant"
|
|
inputBooleanName: "inPlace"
|
|
ruleType: "attribute"
|
|
transformerArgs {
|
|
key: "value"
|
|
transformerArgs {
|
|
name: "inPlace"
|
|
argType: BOOL
|
|
}
|
|
}
|
|
inputFrameworkOpName: "Selu"
|
|
}
|
|
}
|
|
mappings {
|
|
frameworkName: "onnx"
|
|
opName: "argmin"
|
|
inputFrameworkOpName: "ArgMin"
|
|
rule {
|
|
ruleName: "ndarraymapping"
|
|
functionName: "ndarraymapping"
|
|
inputTensorName: "data"
|
|
outputTensorName: "input"
|
|
inputToOutput {
|
|
key: "input"
|
|
value: "data"
|
|
}
|
|
ruleType: "tensor"
|
|
inputFrameworkOpName: "ArgMin"
|
|
}
|
|
rule {
|
|
ruleName: "invertbooleannumber"
|
|
functionName: "invertbooleannumber"
|
|
inputIntName: "keepdims"
|
|
outputBooleanName: "keepDims"
|
|
inputToOutput {
|
|
key: "keepDims"
|
|
value: "keepdims"
|
|
}
|
|
ruleType: "attribute"
|
|
inputFrameworkOpName: "ArgMin"
|
|
}
|
|
rule {
|
|
ruleName: "valuemapping"
|
|
functionName: "valuemapping"
|
|
inputIntName: "axis"
|
|
outputIntName: "dimensions"
|
|
inputToOutput {
|
|
key: "dimensions"
|
|
value: "axis"
|
|
}
|
|
ruleType: "attribute"
|
|
inputFrameworkOpName: "ArgMin"
|
|
}
|
|
}
|
|
mappings {
|
|
frameworkName: "onnx"
|
|
opName: "sigmoid"
|
|
inputFrameworkOpName: "Sigmoid"
|
|
rule {
|
|
ruleName: "ndarraymapping"
|
|
functionName: "ndarraymapping"
|
|
inputTensorName: "X"
|
|
outputTensorName: "input"
|
|
inputToOutput {
|
|
key: "input"
|
|
value: "X"
|
|
}
|
|
ruleType: "tensor"
|
|
inputFrameworkOpName: "Sigmoid"
|
|
}
|
|
rule {
|
|
ruleName: "argdescriptorconstant"
|
|
functionName: "argdescriptorconstant"
|
|
inputBooleanName: "inPlace"
|
|
ruleType: "attribute"
|
|
transformerArgs {
|
|
key: "value"
|
|
transformerArgs {
|
|
name: "inPlace"
|
|
argType: BOOL
|
|
}
|
|
}
|
|
inputFrameworkOpName: "Sigmoid"
|
|
}
|
|
}
|
|
mappings {
|
|
frameworkName: "onnx"
|
|
opName: "avgpool2d"
|
|
inputFrameworkOpName: "AveragePool"
|
|
rule {
|
|
ruleName: "ndarraymapping"
|
|
functionName: "ndarraymapping"
|
|
inputTensorName: "X"
|
|
outputTensorName: "input"
|
|
inputToOutput {
|
|
key: "input"
|
|
value: "X"
|
|
}
|
|
ruleType: "tensor"
|
|
inputFrameworkOpName: "AveragePool"
|
|
}
|
|
rule {
|
|
ruleName: "argdescriptorconstant"
|
|
functionName: "argdescriptorconstant"
|
|
inputFloatName: "isNCHW"
|
|
ruleType: "attribute"
|
|
transformerArgs {
|
|
key: "value"
|
|
transformerArgs {
|
|
name: "isNCHW"
|
|
int64Value: 1
|
|
argIndex: 10
|
|
}
|
|
}
|
|
inputFrameworkOpName: "AveragePool"
|
|
}
|
|
rule {
|
|
ruleName: "argdescriptorconstant"
|
|
functionName: "argdescriptorconstant"
|
|
inputIntName: "dH"
|
|
ruleType: "attribute"
|
|
transformerArgs {
|
|
key: "value"
|
|
transformerArgs {
|
|
name: "dH"
|
|
argType: INT64
|
|
argIndex: 6
|
|
}
|
|
}
|
|
inputFrameworkOpName: "AveragePool"
|
|
}
|
|
rule {
|
|
ruleName: "argdescriptorconstant"
|
|
functionName: "argdescriptorconstant"
|
|
inputIntName: "dW"
|
|
ruleType: "attribute"
|
|
transformerArgs {
|
|
key: "value"
|
|
transformerArgs {
|
|
name: "dW"
|
|
argType: INT64
|
|
argIndex: 7
|
|
}
|
|
}
|
|
inputFrameworkOpName: "AveragePool"
|
|
}
|
|
rule {
|
|
ruleName: "argdescriptorconstant"
|
|
functionName: "argdescriptorconstant"
|
|
inputIntName: "extraParam0"
|
|
ruleType: "attribute"
|
|
transformerArgs {
|
|
key: "value"
|
|
transformerArgs {
|
|
name: "extraParam0"
|
|
argType: INT64
|
|
argIndex: 9
|
|
}
|
|
}
|
|
inputFrameworkOpName: "AveragePool"
|
|
}
|
|
rule {
|
|
ruleName: "stringcontains"
|
|
functionName: "stringcontains"
|
|
inputStringAttrName: "auto_pad"
|
|
outputIntName: "isSameMode"
|
|
inputFloatName: "auto_pad"
|
|
inputToOutput {
|
|
key: "isSameMode"
|
|
value: "auto_pad"
|
|
}
|
|
ruleType: "attribute"
|
|
transformerArgs {
|
|
key: "isSameMode"
|
|
transformerArgs {
|
|
name: "auto_pad"
|
|
argIndex: 8
|
|
stringValue: "SAME"
|
|
}
|
|
}
|
|
inputFrameworkOpName: "AveragePool"
|
|
}
|
|
rule {
|
|
ruleName: "listattributevaluelookuptoindex"
|
|
functionName: "listattributevaluelookuptoindex"
|
|
outputIntName: "pH"
|
|
inputFloatName: "pads"
|
|
inputToOutput {
|
|
key: "pH"
|
|
value: "pads"
|
|
}
|
|
ruleType: "attribute"
|
|
transformerArgs {
|
|
key: "pH"
|
|
transformerArgs {
|
|
name: "pads"
|
|
argIndex: 4
|
|
}
|
|
}
|
|
inputFrameworkOpName: "AveragePool"
|
|
}
|
|
rule {
|
|
ruleName: "listattributevaluelookuptoindex"
|
|
functionName: "listattributevaluelookuptoindex"
|
|
outputIntName: "pW"
|
|
inputFloatName: "pads"
|
|
inputToOutput {
|
|
key: "pW"
|
|
value: "pads"
|
|
}
|
|
ruleType: "attribute"
|
|
transformerArgs {
|
|
key: "pW"
|
|
transformerArgs {
|
|
name: "pads"
|
|
int64Value: 1
|
|
argIndex: 5
|
|
}
|
|
}
|
|
inputFrameworkOpName: "AveragePool"
|
|
}
|
|
rule {
|
|
ruleName: "listattributevaluelookuptoindex"
|
|
functionName: "listattributevaluelookuptoindex"
|
|
outputIntName: "sH"
|
|
inputFloatName: "strides"
|
|
inputToOutput {
|
|
key: "sH"
|
|
value: "strides"
|
|
}
|
|
ruleType: "attribute"
|
|
transformerArgs {
|
|
key: "sH"
|
|
transformerArgs {
|
|
name: "strides"
|
|
argIndex: 2
|
|
}
|
|
}
|
|
inputFrameworkOpName: "AveragePool"
|
|
}
|
|
rule {
|
|
ruleName: "listattributevaluelookuptoindex"
|
|
functionName: "listattributevaluelookuptoindex"
|
|
outputIntName: "sW"
|
|
inputFloatName: "strides"
|
|
inputToOutput {
|
|
key: "sW"
|
|
value: "strides"
|
|
}
|
|
ruleType: "attribute"
|
|
transformerArgs {
|
|
key: "sW"
|
|
transformerArgs {
|
|
name: "strides"
|
|
int64Value: 1
|
|
argIndex: 3
|
|
}
|
|
}
|
|
inputFrameworkOpName: "AveragePool"
|
|
}
|
|
rule {
|
|
ruleName: "listattributevaluelookuptoindex"
|
|
functionName: "listattributevaluelookuptoindex"
|
|
outputIntName: "kW"
|
|
inputFloatName: "kernel_shape"
|
|
inputToOutput {
|
|
key: "kW"
|
|
value: "kernel_shape"
|
|
}
|
|
ruleType: "attribute"
|
|
transformerArgs {
|
|
key: "kW"
|
|
transformerArgs {
|
|
name: "kernel_shape"
|
|
int64Value: 1
|
|
argIndex: 1
|
|
}
|
|
}
|
|
inputFrameworkOpName: "AveragePool"
|
|
}
|
|
rule {
|
|
ruleName: "listattributevaluelookuptoindex"
|
|
functionName: "listattributevaluelookuptoindex"
|
|
outputIntName: "kH"
|
|
inputFloatName: "kernel_shape"
|
|
inputToOutput {
|
|
key: "kH"
|
|
value: "kernel_shape"
|
|
}
|
|
ruleType: "attribute"
|
|
transformerArgs {
|
|
key: "kH"
|
|
transformerArgs {
|
|
name: "kernel_shape"
|
|
}
|
|
}
|
|
inputFrameworkOpName: "AveragePool"
|
|
}
|
|
}
|
|
mappings {
|
|
frameworkName: "onnx"
|
|
opName: "dropout_inverted"
|
|
inputFrameworkOpName: "Dropout"
|
|
rule {
|
|
ruleName: "ndarraymapping"
|
|
functionName: "ndarraymapping"
|
|
inputTensorName: "data"
|
|
outputTensorName: "input"
|
|
inputToOutput {
|
|
key: "input"
|
|
value: "data"
|
|
}
|
|
ruleType: "tensor"
|
|
inputFrameworkOpName: "Dropout"
|
|
}
|
|
rule {
|
|
ruleName: "ndarrayinputtonumericalattribute"
|
|
functionName: "ndarrayinputtonumericalattribute"
|
|
outputDoubleName: "p"
|
|
inputToOutput {
|
|
key: "p"
|
|
value: "ratio"
|
|
}
|
|
ruleType: "attribute"
|
|
inputFrameworkOpName: "Dropout"
|
|
}
|
|
}
|
|
mappings {
|
|
frameworkName: "onnx"
|
|
opName: "atan"
|
|
inputFrameworkOpName: "Atan"
|
|
rule {
|
|
ruleName: "ndarraymapping"
|
|
functionName: "ndarraymapping"
|
|
inputTensorName: "input"
|
|
outputTensorName: "input"
|
|
inputToOutput {
|
|
key: "input"
|
|
value: "input"
|
|
}
|
|
ruleType: "tensor"
|
|
inputFrameworkOpName: "Atan"
|
|
}
|
|
rule {
|
|
ruleName: "argdescriptorconstant"
|
|
functionName: "argdescriptorconstant"
|
|
inputBooleanName: "inPlace"
|
|
ruleType: "attribute"
|
|
transformerArgs {
|
|
key: "value"
|
|
transformerArgs {
|
|
name: "inPlace"
|
|
argType: BOOL
|
|
}
|
|
}
|
|
inputFrameworkOpName: "Atan"
|
|
}
|
|
}
|
|
mappings {
|
|
frameworkName: "onnx"
|
|
opName: "floor"
|
|
inputFrameworkOpName: "Floor"
|
|
rule {
|
|
ruleName: "ndarraymapping"
|
|
functionName: "ndarraymapping"
|
|
inputTensorName: "X"
|
|
outputTensorName: "input"
|
|
inputToOutput {
|
|
key: "input"
|
|
value: "X"
|
|
}
|
|
ruleType: "tensor"
|
|
inputFrameworkOpName: "Floor"
|
|
}
|
|
rule {
|
|
ruleName: "argdescriptorconstant"
|
|
functionName: "argdescriptorconstant"
|
|
inputBooleanName: "inPlace"
|
|
ruleType: "attribute"
|
|
transformerArgs {
|
|
key: "value"
|
|
transformerArgs {
|
|
name: "inPlace"
|
|
argType: BOOL
|
|
}
|
|
}
|
|
inputFrameworkOpName: "Floor"
|
|
}
|
|
}
|
|
mappings {
|
|
frameworkName: "onnx"
|
|
opName: "prelu"
|
|
inputFrameworkOpName: "PRelu"
|
|
rule {
|
|
ruleName: "ndarraymapping"
|
|
functionName: "ndarraymapping"
|
|
inputTensorName: "X"
|
|
inputTensorName: "slope"
|
|
outputTensorName: "input"
|
|
outputTensorName: "alpha"
|
|
inputToOutput {
|
|
key: "input"
|
|
value: "X"
|
|
}
|
|
inputToOutput {
|
|
key: "alpha"
|
|
value: "slope"
|
|
}
|
|
ruleType: "tensor"
|
|
inputFrameworkOpName: "PRelu"
|
|
}
|
|
rule {
|
|
ruleName: "argdescriptorconstant"
|
|
functionName: "argdescriptorconstant"
|
|
inputIntName: "sharedAxes"
|
|
ruleType: "attribute"
|
|
transformerArgs {
|
|
key: "value"
|
|
transformerArgs {
|
|
name: "sharedAxes"
|
|
int64Value: -1
|
|
argType: INT64
|
|
}
|
|
}
|
|
inputFrameworkOpName: "PRelu"
|
|
}
|
|
}
|
|
mappings {
|
|
frameworkName: "onnx"
|
|
opName: "atanh"
|
|
inputFrameworkOpName: "Atanh"
|
|
rule {
|
|
ruleName: "ndarraymapping"
|
|
functionName: "ndarraymapping"
|
|
inputTensorName: "input"
|
|
outputTensorName: "input"
|
|
inputToOutput {
|
|
key: "input"
|
|
value: "input"
|
|
}
|
|
ruleType: "tensor"
|
|
inputFrameworkOpName: "Atanh"
|
|
}
|
|
rule {
|
|
ruleName: "argdescriptorconstant"
|
|
functionName: "argdescriptorconstant"
|
|
inputBooleanName: "inPlace"
|
|
ruleType: "attribute"
|
|
transformerArgs {
|
|
key: "value"
|
|
transformerArgs {
|
|
name: "inPlace"
|
|
argType: BOOL
|
|
}
|
|
}
|
|
inputFrameworkOpName: "Atanh"
|
|
}
|
|
}
|
|
mappings {
|
|
frameworkName: "onnx"
|
|
opName: "mod"
|
|
inputFrameworkOpName: "Mod"
|
|
rule {
|
|
ruleName: "ndarraymapping"
|
|
functionName: "ndarraymapping"
|
|
inputTensorName: "A"
|
|
inputTensorName: "B"
|
|
outputTensorName: "input"
|
|
outputTensorName: "y"
|
|
inputToOutput {
|
|
key: "input"
|
|
value: "A"
|
|
}
|
|
inputToOutput {
|
|
key: "y"
|
|
value: "B"
|
|
}
|
|
ruleType: "tensor"
|
|
inputFrameworkOpName: "Mod"
|
|
}
|
|
rule {
|
|
ruleName: "argdescriptorconstant"
|
|
functionName: "argdescriptorconstant"
|
|
inputBooleanName: "inPlace"
|
|
ruleType: "attribute"
|
|
transformerArgs {
|
|
key: "value"
|
|
transformerArgs {
|
|
name: "inPlace"
|
|
argType: BOOL
|
|
}
|
|
}
|
|
inputFrameworkOpName: "Mod"
|
|
}
|
|
}
|
|
mappings {
|
|
frameworkName: "onnx"
|
|
opName: "lstmLayer"
|
|
inputFrameworkOpName: "LSTM"
|
|
rule {
|
|
ruleName: "ndarraymapping"
|
|
functionName: "ndarraymapping"
|
|
inputTensorName: "X"
|
|
inputTensorName: "W"
|
|
inputTensorName: "R"
|
|
inputTensorName: "P"
|
|
inputTensorName: "B"
|
|
inputTensorName: "sequence_lens"
|
|
inputTensorName: "initial_h"
|
|
inputTensorName: "initial_c"
|
|
outputTensorName: "input"
|
|
outputTensorName: "Wx"
|
|
outputTensorName: "Wr"
|
|
outputTensorName: "Wp"
|
|
outputTensorName: "b"
|
|
outputTensorName: "seqLen"
|
|
outputTensorName: "hI"
|
|
outputTensorName: "cI"
|
|
inputToOutput {
|
|
key: "input"
|
|
value: "X"
|
|
}
|
|
inputToOutput {
|
|
key: "Wx"
|
|
value: "W"
|
|
}
|
|
inputToOutput {
|
|
key: "Wr"
|
|
value: "R"
|
|
}
|
|
inputToOutput {
|
|
key: "Wp"
|
|
value: "P"
|
|
}
|
|
inputToOutput {
|
|
key: "b"
|
|
value: "B"
|
|
}
|
|
inputToOutput {
|
|
key: "seqLen"
|
|
value: "sequence_lens"
|
|
}
|
|
inputToOutput {
|
|
key: "hI"
|
|
value: "initial_h"
|
|
}
|
|
inputToOutput {
|
|
key: "cI"
|
|
value: "initial_c"
|
|
}
|
|
ruleType: "tensor"
|
|
inputFrameworkOpName: "LSTM"
|
|
}
|
|
rule {
|
|
ruleName: "valuemapping"
|
|
functionName: "valuemapping"
|
|
inputFloatName: "clip"
|
|
outputDoubleName: "cellClip"
|
|
inputToOutput {
|
|
key: "cellClip"
|
|
value: "clip"
|
|
}
|
|
ruleType: "attribute"
|
|
inputFrameworkOpName: "LSTM"
|
|
}
|
|
rule {
|
|
ruleName: "stringtoindex"
|
|
functionName: "stringtoindex"
|
|
inputStringAttrName: "direction"
|
|
outputIntName: "directionMode"
|
|
inputFloatName: "directionMode"
|
|
inputFloatName: "directionMode"
|
|
inputFloatName: "directionMode"
|
|
inputToOutput {
|
|
key: "directionMode"
|
|
value: "direction"
|
|
}
|
|
ruleType: "attribute"
|
|
transformerArgs {
|
|
key: "directionMode"
|
|
transformerArgs {
|
|
name: "directionMode"
|
|
argIndex: 1
|
|
stringValue: "forward"
|
|
}
|
|
transformerArgs {
|
|
name: "directionMode"
|
|
argIndex: 1
|
|
stringValue: "reverse"
|
|
}
|
|
transformerArgs {
|
|
name: "directionMode"
|
|
argIndex: 1
|
|
stringValue: "bidirectional"
|
|
}
|
|
}
|
|
transformerArgs {
|
|
key: "directionMode"
|
|
transformerArgs {
|
|
name: "directionMode"
|
|
argIndex: 1
|
|
stringValue: "forward"
|
|
}
|
|
transformerArgs {
|
|
name: "directionMode"
|
|
argIndex: 1
|
|
stringValue: "reverse"
|
|
}
|
|
transformerArgs {
|
|
name: "directionMode"
|
|
argIndex: 1
|
|
stringValue: "bidirectional"
|
|
}
|
|
}
|
|
transformerArgs {
|
|
key: "directionMode"
|
|
transformerArgs {
|
|
name: "directionMode"
|
|
argIndex: 1
|
|
stringValue: "forward"
|
|
}
|
|
transformerArgs {
|
|
name: "directionMode"
|
|
argIndex: 1
|
|
stringValue: "reverse"
|
|
}
|
|
transformerArgs {
|
|
name: "directionMode"
|
|
argIndex: 1
|
|
stringValue: "bidirectional"
|
|
}
|
|
}
|
|
inputFrameworkOpName: "LSTM"
|
|
}
|
|
rule {
|
|
ruleName: "argdescriptorconstant"
|
|
functionName: "argdescriptorconstant"
|
|
inputIntName: "dataFormat"
|
|
ruleType: "attribute"
|
|
transformerArgs {
|
|
key: "value"
|
|
transformerArgs {
|
|
name: "dataFormat"
|
|
argType: INT64
|
|
}
|
|
}
|
|
inputFrameworkOpName: "LSTM"
|
|
}
|
|
rule {
|
|
ruleName: "argdescriptorconstant"
|
|
functionName: "argdescriptorconstant"
|
|
inputBooleanName: "hasBiases"
|
|
ruleType: "attribute"
|
|
transformerArgs {
|
|
key: "value"
|
|
transformerArgs {
|
|
name: "hasBiases"
|
|
boolValue: true
|
|
argType: BOOL
|
|
}
|
|
}
|
|
inputFrameworkOpName: "LSTM"
|
|
}
|
|
rule {
|
|
ruleName: "argdescriptorconstant"
|
|
functionName: "argdescriptorconstant"
|
|
inputBooleanName: "hasSeqLen"
|
|
ruleType: "attribute"
|
|
transformerArgs {
|
|
key: "value"
|
|
transformerArgs {
|
|
name: "hasSeqLen"
|
|
boolValue: true
|
|
argType: BOOL
|
|
argIndex: 1
|
|
}
|
|
}
|
|
inputFrameworkOpName: "LSTM"
|
|
}
|
|
rule {
|
|
ruleName: "argdescriptorconstant"
|
|
functionName: "argdescriptorconstant"
|
|
inputBooleanName: "hasInitH"
|
|
ruleType: "attribute"
|
|
transformerArgs {
|
|
key: "value"
|
|
transformerArgs {
|
|
name: "hasInitH"
|
|
boolValue: true
|
|
argType: BOOL
|
|
argIndex: 2
|
|
}
|
|
}
|
|
inputFrameworkOpName: "LSTM"
|
|
}
|
|
rule {
|
|
ruleName: "argdescriptorconstant"
|
|
functionName: "argdescriptorconstant"
|
|
inputBooleanName: "hasInitC"
|
|
ruleType: "attribute"
|
|
transformerArgs {
|
|
key: "value"
|
|
transformerArgs {
|
|
name: "hasInitC"
|
|
boolValue: true
|
|
argType: BOOL
|
|
argIndex: 3
|
|
}
|
|
}
|
|
inputFrameworkOpName: "LSTM"
|
|
}
|
|
rule {
|
|
ruleName: "argdescriptorconstant"
|
|
functionName: "argdescriptorconstant"
|
|
inputBooleanName: "hasPH"
|
|
ruleType: "attribute"
|
|
transformerArgs {
|
|
key: "value"
|
|
transformerArgs {
|
|
name: "hasPH"
|
|
boolValue: true
|
|
argType: BOOL
|
|
argIndex: 4
|
|
}
|
|
}
|
|
inputFrameworkOpName: "LSTM"
|
|
}
|
|
rule {
|
|
ruleName: "argdescriptorconstant"
|
|
functionName: "argdescriptorconstant"
|
|
inputBooleanName: "retFullSeq"
|
|
ruleType: "attribute"
|
|
transformerArgs {
|
|
key: "value"
|
|
transformerArgs {
|
|
name: "retFullSeq"
|
|
boolValue: true
|
|
argType: BOOL
|
|
argIndex: 5
|
|
}
|
|
}
|
|
inputFrameworkOpName: "LSTM"
|
|
}
|
|
rule {
|
|
ruleName: "argdescriptorconstant"
|
|
functionName: "argdescriptorconstant"
|
|
inputBooleanName: "retLastH"
|
|
ruleType: "attribute"
|
|
transformerArgs {
|
|
key: "value"
|
|
transformerArgs {
|
|
name: "retLastH"
|
|
boolValue: true
|
|
argType: BOOL
|
|
argIndex: 6
|
|
}
|
|
}
|
|
inputFrameworkOpName: "LSTM"
|
|
}
|
|
rule {
|
|
ruleName: "argdescriptorconstant"
|
|
functionName: "argdescriptorconstant"
|
|
inputBooleanName: "retLastC"
|
|
ruleType: "attribute"
|
|
transformerArgs {
|
|
key: "value"
|
|
transformerArgs {
|
|
name: "retLastC"
|
|
boolValue: true
|
|
argType: BOOL
|
|
argIndex: 7
|
|
}
|
|
}
|
|
inputFrameworkOpName: "LSTM"
|
|
}
|
|
rule {
|
|
ruleName: "listattributevaluelookuptoindex"
|
|
functionName: "listattributevaluelookuptoindex"
|
|
inputFloatName: "activation_alpha"
|
|
outputDoubleName: "gateAlpha"
|
|
inputToOutput {
|
|
key: "gateAlpha"
|
|
value: "activation_alpha"
|
|
}
|
|
ruleType: "attribute"
|
|
transformerArgs {
|
|
key: "gateAlpha"
|
|
transformerArgs {
|
|
name: "activation_alpha"
|
|
argIndex: 1
|
|
}
|
|
}
|
|
inputFrameworkOpName: "LSTM"
|
|
}
|
|
rule {
|
|
ruleName: "listattributevaluelookuptoindex"
|
|
functionName: "listattributevaluelookuptoindex"
|
|
inputFloatName: "activation_alpha"
|
|
outputDoubleName: "cellAlpha"
|
|
inputToOutput {
|
|
key: "cellAlpha"
|
|
value: "activation_alpha"
|
|
}
|
|
ruleType: "attribute"
|
|
transformerArgs {
|
|
key: "cellAlpha"
|
|
transformerArgs {
|
|
name: "activation_alpha"
|
|
int64Value: 1
|
|
argIndex: 3
|
|
}
|
|
}
|
|
inputFrameworkOpName: "LSTM"
|
|
}
|
|
rule {
|
|
ruleName: "listattributevaluelookuptoindex"
|
|
functionName: "listattributevaluelookuptoindex"
|
|
inputFloatName: "activation_alpha"
|
|
outputDoubleName: "outAlpha"
|
|
inputToOutput {
|
|
key: "outAlpha"
|
|
value: "activation_alpha"
|
|
}
|
|
ruleType: "attribute"
|
|
transformerArgs {
|
|
key: "outAlpha"
|
|
transformerArgs {
|
|
name: "activation_alpha"
|
|
int64Value: 2
|
|
argIndex: 5
|
|
}
|
|
}
|
|
inputFrameworkOpName: "LSTM"
|
|
}
|
|
rule {
|
|
ruleName: "listattributevaluelookuptoindex"
|
|
functionName: "listattributevaluelookuptoindex"
|
|
inputFloatName: "activation_beta"
|
|
outputDoubleName: "gateBeta"
|
|
inputToOutput {
|
|
key: "gateBeta"
|
|
value: "activation_beta"
|
|
}
|
|
ruleType: "attribute"
|
|
transformerArgs {
|
|
key: "gateBeta"
|
|
transformerArgs {
|
|
name: "activation_beta"
|
|
argIndex: 2
|
|
}
|
|
}
|
|
inputFrameworkOpName: "LSTM"
|
|
}
|
|
rule {
|
|
ruleName: "listattributevaluelookuptoindex"
|
|
functionName: "listattributevaluelookuptoindex"
|
|
inputFloatName: "activation_beta"
|
|
outputDoubleName: "cellBeta"
|
|
inputToOutput {
|
|
key: "cellBeta"
|
|
value: "activation_beta"
|
|
}
|
|
ruleType: "attribute"
|
|
transformerArgs {
|
|
key: "cellBeta"
|
|
transformerArgs {
|
|
name: "activation_beta"
|
|
int64Value: 1
|
|
argIndex: 4
|
|
}
|
|
}
|
|
inputFrameworkOpName: "LSTM"
|
|
}
|
|
rule {
|
|
ruleName: "listattributevaluelookuptoindex"
|
|
functionName: "listattributevaluelookuptoindex"
|
|
inputFloatName: "activation_beta"
|
|
outputDoubleName: "outBeta"
|
|
inputToOutput {
|
|
key: "outBeta"
|
|
value: "activation_beta"
|
|
}
|
|
ruleType: "attribute"
|
|
transformerArgs {
|
|
key: "outBeta"
|
|
transformerArgs {
|
|
name: "activation_beta"
|
|
int64Value: 2
|
|
argIndex: 6
|
|
}
|
|
}
|
|
inputFrameworkOpName: "LSTM"
|
|
}
|
|
rule {
|
|
ruleName: "mapstringtoindex"
|
|
functionName: "mapstringtoindex"
|
|
outputIntName: "gateAct"
|
|
inputFloatName: "Relu"
|
|
inputFloatName: "Tanh"
|
|
inputFloatName: "Sigmoid"
|
|
inputFloatName: "Affine"
|
|
inputFloatName: "LeakyRelu"
|
|
inputFloatName: "ThresholdedRelu"
|
|
inputFloatName: "ScaledTanh"
|
|
inputFloatName: "HardSigmoid"
|
|
inputFloatName: "Elu"
|
|
inputFloatName: "Softsign"
|
|
inputFloatName: "Softplus"
|
|
inputFloatName: "index"
|
|
inputToOutput {
|
|
key: "gateAct"
|
|
value: "activations"
|
|
}
|
|
ruleType: "attribute"
|
|
transformerArgs {
|
|
key: "gateAct"
|
|
transformerArgs {
|
|
name: "Relu"
|
|
int64Value: 1
|
|
argIndex: 2
|
|
}
|
|
transformerArgs {
|
|
name: "Tanh"
|
|
argIndex: 2
|
|
}
|
|
transformerArgs {
|
|
name: "Sigmoid"
|
|
int64Value: 2
|
|
argIndex: 2
|
|
}
|
|
transformerArgs {
|
|
name: "Affine"
|
|
int64Value: 3
|
|
argIndex: 2
|
|
}
|
|
transformerArgs {
|
|
name: "LeakyRelu"
|
|
int64Value: 4
|
|
argIndex: 2
|
|
}
|
|
transformerArgs {
|
|
name: "ThresholdedRelu"
|
|
int64Value: 5
|
|
argIndex: 2
|
|
}
|
|
transformerArgs {
|
|
name: "ScaledTanh"
|
|
int64Value: 6
|
|
argIndex: 2
|
|
}
|
|
transformerArgs {
|
|
name: "HardSigmoid"
|
|
int64Value: 7
|
|
argIndex: 2
|
|
}
|
|
transformerArgs {
|
|
name: "Elu"
|
|
int64Value: 8
|
|
argIndex: 2
|
|
}
|
|
transformerArgs {
|
|
name: "Softsign"
|
|
int64Value: 9
|
|
argIndex: 2
|
|
}
|
|
transformerArgs {
|
|
name: "Softplus"
|
|
int64Value: 10
|
|
argIndex: 2
|
|
}
|
|
}
|
|
transformerArgs {
|
|
key: "gateAct"
|
|
transformerArgs {
|
|
name: "Relu"
|
|
int64Value: 1
|
|
argIndex: 2
|
|
}
|
|
transformerArgs {
|
|
name: "Tanh"
|
|
argIndex: 2
|
|
}
|
|
transformerArgs {
|
|
name: "Sigmoid"
|
|
int64Value: 2
|
|
argIndex: 2
|
|
}
|
|
transformerArgs {
|
|
name: "Affine"
|
|
int64Value: 3
|
|
argIndex: 2
|
|
}
|
|
transformerArgs {
|
|
name: "LeakyRelu"
|
|
int64Value: 4
|
|
argIndex: 2
|
|
}
|
|
transformerArgs {
|
|
name: "ThresholdedRelu"
|
|
int64Value: 5
|
|
argIndex: 2
|
|
}
|
|
transformerArgs {
|
|
name: "ScaledTanh"
|
|
int64Value: 6
|
|
argIndex: 2
|
|
}
|
|
transformerArgs {
|
|
name: "HardSigmoid"
|
|
int64Value: 7
|
|
argIndex: 2
|
|
}
|
|
transformerArgs {
|
|
name: "Elu"
|
|
int64Value: 8
|
|
argIndex: 2
|
|
}
|
|
transformerArgs {
|
|
name: "Softsign"
|
|
int64Value: 9
|
|
argIndex: 2
|
|
}
|
|
transformerArgs {
|
|
name: "Softplus"
|
|
int64Value: 10
|
|
argIndex: 2
|
|
}
|
|
}
|
|
transformerArgs {
|
|
key: "gateAct"
|
|
transformerArgs {
|
|
name: "Relu"
|
|
int64Value: 1
|
|
argIndex: 2
|
|
}
|
|
transformerArgs {
|
|
name: "Tanh"
|
|
argIndex: 2
|
|
}
|
|
transformerArgs {
|
|
name: "Sigmoid"
|
|
int64Value: 2
|
|
argIndex: 2
|
|
}
|
|
transformerArgs {
|
|
name: "Affine"
|
|
int64Value: 3
|
|
argIndex: 2
|
|
}
|
|
transformerArgs {
|
|
name: "LeakyRelu"
|
|
int64Value: 4
|
|
argIndex: 2
|
|
}
|
|
transformerArgs {
|
|
name: "ThresholdedRelu"
|
|
int64Value: 5
|
|
argIndex: 2
|
|
}
|
|
transformerArgs {
|
|
name: "ScaledTanh"
|
|
int64Value: 6
|
|
argIndex: 2
|
|
}
|
|
transformerArgs {
|
|
name: "HardSigmoid"
|
|
int64Value: 7
|
|
argIndex: 2
|
|
}
|
|
transformerArgs {
|
|
name: "Elu"
|
|
int64Value: 8
|
|
argIndex: 2
|
|
}
|
|
transformerArgs {
|
|
name: "Softsign"
|
|
int64Value: 9
|
|
argIndex: 2
|
|
}
|
|
transformerArgs {
|
|
name: "Softplus"
|
|
int64Value: 10
|
|
argIndex: 2
|
|
}
|
|
}
|
|
transformerArgs {
|
|
key: "gateAct"
|
|
transformerArgs {
|
|
name: "Relu"
|
|
int64Value: 1
|
|
argIndex: 2
|
|
}
|
|
transformerArgs {
|
|
name: "Tanh"
|
|
argIndex: 2
|
|
}
|
|
transformerArgs {
|
|
name: "Sigmoid"
|
|
int64Value: 2
|
|
argIndex: 2
|
|
}
|
|
transformerArgs {
|
|
name: "Affine"
|
|
int64Value: 3
|
|
argIndex: 2
|
|
}
|
|
transformerArgs {
|
|
name: "LeakyRelu"
|
|
int64Value: 4
|
|
argIndex: 2
|
|
}
|
|
transformerArgs {
|
|
name: "ThresholdedRelu"
|
|
int64Value: 5
|
|
argIndex: 2
|
|
}
|
|
transformerArgs {
|
|
name: "ScaledTanh"
|
|
int64Value: 6
|
|
argIndex: 2
|
|
}
|
|
transformerArgs {
|
|
name: "HardSigmoid"
|
|
int64Value: 7
|
|
argIndex: 2
|
|
}
|
|
transformerArgs {
|
|
name: "Elu"
|
|
int64Value: 8
|
|
argIndex: 2
|
|
}
|
|
transformerArgs {
|
|
name: "Softsign"
|
|
int64Value: 9
|
|
argIndex: 2
|
|
}
|
|
transformerArgs {
|
|
name: "Softplus"
|
|
int64Value: 10
|
|
argIndex: 2
|
|
}
|
|
}
|
|
transformerArgs {
|
|
key: "gateAct"
|
|
transformerArgs {
|
|
name: "Relu"
|
|
int64Value: 1
|
|
argIndex: 2
|
|
}
|
|
transformerArgs {
|
|
name: "Tanh"
|
|
argIndex: 2
|
|
}
|
|
transformerArgs {
|
|
name: "Sigmoid"
|
|
int64Value: 2
|
|
argIndex: 2
|
|
}
|
|
transformerArgs {
|
|
name: "Affine"
|
|
int64Value: 3
|
|
argIndex: 2
|
|
}
|
|
transformerArgs {
|
|
name: "LeakyRelu"
|
|
int64Value: 4
|
|
argIndex: 2
|
|
}
|
|
transformerArgs {
|
|
name: "ThresholdedRelu"
|
|
int64Value: 5
|
|
argIndex: 2
|
|
}
|
|
transformerArgs {
|
|
name: "ScaledTanh"
|
|
int64Value: 6
|
|
argIndex: 2
|
|
}
|
|
transformerArgs {
|
|
name: "HardSigmoid"
|
|
int64Value: 7
|
|
argIndex: 2
|
|
}
|
|
transformerArgs {
|
|
name: "Elu"
|
|
int64Value: 8
|
|
argIndex: 2
|
|
}
|
|
transformerArgs {
|
|
name: "Softsign"
|
|
int64Value: 9
|
|
argIndex: 2
|
|
}
|
|
transformerArgs {
|
|
name: "Softplus"
|
|
int64Value: 10
|
|
argIndex: 2
|
|
}
|
|
}
|
|
transformerArgs {
|
|
key: "gateAct"
|
|
transformerArgs {
|
|
name: "Relu"
|
|
int64Value: 1
|
|
argIndex: 2
|
|
}
|
|
transformerArgs {
|
|
name: "Tanh"
|
|
argIndex: 2
|
|
}
|
|
transformerArgs {
|
|
name: "Sigmoid"
|
|
int64Value: 2
|
|
argIndex: 2
|
|
}
|
|
transformerArgs {
|
|
name: "Affine"
|
|
int64Value: 3
|
|
argIndex: 2
|
|
}
|
|
transformerArgs {
|
|
name: "LeakyRelu"
|
|
int64Value: 4
|
|
argIndex: 2
|
|
}
|
|
transformerArgs {
|
|
name: "ThresholdedRelu"
|
|
int64Value: 5
|
|
argIndex: 2
|
|
}
|
|
transformerArgs {
|
|
name: "ScaledTanh"
|
|
int64Value: 6
|
|
argIndex: 2
|
|
}
|
|
transformerArgs {
|
|
name: "HardSigmoid"
|
|
int64Value: 7
|
|
argIndex: 2
|
|
}
|
|
transformerArgs {
|
|
name: "Elu"
|
|
int64Value: 8
|
|
argIndex: 2
|
|
}
|
|
transformerArgs {
|
|
name: "Softsign"
|
|
int64Value: 9
|
|
argIndex: 2
|
|
}
|
|
transformerArgs {
|
|
name: "Softplus"
|
|
int64Value: 10
|
|
argIndex: 2
|
|
}
|
|
}
|
|
transformerArgs {
|
|
key: "gateAct"
|
|
transformerArgs {
|
|
name: "Relu"
|
|
int64Value: 1
|
|
argIndex: 2
|
|
}
|
|
transformerArgs {
|
|
name: "Tanh"
|
|
argIndex: 2
|
|
}
|
|
transformerArgs {
|
|
name: "Sigmoid"
|
|
int64Value: 2
|
|
argIndex: 2
|
|
}
|
|
transformerArgs {
|
|
name: "Affine"
|
|
int64Value: 3
|
|
argIndex: 2
|
|
}
|
|
transformerArgs {
|
|
name: "LeakyRelu"
|
|
int64Value: 4
|
|
argIndex: 2
|
|
}
|
|
transformerArgs {
|
|
name: "ThresholdedRelu"
|
|
int64Value: 5
|
|
argIndex: 2
|
|
}
|
|
transformerArgs {
|
|
name: "ScaledTanh"
|
|
int64Value: 6
|
|
argIndex: 2
|
|
}
|
|
transformerArgs {
|
|
name: "HardSigmoid"
|
|
int64Value: 7
|
|
argIndex: 2
|
|
}
|
|
transformerArgs {
|
|
name: "Elu"
|
|
int64Value: 8
|
|
argIndex: 2
|
|
}
|
|
transformerArgs {
|
|
name: "Softsign"
|
|
int64Value: 9
|
|
argIndex: 2
|
|
}
|
|
transformerArgs {
|
|
name: "Softplus"
|
|
int64Value: 10
|
|
argIndex: 2
|
|
}
|
|
}
|
|
transformerArgs {
|
|
key: "gateAct"
|
|
transformerArgs {
|
|
name: "Relu"
|
|
int64Value: 1
|
|
argIndex: 2
|
|
}
|
|
transformerArgs {
|
|
name: "Tanh"
|
|
argIndex: 2
|
|
}
|
|
transformerArgs {
|
|
name: "Sigmoid"
|
|
int64Value: 2
|
|
argIndex: 2
|
|
}
|
|
transformerArgs {
|
|
name: "Affine"
|
|
int64Value: 3
|
|
argIndex: 2
|
|
}
|
|
transformerArgs {
|
|
name: "LeakyRelu"
|
|
int64Value: 4
|
|
argIndex: 2
|
|
}
|
|
transformerArgs {
|
|
name: "ThresholdedRelu"
|
|
int64Value: 5
|
|
argIndex: 2
|
|
}
|
|
transformerArgs {
|
|
name: "ScaledTanh"
|
|
int64Value: 6
|
|
argIndex: 2
|
|
}
|
|
transformerArgs {
|
|
name: "HardSigmoid"
|
|
int64Value: 7
|
|
argIndex: 2
|
|
}
|
|
transformerArgs {
|
|
name: "Elu"
|
|
int64Value: 8
|
|
argIndex: 2
|
|
}
|
|
transformerArgs {
|
|
name: "Softsign"
|
|
int64Value: 9
|
|
argIndex: 2
|
|
}
|
|
transformerArgs {
|
|
name: "Softplus"
|
|
int64Value: 10
|
|
argIndex: 2
|
|
}
|
|
}
|
|
transformerArgs {
|
|
key: "gateAct"
|
|
transformerArgs {
|
|
name: "Relu"
|
|
int64Value: 1
|
|
argIndex: 2
|
|
}
|
|
transformerArgs {
|
|
name: "Tanh"
|
|
argIndex: 2
|
|
}
|
|
transformerArgs {
|
|
name: "Sigmoid"
|
|
int64Value: 2
|
|
argIndex: 2
|
|
}
|
|
transformerArgs {
|
|
name: "Affine"
|
|
int64Value: 3
|
|
argIndex: 2
|
|
}
|
|
transformerArgs {
|
|
name: "LeakyRelu"
|
|
int64Value: 4
|
|
argIndex: 2
|
|
}
|
|
transformerArgs {
|
|
name: "ThresholdedRelu"
|
|
int64Value: 5
|
|
argIndex: 2
|
|
}
|
|
transformerArgs {
|
|
name: "ScaledTanh"
|
|
int64Value: 6
|
|
argIndex: 2
|
|
}
|
|
transformerArgs {
|
|
name: "HardSigmoid"
|
|
int64Value: 7
|
|
argIndex: 2
|
|
}
|
|
transformerArgs {
|
|
name: "Elu"
|
|
int64Value: 8
|
|
argIndex: 2
|
|
}
|
|
transformerArgs {
|
|
name: "Softsign"
|
|
int64Value: 9
|
|
argIndex: 2
|
|
}
|
|
transformerArgs {
|
|
name: "Softplus"
|
|
int64Value: 10
|
|
argIndex: 2
|
|
}
|
|
}
|
|
transformerArgs {
|
|
key: "gateAct"
|
|
transformerArgs {
|
|
name: "Relu"
|
|
int64Value: 1
|
|
argIndex: 2
|
|
}
|
|
transformerArgs {
|
|
name: "Tanh"
|
|
argIndex: 2
|
|
}
|
|
transformerArgs {
|
|
name: "Sigmoid"
|
|
int64Value: 2
|
|
argIndex: 2
|
|
}
|
|
transformerArgs {
|
|
name: "Affine"
|
|
int64Value: 3
|
|
argIndex: 2
|
|
}
|
|
transformerArgs {
|
|
name: "LeakyRelu"
|
|
int64Value: 4
|
|
argIndex: 2
|
|
}
|
|
transformerArgs {
|
|
name: "ThresholdedRelu"
|
|
int64Value: 5
|
|
argIndex: 2
|
|
}
|
|
transformerArgs {
|
|
name: "ScaledTanh"
|
|
int64Value: 6
|
|
argIndex: 2
|
|
}
|
|
transformerArgs {
|
|
name: "HardSigmoid"
|
|
int64Value: 7
|
|
argIndex: 2
|
|
}
|
|
transformerArgs {
|
|
name: "Elu"
|
|
int64Value: 8
|
|
argIndex: 2
|
|
}
|
|
transformerArgs {
|
|
name: "Softsign"
|
|
int64Value: 9
|
|
argIndex: 2
|
|
}
|
|
transformerArgs {
|
|
name: "Softplus"
|
|
int64Value: 10
|
|
argIndex: 2
|
|
}
|
|
}
|
|
transformerArgs {
|
|
key: "gateAct"
|
|
transformerArgs {
|
|
name: "Relu"
|
|
int64Value: 1
|
|
argIndex: 2
|
|
}
|
|
transformerArgs {
|
|
name: "Tanh"
|
|
argIndex: 2
|
|
}
|
|
transformerArgs {
|
|
name: "Sigmoid"
|
|
int64Value: 2
|
|
argIndex: 2
|
|
}
|
|
transformerArgs {
|
|
name: "Affine"
|
|
int64Value: 3
|
|
argIndex: 2
|
|
}
|
|
transformerArgs {
|
|
name: "LeakyRelu"
|
|
int64Value: 4
|
|
argIndex: 2
|
|
}
|
|
transformerArgs {
|
|
name: "ThresholdedRelu"
|
|
int64Value: 5
|
|
argIndex: 2
|
|
}
|
|
transformerArgs {
|
|
name: "ScaledTanh"
|
|
int64Value: 6
|
|
argIndex: 2
|
|
}
|
|
transformerArgs {
|
|
name: "HardSigmoid"
|
|
int64Value: 7
|
|
argIndex: 2
|
|
}
|
|
transformerArgs {
|
|
name: "Elu"
|
|
int64Value: 8
|
|
argIndex: 2
|
|
}
|
|
transformerArgs {
|
|
name: "Softsign"
|
|
int64Value: 9
|
|
argIndex: 2
|
|
}
|
|
transformerArgs {
|
|
name: "Softplus"
|
|
int64Value: 10
|
|
argIndex: 2
|
|
}
|
|
}
|
|
transformerArgs {
|
|
key: "index"
|
|
transformerArgs {
|
|
name: "index"
|
|
}
|
|
}
|
|
inputFrameworkOpName: "LSTM"
|
|
}
|
|
rule {
|
|
ruleName: "mapstringtoindex"
|
|
functionName: "mapstringtoindex"
|
|
outputIntName: "cellAct"
|
|
inputFloatName: "Relu"
|
|
inputFloatName: "Tanh"
|
|
inputFloatName: "Sigmoid"
|
|
inputFloatName: "Affine"
|
|
inputFloatName: "LeakyRelu"
|
|
inputFloatName: "ThresholdedRelu"
|
|
inputFloatName: "ScaledTanh"
|
|
inputFloatName: "HardSigmoid"
|
|
inputFloatName: "Elu"
|
|
inputFloatName: "Softsign"
|
|
inputFloatName: "Softplus"
|
|
inputFloatName: "index"
|
|
inputToOutput {
|
|
key: "cellAct"
|
|
value: "activations"
|
|
}
|
|
ruleType: "attribute"
|
|
transformerArgs {
|
|
key: "cellAct"
|
|
transformerArgs {
|
|
name: "Relu"
|
|
int64Value: 1
|
|
argIndex: 3
|
|
}
|
|
transformerArgs {
|
|
name: "Tanh"
|
|
argIndex: 3
|
|
}
|
|
transformerArgs {
|
|
name: "Sigmoid"
|
|
int64Value: 2
|
|
argIndex: 3
|
|
}
|
|
transformerArgs {
|
|
name: "Affine"
|
|
int64Value: 3
|
|
argIndex: 3
|
|
}
|
|
transformerArgs {
|
|
name: "LeakyRelu"
|
|
int64Value: 4
|
|
argIndex: 3
|
|
}
|
|
transformerArgs {
|
|
name: "ThresholdedRelu"
|
|
int64Value: 5
|
|
argIndex: 3
|
|
}
|
|
transformerArgs {
|
|
name: "ScaledTanh"
|
|
int64Value: 6
|
|
argIndex: 3
|
|
}
|
|
transformerArgs {
|
|
name: "HardSigmoid"
|
|
int64Value: 7
|
|
argIndex: 3
|
|
}
|
|
transformerArgs {
|
|
name: "Elu"
|
|
int64Value: 8
|
|
argIndex: 3
|
|
}
|
|
transformerArgs {
|
|
name: "Softsign"
|
|
int64Value: 9
|
|
argIndex: 3
|
|
}
|
|
transformerArgs {
|
|
name: "Softplus"
|
|
int64Value: 10
|
|
argIndex: 3
|
|
}
|
|
}
|
|
transformerArgs {
|
|
key: "cellAct"
|
|
transformerArgs {
|
|
name: "Relu"
|
|
int64Value: 1
|
|
argIndex: 3
|
|
}
|
|
transformerArgs {
|
|
name: "Tanh"
|
|
argIndex: 3
|
|
}
|
|
transformerArgs {
|
|
name: "Sigmoid"
|
|
int64Value: 2
|
|
argIndex: 3
|
|
}
|
|
transformerArgs {
|
|
name: "Affine"
|
|
int64Value: 3
|
|
argIndex: 3
|
|
}
|
|
transformerArgs {
|
|
name: "LeakyRelu"
|
|
int64Value: 4
|
|
argIndex: 3
|
|
}
|
|
transformerArgs {
|
|
name: "ThresholdedRelu"
|
|
int64Value: 5
|
|
argIndex: 3
|
|
}
|
|
transformerArgs {
|
|
name: "ScaledTanh"
|
|
int64Value: 6
|
|
argIndex: 3
|
|
}
|
|
transformerArgs {
|
|
name: "HardSigmoid"
|
|
int64Value: 7
|
|
argIndex: 3
|
|
}
|
|
transformerArgs {
|
|
name: "Elu"
|
|
int64Value: 8
|
|
argIndex: 3
|
|
}
|
|
transformerArgs {
|
|
name: "Softsign"
|
|
int64Value: 9
|
|
argIndex: 3
|
|
}
|
|
transformerArgs {
|
|
name: "Softplus"
|
|
int64Value: 10
|
|
argIndex: 3
|
|
}
|
|
}
|
|
transformerArgs {
|
|
key: "cellAct"
|
|
transformerArgs {
|
|
name: "Relu"
|
|
int64Value: 1
|
|
argIndex: 3
|
|
}
|
|
transformerArgs {
|
|
name: "Tanh"
|
|
argIndex: 3
|
|
}
|
|
transformerArgs {
|
|
name: "Sigmoid"
|
|
int64Value: 2
|
|
argIndex: 3
|
|
}
|
|
transformerArgs {
|
|
name: "Affine"
|
|
int64Value: 3
|
|
argIndex: 3
|
|
}
|
|
transformerArgs {
|
|
name: "LeakyRelu"
|
|
int64Value: 4
|
|
argIndex: 3
|
|
}
|
|
transformerArgs {
|
|
name: "ThresholdedRelu"
|
|
int64Value: 5
|
|
argIndex: 3
|
|
}
|
|
transformerArgs {
|
|
name: "ScaledTanh"
|
|
int64Value: 6
|
|
argIndex: 3
|
|
}
|
|
transformerArgs {
|
|
name: "HardSigmoid"
|
|
int64Value: 7
|
|
argIndex: 3
|
|
}
|
|
transformerArgs {
|
|
name: "Elu"
|
|
int64Value: 8
|
|
argIndex: 3
|
|
}
|
|
transformerArgs {
|
|
name: "Softsign"
|
|
int64Value: 9
|
|
argIndex: 3
|
|
}
|
|
transformerArgs {
|
|
name: "Softplus"
|
|
int64Value: 10
|
|
argIndex: 3
|
|
}
|
|
}
|
|
transformerArgs {
|
|
key: "cellAct"
|
|
transformerArgs {
|
|
name: "Relu"
|
|
int64Value: 1
|
|
argIndex: 3
|
|
}
|
|
transformerArgs {
|
|
name: "Tanh"
|
|
argIndex: 3
|
|
}
|
|
transformerArgs {
|
|
name: "Sigmoid"
|
|
int64Value: 2
|
|
argIndex: 3
|
|
}
|
|
transformerArgs {
|
|
name: "Affine"
|
|
int64Value: 3
|
|
argIndex: 3
|
|
}
|
|
transformerArgs {
|
|
name: "LeakyRelu"
|
|
int64Value: 4
|
|
argIndex: 3
|
|
}
|
|
transformerArgs {
|
|
name: "ThresholdedRelu"
|
|
int64Value: 5
|
|
argIndex: 3
|
|
}
|
|
transformerArgs {
|
|
name: "ScaledTanh"
|
|
int64Value: 6
|
|
argIndex: 3
|
|
}
|
|
transformerArgs {
|
|
name: "HardSigmoid"
|
|
int64Value: 7
|
|
argIndex: 3
|
|
}
|
|
transformerArgs {
|
|
name: "Elu"
|
|
int64Value: 8
|
|
argIndex: 3
|
|
}
|
|
transformerArgs {
|
|
name: "Softsign"
|
|
int64Value: 9
|
|
argIndex: 3
|
|
}
|
|
transformerArgs {
|
|
name: "Softplus"
|
|
int64Value: 10
|
|
argIndex: 3
|
|
}
|
|
}
|
|
transformerArgs {
|
|
key: "cellAct"
|
|
transformerArgs {
|
|
name: "Relu"
|
|
int64Value: 1
|
|
argIndex: 3
|
|
}
|
|
transformerArgs {
|
|
name: "Tanh"
|
|
argIndex: 3
|
|
}
|
|
transformerArgs {
|
|
name: "Sigmoid"
|
|
int64Value: 2
|
|
argIndex: 3
|
|
}
|
|
transformerArgs {
|
|
name: "Affine"
|
|
int64Value: 3
|
|
argIndex: 3
|
|
}
|
|
transformerArgs {
|
|
name: "LeakyRelu"
|
|
int64Value: 4
|
|
argIndex: 3
|
|
}
|
|
transformerArgs {
|
|
name: "ThresholdedRelu"
|
|
int64Value: 5
|
|
argIndex: 3
|
|
}
|
|
transformerArgs {
|
|
name: "ScaledTanh"
|
|
int64Value: 6
|
|
argIndex: 3
|
|
}
|
|
transformerArgs {
|
|
name: "HardSigmoid"
|
|
int64Value: 7
|
|
argIndex: 3
|
|
}
|
|
transformerArgs {
|
|
name: "Elu"
|
|
int64Value: 8
|
|
argIndex: 3
|
|
}
|
|
transformerArgs {
|
|
name: "Softsign"
|
|
int64Value: 9
|
|
argIndex: 3
|
|
}
|
|
transformerArgs {
|
|
name: "Softplus"
|
|
int64Value: 10
|
|
argIndex: 3
|
|
}
|
|
}
|
|
transformerArgs {
|
|
key: "cellAct"
|
|
transformerArgs {
|
|
name: "Relu"
|
|
int64Value: 1
|
|
argIndex: 3
|
|
}
|
|
transformerArgs {
|
|
name: "Tanh"
|
|
argIndex: 3
|
|
}
|
|
transformerArgs {
|
|
name: "Sigmoid"
|
|
int64Value: 2
|
|
argIndex: 3
|
|
}
|
|
transformerArgs {
|
|
name: "Affine"
|
|
int64Value: 3
|
|
argIndex: 3
|
|
}
|
|
transformerArgs {
|
|
name: "LeakyRelu"
|
|
int64Value: 4
|
|
argIndex: 3
|
|
}
|
|
transformerArgs {
|
|
name: "ThresholdedRelu"
|
|
int64Value: 5
|
|
argIndex: 3
|
|
}
|
|
transformerArgs {
|
|
name: "ScaledTanh"
|
|
int64Value: 6
|
|
argIndex: 3
|
|
}
|
|
transformerArgs {
|
|
name: "HardSigmoid"
|
|
int64Value: 7
|
|
argIndex: 3
|
|
}
|
|
transformerArgs {
|
|
name: "Elu"
|
|
int64Value: 8
|
|
argIndex: 3
|
|
}
|
|
transformerArgs {
|
|
name: "Softsign"
|
|
int64Value: 9
|
|
argIndex: 3
|
|
}
|
|
transformerArgs {
|
|
name: "Softplus"
|
|
int64Value: 10
|
|
argIndex: 3
|
|
}
|
|
}
|
|
transformerArgs {
|
|
key: "cellAct"
|
|
transformerArgs {
|
|
name: "Relu"
|
|
int64Value: 1
|
|
argIndex: 3
|
|
}
|
|
transformerArgs {
|
|
name: "Tanh"
|
|
argIndex: 3
|
|
}
|
|
transformerArgs {
|
|
name: "Sigmoid"
|
|
int64Value: 2
|
|
argIndex: 3
|
|
}
|
|
transformerArgs {
|
|
name: "Affine"
|
|
int64Value: 3
|
|
argIndex: 3
|
|
}
|
|
transformerArgs {
|
|
name: "LeakyRelu"
|
|
int64Value: 4
|
|
argIndex: 3
|
|
}
|
|
transformerArgs {
|
|
name: "ThresholdedRelu"
|
|
int64Value: 5
|
|
argIndex: 3
|
|
}
|
|
transformerArgs {
|
|
name: "ScaledTanh"
|
|
int64Value: 6
|
|
argIndex: 3
|
|
}
|
|
transformerArgs {
|
|
name: "HardSigmoid"
|
|
int64Value: 7
|
|
argIndex: 3
|
|
}
|
|
transformerArgs {
|
|
name: "Elu"
|
|
int64Value: 8
|
|
argIndex: 3
|
|
}
|
|
transformerArgs {
|
|
name: "Softsign"
|
|
int64Value: 9
|
|
argIndex: 3
|
|
}
|
|
transformerArgs {
|
|
name: "Softplus"
|
|
int64Value: 10
|
|
argIndex: 3
|
|
}
|
|
}
|
|
transformerArgs {
|
|
key: "cellAct"
|
|
transformerArgs {
|
|
name: "Relu"
|
|
int64Value: 1
|
|
argIndex: 3
|
|
}
|
|
transformerArgs {
|
|
name: "Tanh"
|
|
argIndex: 3
|
|
}
|
|
transformerArgs {
|
|
name: "Sigmoid"
|
|
int64Value: 2
|
|
argIndex: 3
|
|
}
|
|
transformerArgs {
|
|
name: "Affine"
|
|
int64Value: 3
|
|
argIndex: 3
|
|
}
|
|
transformerArgs {
|
|
name: "LeakyRelu"
|
|
int64Value: 4
|
|
argIndex: 3
|
|
}
|
|
transformerArgs {
|
|
name: "ThresholdedRelu"
|
|
int64Value: 5
|
|
argIndex: 3
|
|
}
|
|
transformerArgs {
|
|
name: "ScaledTanh"
|
|
int64Value: 6
|
|
argIndex: 3
|
|
}
|
|
transformerArgs {
|
|
name: "HardSigmoid"
|
|
int64Value: 7
|
|
argIndex: 3
|
|
}
|
|
transformerArgs {
|
|
name: "Elu"
|
|
int64Value: 8
|
|
argIndex: 3
|
|
}
|
|
transformerArgs {
|
|
name: "Softsign"
|
|
int64Value: 9
|
|
argIndex: 3
|
|
}
|
|
transformerArgs {
|
|
name: "Softplus"
|
|
int64Value: 10
|
|
argIndex: 3
|
|
}
|
|
}
|
|
transformerArgs {
|
|
key: "cellAct"
|
|
transformerArgs {
|
|
name: "Relu"
|
|
int64Value: 1
|
|
argIndex: 3
|
|
}
|
|
transformerArgs {
|
|
name: "Tanh"
|
|
argIndex: 3
|
|
}
|
|
transformerArgs {
|
|
name: "Sigmoid"
|
|
int64Value: 2
|
|
argIndex: 3
|
|
}
|
|
transformerArgs {
|
|
name: "Affine"
|
|
int64Value: 3
|
|
argIndex: 3
|
|
}
|
|
transformerArgs {
|
|
name: "LeakyRelu"
|
|
int64Value: 4
|
|
argIndex: 3
|
|
}
|
|
transformerArgs {
|
|
name: "ThresholdedRelu"
|
|
int64Value: 5
|
|
argIndex: 3
|
|
}
|
|
transformerArgs {
|
|
name: "ScaledTanh"
|
|
int64Value: 6
|
|
argIndex: 3
|
|
}
|
|
transformerArgs {
|
|
name: "HardSigmoid"
|
|
int64Value: 7
|
|
argIndex: 3
|
|
}
|
|
transformerArgs {
|
|
name: "Elu"
|
|
int64Value: 8
|
|
argIndex: 3
|
|
}
|
|
transformerArgs {
|
|
name: "Softsign"
|
|
int64Value: 9
|
|
argIndex: 3
|
|
}
|
|
transformerArgs {
|
|
name: "Softplus"
|
|
int64Value: 10
|
|
argIndex: 3
|
|
}
|
|
}
|
|
transformerArgs {
|
|
key: "cellAct"
|
|
transformerArgs {
|
|
name: "Relu"
|
|
int64Value: 1
|
|
argIndex: 3
|
|
}
|
|
transformerArgs {
|
|
name: "Tanh"
|
|
argIndex: 3
|
|
}
|
|
transformerArgs {
|
|
name: "Sigmoid"
|
|
int64Value: 2
|
|
argIndex: 3
|
|
}
|
|
transformerArgs {
|
|
name: "Affine"
|
|
int64Value: 3
|
|
argIndex: 3
|
|
}
|
|
transformerArgs {
|
|
name: "LeakyRelu"
|
|
int64Value: 4
|
|
argIndex: 3
|
|
}
|
|
transformerArgs {
|
|
name: "ThresholdedRelu"
|
|
int64Value: 5
|
|
argIndex: 3
|
|
}
|
|
transformerArgs {
|
|
name: "ScaledTanh"
|
|
int64Value: 6
|
|
argIndex: 3
|
|
}
|
|
transformerArgs {
|
|
name: "HardSigmoid"
|
|
int64Value: 7
|
|
argIndex: 3
|
|
}
|
|
transformerArgs {
|
|
name: "Elu"
|
|
int64Value: 8
|
|
argIndex: 3
|
|
}
|
|
transformerArgs {
|
|
name: "Softsign"
|
|
int64Value: 9
|
|
argIndex: 3
|
|
}
|
|
transformerArgs {
|
|
name: "Softplus"
|
|
int64Value: 10
|
|
argIndex: 3
|
|
}
|
|
}
|
|
transformerArgs {
|
|
key: "cellAct"
|
|
transformerArgs {
|
|
name: "Relu"
|
|
int64Value: 1
|
|
argIndex: 3
|
|
}
|
|
transformerArgs {
|
|
name: "Tanh"
|
|
argIndex: 3
|
|
}
|
|
transformerArgs {
|
|
name: "Sigmoid"
|
|
int64Value: 2
|
|
argIndex: 3
|
|
}
|
|
transformerArgs {
|
|
name: "Affine"
|
|
int64Value: 3
|
|
argIndex: 3
|
|
}
|
|
transformerArgs {
|
|
name: "LeakyRelu"
|
|
int64Value: 4
|
|
argIndex: 3
|
|
}
|
|
transformerArgs {
|
|
name: "ThresholdedRelu"
|
|
int64Value: 5
|
|
argIndex: 3
|
|
}
|
|
transformerArgs {
|
|
name: "ScaledTanh"
|
|
int64Value: 6
|
|
argIndex: 3
|
|
}
|
|
transformerArgs {
|
|
name: "HardSigmoid"
|
|
int64Value: 7
|
|
argIndex: 3
|
|
}
|
|
transformerArgs {
|
|
name: "Elu"
|
|
int64Value: 8
|
|
argIndex: 3
|
|
}
|
|
transformerArgs {
|
|
name: "Softsign"
|
|
int64Value: 9
|
|
argIndex: 3
|
|
}
|
|
transformerArgs {
|
|
name: "Softplus"
|
|
int64Value: 10
|
|
argIndex: 3
|
|
}
|
|
}
|
|
transformerArgs {
|
|
key: "index"
|
|
transformerArgs {
|
|
name: "index"
|
|
int64Value: 1
|
|
}
|
|
}
|
|
inputFrameworkOpName: "LSTM"
|
|
}
|
|
rule {
|
|
ruleName: "mapstringtoindex"
|
|
functionName: "mapstringtoindex"
|
|
outputIntName: "outAct"
|
|
inputFloatName: "Relu"
|
|
inputFloatName: "Tanh"
|
|
inputFloatName: "Sigmoid"
|
|
inputFloatName: "Affine"
|
|
inputFloatName: "LeakyRelu"
|
|
inputFloatName: "ThresholdedRelu"
|
|
inputFloatName: "ScaledTanh"
|
|
inputFloatName: "HardSigmoid"
|
|
inputFloatName: "Elu"
|
|
inputFloatName: "Softsign"
|
|
inputFloatName: "Softplus"
|
|
inputFloatName: "index"
|
|
inputToOutput {
|
|
key: "outAct"
|
|
value: "activations"
|
|
}
|
|
ruleType: "attribute"
|
|
transformerArgs {
|
|
key: "outAct"
|
|
transformerArgs {
|
|
name: "Relu"
|
|
int64Value: 1
|
|
argIndex: 4
|
|
}
|
|
transformerArgs {
|
|
name: "Tanh"
|
|
argIndex: 4
|
|
}
|
|
transformerArgs {
|
|
name: "Sigmoid"
|
|
int64Value: 2
|
|
argIndex: 4
|
|
}
|
|
transformerArgs {
|
|
name: "Affine"
|
|
int64Value: 3
|
|
argIndex: 4
|
|
}
|
|
transformerArgs {
|
|
name: "LeakyRelu"
|
|
int64Value: 4
|
|
argIndex: 4
|
|
}
|
|
transformerArgs {
|
|
name: "ThresholdedRelu"
|
|
int64Value: 5
|
|
argIndex: 4
|
|
}
|
|
transformerArgs {
|
|
name: "ScaledTanh"
|
|
int64Value: 6
|
|
argIndex: 4
|
|
}
|
|
transformerArgs {
|
|
name: "HardSigmoid"
|
|
int64Value: 7
|
|
argIndex: 4
|
|
}
|
|
transformerArgs {
|
|
name: "Elu"
|
|
int64Value: 8
|
|
argIndex: 4
|
|
}
|
|
transformerArgs {
|
|
name: "Softsign"
|
|
int64Value: 9
|
|
argIndex: 4
|
|
}
|
|
transformerArgs {
|
|
name: "Softplus"
|
|
int64Value: 10
|
|
argIndex: 4
|
|
}
|
|
}
|
|
transformerArgs {
|
|
key: "outAct"
|
|
transformerArgs {
|
|
name: "Relu"
|
|
int64Value: 1
|
|
argIndex: 4
|
|
}
|
|
transformerArgs {
|
|
name: "Tanh"
|
|
argIndex: 4
|
|
}
|
|
transformerArgs {
|
|
name: "Sigmoid"
|
|
int64Value: 2
|
|
argIndex: 4
|
|
}
|
|
transformerArgs {
|
|
name: "Affine"
|
|
int64Value: 3
|
|
argIndex: 4
|
|
}
|
|
transformerArgs {
|
|
name: "LeakyRelu"
|
|
int64Value: 4
|
|
argIndex: 4
|
|
}
|
|
transformerArgs {
|
|
name: "ThresholdedRelu"
|
|
int64Value: 5
|
|
argIndex: 4
|
|
}
|
|
transformerArgs {
|
|
name: "ScaledTanh"
|
|
int64Value: 6
|
|
argIndex: 4
|
|
}
|
|
transformerArgs {
|
|
name: "HardSigmoid"
|
|
int64Value: 7
|
|
argIndex: 4
|
|
}
|
|
transformerArgs {
|
|
name: "Elu"
|
|
int64Value: 8
|
|
argIndex: 4
|
|
}
|
|
transformerArgs {
|
|
name: "Softsign"
|
|
int64Value: 9
|
|
argIndex: 4
|
|
}
|
|
transformerArgs {
|
|
name: "Softplus"
|
|
int64Value: 10
|
|
argIndex: 4
|
|
}
|
|
}
|
|
transformerArgs {
|
|
key: "outAct"
|
|
transformerArgs {
|
|
name: "Relu"
|
|
int64Value: 1
|
|
argIndex: 4
|
|
}
|
|
transformerArgs {
|
|
name: "Tanh"
|
|
argIndex: 4
|
|
}
|
|
transformerArgs {
|
|
name: "Sigmoid"
|
|
int64Value: 2
|
|
argIndex: 4
|
|
}
|
|
transformerArgs {
|
|
name: "Affine"
|
|
int64Value: 3
|
|
argIndex: 4
|
|
}
|
|
transformerArgs {
|
|
name: "LeakyRelu"
|
|
int64Value: 4
|
|
argIndex: 4
|
|
}
|
|
transformerArgs {
|
|
name: "ThresholdedRelu"
|
|
int64Value: 5
|
|
argIndex: 4
|
|
}
|
|
transformerArgs {
|
|
name: "ScaledTanh"
|
|
int64Value: 6
|
|
argIndex: 4
|
|
}
|
|
transformerArgs {
|
|
name: "HardSigmoid"
|
|
int64Value: 7
|
|
argIndex: 4
|
|
}
|
|
transformerArgs {
|
|
name: "Elu"
|
|
int64Value: 8
|
|
argIndex: 4
|
|
}
|
|
transformerArgs {
|
|
name: "Softsign"
|
|
int64Value: 9
|
|
argIndex: 4
|
|
}
|
|
transformerArgs {
|
|
name: "Softplus"
|
|
int64Value: 10
|
|
argIndex: 4
|
|
}
|
|
}
|
|
transformerArgs {
|
|
key: "outAct"
|
|
transformerArgs {
|
|
name: "Relu"
|
|
int64Value: 1
|
|
argIndex: 4
|
|
}
|
|
transformerArgs {
|
|
name: "Tanh"
|
|
argIndex: 4
|
|
}
|
|
transformerArgs {
|
|
name: "Sigmoid"
|
|
int64Value: 2
|
|
argIndex: 4
|
|
}
|
|
transformerArgs {
|
|
name: "Affine"
|
|
int64Value: 3
|
|
argIndex: 4
|
|
}
|
|
transformerArgs {
|
|
name: "LeakyRelu"
|
|
int64Value: 4
|
|
argIndex: 4
|
|
}
|
|
transformerArgs {
|
|
name: "ThresholdedRelu"
|
|
int64Value: 5
|
|
argIndex: 4
|
|
}
|
|
transformerArgs {
|
|
name: "ScaledTanh"
|
|
int64Value: 6
|
|
argIndex: 4
|
|
}
|
|
transformerArgs {
|
|
name: "HardSigmoid"
|
|
int64Value: 7
|
|
argIndex: 4
|
|
}
|
|
transformerArgs {
|
|
name: "Elu"
|
|
int64Value: 8
|
|
argIndex: 4
|
|
}
|
|
transformerArgs {
|
|
name: "Softsign"
|
|
int64Value: 9
|
|
argIndex: 4
|
|
}
|
|
transformerArgs {
|
|
name: "Softplus"
|
|
int64Value: 10
|
|
argIndex: 4
|
|
}
|
|
}
|
|
transformerArgs {
|
|
key: "outAct"
|
|
transformerArgs {
|
|
name: "Relu"
|
|
int64Value: 1
|
|
argIndex: 4
|
|
}
|
|
transformerArgs {
|
|
name: "Tanh"
|
|
argIndex: 4
|
|
}
|
|
transformerArgs {
|
|
name: "Sigmoid"
|
|
int64Value: 2
|
|
argIndex: 4
|
|
}
|
|
transformerArgs {
|
|
name: "Affine"
|
|
int64Value: 3
|
|
argIndex: 4
|
|
}
|
|
transformerArgs {
|
|
name: "LeakyRelu"
|
|
int64Value: 4
|
|
argIndex: 4
|
|
}
|
|
transformerArgs {
|
|
name: "ThresholdedRelu"
|
|
int64Value: 5
|
|
argIndex: 4
|
|
}
|
|
transformerArgs {
|
|
name: "ScaledTanh"
|
|
int64Value: 6
|
|
argIndex: 4
|
|
}
|
|
transformerArgs {
|
|
name: "HardSigmoid"
|
|
int64Value: 7
|
|
argIndex: 4
|
|
}
|
|
transformerArgs {
|
|
name: "Elu"
|
|
int64Value: 8
|
|
argIndex: 4
|
|
}
|
|
transformerArgs {
|
|
name: "Softsign"
|
|
int64Value: 9
|
|
argIndex: 4
|
|
}
|
|
transformerArgs {
|
|
name: "Softplus"
|
|
int64Value: 10
|
|
argIndex: 4
|
|
}
|
|
}
|
|
transformerArgs {
|
|
key: "outAct"
|
|
transformerArgs {
|
|
name: "Relu"
|
|
int64Value: 1
|
|
argIndex: 4
|
|
}
|
|
transformerArgs {
|
|
name: "Tanh"
|
|
argIndex: 4
|
|
}
|
|
transformerArgs {
|
|
name: "Sigmoid"
|
|
int64Value: 2
|
|
argIndex: 4
|
|
}
|
|
transformerArgs {
|
|
name: "Affine"
|
|
int64Value: 3
|
|
argIndex: 4
|
|
}
|
|
transformerArgs {
|
|
name: "LeakyRelu"
|
|
int64Value: 4
|
|
argIndex: 4
|
|
}
|
|
transformerArgs {
|
|
name: "ThresholdedRelu"
|
|
int64Value: 5
|
|
argIndex: 4
|
|
}
|
|
transformerArgs {
|
|
name: "ScaledTanh"
|
|
int64Value: 6
|
|
argIndex: 4
|
|
}
|
|
transformerArgs {
|
|
name: "HardSigmoid"
|
|
int64Value: 7
|
|
argIndex: 4
|
|
}
|
|
transformerArgs {
|
|
name: "Elu"
|
|
int64Value: 8
|
|
argIndex: 4
|
|
}
|
|
transformerArgs {
|
|
name: "Softsign"
|
|
int64Value: 9
|
|
argIndex: 4
|
|
}
|
|
transformerArgs {
|
|
name: "Softplus"
|
|
int64Value: 10
|
|
argIndex: 4
|
|
}
|
|
}
|
|
transformerArgs {
|
|
key: "outAct"
|
|
transformerArgs {
|
|
name: "Relu"
|
|
int64Value: 1
|
|
argIndex: 4
|
|
}
|
|
transformerArgs {
|
|
name: "Tanh"
|
|
argIndex: 4
|
|
}
|
|
transformerArgs {
|
|
name: "Sigmoid"
|
|
int64Value: 2
|
|
argIndex: 4
|
|
}
|
|
transformerArgs {
|
|
name: "Affine"
|
|
int64Value: 3
|
|
argIndex: 4
|
|
}
|
|
transformerArgs {
|
|
name: "LeakyRelu"
|
|
int64Value: 4
|
|
argIndex: 4
|
|
}
|
|
transformerArgs {
|
|
name: "ThresholdedRelu"
|
|
int64Value: 5
|
|
argIndex: 4
|
|
}
|
|
transformerArgs {
|
|
name: "ScaledTanh"
|
|
int64Value: 6
|
|
argIndex: 4
|
|
}
|
|
transformerArgs {
|
|
name: "HardSigmoid"
|
|
int64Value: 7
|
|
argIndex: 4
|
|
}
|
|
transformerArgs {
|
|
name: "Elu"
|
|
int64Value: 8
|
|
argIndex: 4
|
|
}
|
|
transformerArgs {
|
|
name: "Softsign"
|
|
int64Value: 9
|
|
argIndex: 4
|
|
}
|
|
transformerArgs {
|
|
name: "Softplus"
|
|
int64Value: 10
|
|
argIndex: 4
|
|
}
|
|
}
|
|
transformerArgs {
|
|
key: "outAct"
|
|
transformerArgs {
|
|
name: "Relu"
|
|
int64Value: 1
|
|
argIndex: 4
|
|
}
|
|
transformerArgs {
|
|
name: "Tanh"
|
|
argIndex: 4
|
|
}
|
|
transformerArgs {
|
|
name: "Sigmoid"
|
|
int64Value: 2
|
|
argIndex: 4
|
|
}
|
|
transformerArgs {
|
|
name: "Affine"
|
|
int64Value: 3
|
|
argIndex: 4
|
|
}
|
|
transformerArgs {
|
|
name: "LeakyRelu"
|
|
int64Value: 4
|
|
argIndex: 4
|
|
}
|
|
transformerArgs {
|
|
name: "ThresholdedRelu"
|
|
int64Value: 5
|
|
argIndex: 4
|
|
}
|
|
transformerArgs {
|
|
name: "ScaledTanh"
|
|
int64Value: 6
|
|
argIndex: 4
|
|
}
|
|
transformerArgs {
|
|
name: "HardSigmoid"
|
|
int64Value: 7
|
|
argIndex: 4
|
|
}
|
|
transformerArgs {
|
|
name: "Elu"
|
|
int64Value: 8
|
|
argIndex: 4
|
|
}
|
|
transformerArgs {
|
|
name: "Softsign"
|
|
int64Value: 9
|
|
argIndex: 4
|
|
}
|
|
transformerArgs {
|
|
name: "Softplus"
|
|
int64Value: 10
|
|
argIndex: 4
|
|
}
|
|
}
|
|
transformerArgs {
|
|
key: "outAct"
|
|
transformerArgs {
|
|
name: "Relu"
|
|
int64Value: 1
|
|
argIndex: 4
|
|
}
|
|
transformerArgs {
|
|
name: "Tanh"
|
|
argIndex: 4
|
|
}
|
|
transformerArgs {
|
|
name: "Sigmoid"
|
|
int64Value: 2
|
|
argIndex: 4
|
|
}
|
|
transformerArgs {
|
|
name: "Affine"
|
|
int64Value: 3
|
|
argIndex: 4
|
|
}
|
|
transformerArgs {
|
|
name: "LeakyRelu"
|
|
int64Value: 4
|
|
argIndex: 4
|
|
}
|
|
transformerArgs {
|
|
name: "ThresholdedRelu"
|
|
int64Value: 5
|
|
argIndex: 4
|
|
}
|
|
transformerArgs {
|
|
name: "ScaledTanh"
|
|
int64Value: 6
|
|
argIndex: 4
|
|
}
|
|
transformerArgs {
|
|
name: "HardSigmoid"
|
|
int64Value: 7
|
|
argIndex: 4
|
|
}
|
|
transformerArgs {
|
|
name: "Elu"
|
|
int64Value: 8
|
|
argIndex: 4
|
|
}
|
|
transformerArgs {
|
|
name: "Softsign"
|
|
int64Value: 9
|
|
argIndex: 4
|
|
}
|
|
transformerArgs {
|
|
name: "Softplus"
|
|
int64Value: 10
|
|
argIndex: 4
|
|
}
|
|
}
|
|
transformerArgs {
|
|
key: "outAct"
|
|
transformerArgs {
|
|
name: "Relu"
|
|
int64Value: 1
|
|
argIndex: 4
|
|
}
|
|
transformerArgs {
|
|
name: "Tanh"
|
|
argIndex: 4
|
|
}
|
|
transformerArgs {
|
|
name: "Sigmoid"
|
|
int64Value: 2
|
|
argIndex: 4
|
|
}
|
|
transformerArgs {
|
|
name: "Affine"
|
|
int64Value: 3
|
|
argIndex: 4
|
|
}
|
|
transformerArgs {
|
|
name: "LeakyRelu"
|
|
int64Value: 4
|
|
argIndex: 4
|
|
}
|
|
transformerArgs {
|
|
name: "ThresholdedRelu"
|
|
int64Value: 5
|
|
argIndex: 4
|
|
}
|
|
transformerArgs {
|
|
name: "ScaledTanh"
|
|
int64Value: 6
|
|
argIndex: 4
|
|
}
|
|
transformerArgs {
|
|
name: "HardSigmoid"
|
|
int64Value: 7
|
|
argIndex: 4
|
|
}
|
|
transformerArgs {
|
|
name: "Elu"
|
|
int64Value: 8
|
|
argIndex: 4
|
|
}
|
|
transformerArgs {
|
|
name: "Softsign"
|
|
int64Value: 9
|
|
argIndex: 4
|
|
}
|
|
transformerArgs {
|
|
name: "Softplus"
|
|
int64Value: 10
|
|
argIndex: 4
|
|
}
|
|
}
|
|
transformerArgs {
|
|
key: "outAct"
|
|
transformerArgs {
|
|
name: "Relu"
|
|
int64Value: 1
|
|
argIndex: 4
|
|
}
|
|
transformerArgs {
|
|
name: "Tanh"
|
|
argIndex: 4
|
|
}
|
|
transformerArgs {
|
|
name: "Sigmoid"
|
|
int64Value: 2
|
|
argIndex: 4
|
|
}
|
|
transformerArgs {
|
|
name: "Affine"
|
|
int64Value: 3
|
|
argIndex: 4
|
|
}
|
|
transformerArgs {
|
|
name: "LeakyRelu"
|
|
int64Value: 4
|
|
argIndex: 4
|
|
}
|
|
transformerArgs {
|
|
name: "ThresholdedRelu"
|
|
int64Value: 5
|
|
argIndex: 4
|
|
}
|
|
transformerArgs {
|
|
name: "ScaledTanh"
|
|
int64Value: 6
|
|
argIndex: 4
|
|
}
|
|
transformerArgs {
|
|
name: "HardSigmoid"
|
|
int64Value: 7
|
|
argIndex: 4
|
|
}
|
|
transformerArgs {
|
|
name: "Elu"
|
|
int64Value: 8
|
|
argIndex: 4
|
|
}
|
|
transformerArgs {
|
|
name: "Softsign"
|
|
int64Value: 9
|
|
argIndex: 4
|
|
}
|
|
transformerArgs {
|
|
name: "Softplus"
|
|
int64Value: 10
|
|
argIndex: 4
|
|
}
|
|
}
|
|
transformerArgs {
|
|
key: "index"
|
|
transformerArgs {
|
|
name: "index"
|
|
int64Value: 2
|
|
}
|
|
}
|
|
inputFrameworkOpName: "LSTM"
|
|
}
|
|
}
|
|
mappings {
|
|
frameworkName: "onnx"
|
|
opName: "cos"
|
|
inputFrameworkOpName: "Cos"
|
|
rule {
|
|
ruleName: "ndarraymapping"
|
|
functionName: "ndarraymapping"
|
|
inputTensorName: "input"
|
|
outputTensorName: "input"
|
|
inputToOutput {
|
|
key: "input"
|
|
value: "input"
|
|
}
|
|
ruleType: "tensor"
|
|
inputFrameworkOpName: "Cos"
|
|
}
|
|
rule {
|
|
ruleName: "argdescriptorconstant"
|
|
functionName: "argdescriptorconstant"
|
|
inputBooleanName: "inPlace"
|
|
ruleType: "attribute"
|
|
transformerArgs {
|
|
key: "value"
|
|
transformerArgs {
|
|
name: "inPlace"
|
|
argType: BOOL
|
|
}
|
|
}
|
|
inputFrameworkOpName: "Cos"
|
|
}
|
|
}
|
|
mappings {
|
|
frameworkName: "onnx"
|
|
opName: "sqrt"
|
|
inputFrameworkOpName: "Sqrt"
|
|
rule {
|
|
ruleName: "ndarraymapping"
|
|
functionName: "ndarraymapping"
|
|
inputTensorName: "X"
|
|
outputTensorName: "input"
|
|
inputToOutput {
|
|
key: "input"
|
|
value: "X"
|
|
}
|
|
ruleType: "tensor"
|
|
inputFrameworkOpName: "Sqrt"
|
|
}
|
|
rule {
|
|
ruleName: "argdescriptorconstant"
|
|
functionName: "argdescriptorconstant"
|
|
inputBooleanName: "inPlace"
|
|
ruleType: "attribute"
|
|
transformerArgs {
|
|
key: "value"
|
|
transformerArgs {
|
|
name: "inPlace"
|
|
argType: BOOL
|
|
}
|
|
}
|
|
inputFrameworkOpName: "Sqrt"
|
|
}
|
|
}
|
|
mappings {
|
|
frameworkName: "onnx"
|
|
opName: "asin"
|
|
inputFrameworkOpName: "Asin"
|
|
rule {
|
|
ruleName: "ndarraymapping"
|
|
functionName: "ndarraymapping"
|
|
inputTensorName: "input"
|
|
outputTensorName: "input"
|
|
inputToOutput {
|
|
key: "input"
|
|
value: "input"
|
|
}
|
|
ruleType: "tensor"
|
|
inputFrameworkOpName: "Asin"
|
|
}
|
|
rule {
|
|
ruleName: "argdescriptorconstant"
|
|
functionName: "argdescriptorconstant"
|
|
inputBooleanName: "inPlace"
|
|
ruleType: "attribute"
|
|
transformerArgs {
|
|
key: "value"
|
|
transformerArgs {
|
|
name: "inPlace"
|
|
argType: BOOL
|
|
}
|
|
}
|
|
inputFrameworkOpName: "Asin"
|
|
}
|
|
}
|
|
mappings {
|
|
frameworkName: "onnx"
|
|
opName: "space_to_depth"
|
|
inputFrameworkOpName: "SpaceToDepth"
|
|
rule {
|
|
ruleName: "ndarraymapping"
|
|
functionName: "ndarraymapping"
|
|
inputTensorName: "input"
|
|
outputTensorName: "input"
|
|
inputToOutput {
|
|
key: "input"
|
|
value: "input"
|
|
}
|
|
ruleType: "tensor"
|
|
inputFrameworkOpName: "SpaceToDepth"
|
|
}
|
|
rule {
|
|
ruleName: "valuemapping"
|
|
functionName: "valuemapping"
|
|
inputIntName: "blocksize"
|
|
outputIntName: "block_size"
|
|
inputToOutput {
|
|
key: "block_size"
|
|
value: "blocksize"
|
|
}
|
|
ruleType: "attribute"
|
|
inputFrameworkOpName: "SpaceToDepth"
|
|
}
|
|
rule {
|
|
ruleName: "argdescriptorconstant"
|
|
functionName: "argdescriptorconstant"
|
|
inputIntName: "isNHWC"
|
|
ruleType: "attribute"
|
|
transformerArgs {
|
|
key: "value"
|
|
transformerArgs {
|
|
name: "isNHWC"
|
|
int64Value: 1
|
|
argType: INT64
|
|
argIndex: 1
|
|
}
|
|
}
|
|
inputFrameworkOpName: "SpaceToDepth"
|
|
}
|
|
}
|
|
mappings {
|
|
frameworkName: "onnx"
|
|
opName: "tile"
|
|
inputFrameworkOpName: "Tile"
|
|
rule {
|
|
ruleName: "ndarraymapping"
|
|
functionName: "ndarraymapping"
|
|
inputTensorName: "input"
|
|
inputTensorName: "repeats"
|
|
outputTensorName: "input"
|
|
outputTensorName: "reps_vector"
|
|
inputToOutput {
|
|
key: "input"
|
|
value: "input"
|
|
}
|
|
inputToOutput {
|
|
key: "reps_vector"
|
|
value: "repeats"
|
|
}
|
|
ruleType: "tensor"
|
|
inputFrameworkOpName: "Tile"
|
|
}
|
|
rule {
|
|
ruleName: "argdescriptorconstant"
|
|
functionName: "argdescriptorconstant"
|
|
inputBooleanName: "is_static_reps"
|
|
ruleType: "attribute"
|
|
transformerArgs {
|
|
key: "value"
|
|
transformerArgs {
|
|
name: "is_static_reps"
|
|
boolValue: true
|
|
argType: BOOL
|
|
}
|
|
}
|
|
inputFrameworkOpName: "Tile"
|
|
}
|
|
rule {
|
|
ruleName: "argdescriptorconstant"
|
|
functionName: "argdescriptorconstant"
|
|
inputIntName: "dimensions"
|
|
ruleType: "attribute"
|
|
transformerArgs {
|
|
key: "value"
|
|
transformerArgs {
|
|
name: "dimensions"
|
|
argType: INT64
|
|
}
|
|
}
|
|
inputFrameworkOpName: "Tile"
|
|
}
|
|
}
|
|
mappings {
|
|
frameworkName: "onnx"
|
|
opName: "greater_equal"
|
|
inputFrameworkOpName: "GreaterOrEqual"
|
|
rule {
|
|
ruleName: "ndarraymapping"
|
|
functionName: "ndarraymapping"
|
|
inputTensorName: "A"
|
|
inputTensorName: "B"
|
|
outputTensorName: "input"
|
|
outputTensorName: "y"
|
|
inputToOutput {
|
|
key: "input"
|
|
value: "A"
|
|
}
|
|
inputToOutput {
|
|
key: "y"
|
|
value: "B"
|
|
}
|
|
ruleType: "tensor"
|
|
inputFrameworkOpName: "GreaterOrEqual"
|
|
}
|
|
rule {
|
|
ruleName: "argdescriptorconstant"
|
|
functionName: "argdescriptorconstant"
|
|
inputBooleanName: "inPlace"
|
|
ruleType: "attribute"
|
|
transformerArgs {
|
|
key: "value"
|
|
transformerArgs {
|
|
name: "inPlace"
|
|
argType: BOOL
|
|
}
|
|
}
|
|
inputFrameworkOpName: "GreaterOrEqual"
|
|
}
|
|
}
|
|
mappings {
|
|
frameworkName: "onnx"
|
|
opName: "depth_to_space"
|
|
inputFrameworkOpName: "DepthToSpace"
|
|
rule {
|
|
ruleName: "ndarraymapping"
|
|
functionName: "ndarraymapping"
|
|
inputTensorName: "input"
|
|
outputTensorName: "input"
|
|
inputToOutput {
|
|
key: "input"
|
|
value: "input"
|
|
}
|
|
ruleType: "tensor"
|
|
inputFrameworkOpName: "DepthToSpace"
|
|
}
|
|
rule {
|
|
ruleName: "valuemapping"
|
|
functionName: "valuemapping"
|
|
inputIntName: "blocksize"
|
|
outputIntName: "block_size"
|
|
inputToOutput {
|
|
key: "block_size"
|
|
value: "blocksize"
|
|
}
|
|
ruleType: "attribute"
|
|
inputFrameworkOpName: "DepthToSpace"
|
|
}
|
|
rule {
|
|
ruleName: "argdescriptorconstant"
|
|
functionName: "argdescriptorconstant"
|
|
inputIntName: "isNHWC"
|
|
ruleType: "attribute"
|
|
transformerArgs {
|
|
key: "value"
|
|
transformerArgs {
|
|
name: "isNHWC"
|
|
int64Value: 1
|
|
argType: INT64
|
|
argIndex: 1
|
|
}
|
|
}
|
|
inputFrameworkOpName: "DepthToSpace"
|
|
}
|
|
}
|
|
mappings {
|
|
frameworkName: "onnx"
|
|
opName: "isnan"
|
|
inputFrameworkOpName: "IsNaN"
|
|
rule {
|
|
ruleName: "ndarraymapping"
|
|
functionName: "ndarraymapping"
|
|
inputTensorName: "X"
|
|
outputTensorName: "input"
|
|
inputToOutput {
|
|
key: "input"
|
|
value: "X"
|
|
}
|
|
ruleType: "tensor"
|
|
inputFrameworkOpName: "IsNaN"
|
|
}
|
|
rule {
|
|
ruleName: "argdescriptorconstant"
|
|
functionName: "argdescriptorconstant"
|
|
inputBooleanName: "inPlace"
|
|
ruleType: "attribute"
|
|
transformerArgs {
|
|
key: "value"
|
|
transformerArgs {
|
|
name: "inPlace"
|
|
argType: BOOL
|
|
}
|
|
}
|
|
inputFrameworkOpName: "IsNaN"
|
|
}
|
|
}
|
|
mappings {
|
|
frameworkName: "onnx"
|
|
opName: "divide"
|
|
inputFrameworkOpName: "Div"
|
|
rule {
|
|
ruleName: "ndarraymapping"
|
|
functionName: "ndarraymapping"
|
|
inputTensorName: "A"
|
|
inputTensorName: "B"
|
|
outputTensorName: "input"
|
|
outputTensorName: "y"
|
|
inputToOutput {
|
|
key: "input"
|
|
value: "A"
|
|
}
|
|
inputToOutput {
|
|
key: "y"
|
|
value: "B"
|
|
}
|
|
ruleType: "tensor"
|
|
inputFrameworkOpName: "Div"
|
|
}
|
|
rule {
|
|
ruleName: "argdescriptorconstant"
|
|
functionName: "argdescriptorconstant"
|
|
inputBooleanName: "inPlace"
|
|
ruleType: "attribute"
|
|
transformerArgs {
|
|
key: "value"
|
|
transformerArgs {
|
|
name: "inPlace"
|
|
argType: BOOL
|
|
}
|
|
}
|
|
inputFrameworkOpName: "Div"
|
|
}
|
|
}
|
|
mappings {
|
|
frameworkName: "onnx"
|
|
opName: "neg"
|
|
inputFrameworkOpName: "Neg"
|
|
rule {
|
|
ruleName: "ndarraymapping"
|
|
functionName: "ndarraymapping"
|
|
inputTensorName: "X"
|
|
outputTensorName: "input"
|
|
inputToOutput {
|
|
key: "input"
|
|
value: "X"
|
|
}
|
|
ruleType: "tensor"
|
|
inputFrameworkOpName: "Neg"
|
|
}
|
|
rule {
|
|
ruleName: "argdescriptorconstant"
|
|
functionName: "argdescriptorconstant"
|
|
inputBooleanName: "inPlace"
|
|
ruleType: "attribute"
|
|
transformerArgs {
|
|
key: "value"
|
|
transformerArgs {
|
|
name: "inPlace"
|
|
argType: BOOL
|
|
}
|
|
}
|
|
inputFrameworkOpName: "Neg"
|
|
}
|
|
}
|
|
mappings {
|
|
frameworkName: "onnx"
|
|
opName: "matrix_determinant"
|
|
inputFrameworkOpName: "Det"
|
|
rule {
|
|
ruleName: "ndarraymapping"
|
|
functionName: "ndarraymapping"
|
|
inputTensorName: "X"
|
|
outputTensorName: "input"
|
|
inputToOutput {
|
|
key: "input"
|
|
value: "X"
|
|
}
|
|
ruleType: "tensor"
|
|
inputFrameworkOpName: "Det"
|
|
}
|
|
rule {
|
|
ruleName: "argdescriptorconstant"
|
|
functionName: "argdescriptorconstant"
|
|
inputBooleanName: "inPlace"
|
|
ruleType: "attribute"
|
|
transformerArgs {
|
|
key: "value"
|
|
transformerArgs {
|
|
name: "inPlace"
|
|
argType: BOOL
|
|
}
|
|
}
|
|
inputFrameworkOpName: "Det"
|
|
}
|
|
}
|
|
mappings {
|
|
frameworkName: "onnx"
|
|
opName: "pad"
|
|
inputFrameworkOpName: "Pad"
|
|
rule {
|
|
ruleName: "ndarraymapping"
|
|
functionName: "ndarraymapping"
|
|
inputTensorName: "data"
|
|
inputTensorName: "pads"
|
|
outputTensorName: "input"
|
|
outputTensorName: "paddings"
|
|
inputToOutput {
|
|
key: "input"
|
|
value: "data"
|
|
}
|
|
inputToOutput {
|
|
key: "paddings"
|
|
value: "pads"
|
|
}
|
|
ruleType: "tensor"
|
|
inputFrameworkOpName: "Pad"
|
|
}
|
|
rule {
|
|
ruleName: "stringtoindex"
|
|
functionName: "stringtoindex"
|
|
inputStringAttrName: "mode"
|
|
outputIntName: "mode"
|
|
inputFloatName: "mode"
|
|
inputFloatName: "mode"
|
|
inputFloatName: "mode"
|
|
inputToOutput {
|
|
key: "mode"
|
|
value: "mode"
|
|
}
|
|
ruleType: "attribute"
|
|
transformerArgs {
|
|
key: "mode"
|
|
transformerArgs {
|
|
name: "mode"
|
|
stringValue: "constant"
|
|
}
|
|
transformerArgs {
|
|
name: "mode"
|
|
stringValue: "reflect"
|
|
}
|
|
transformerArgs {
|
|
name: "mode"
|
|
stringValue: "edge"
|
|
}
|
|
}
|
|
transformerArgs {
|
|
key: "mode"
|
|
transformerArgs {
|
|
name: "mode"
|
|
stringValue: "constant"
|
|
}
|
|
transformerArgs {
|
|
name: "mode"
|
|
stringValue: "reflect"
|
|
}
|
|
transformerArgs {
|
|
name: "mode"
|
|
stringValue: "edge"
|
|
}
|
|
}
|
|
transformerArgs {
|
|
key: "mode"
|
|
transformerArgs {
|
|
name: "mode"
|
|
stringValue: "constant"
|
|
}
|
|
transformerArgs {
|
|
name: "mode"
|
|
stringValue: "reflect"
|
|
}
|
|
transformerArgs {
|
|
name: "mode"
|
|
stringValue: "edge"
|
|
}
|
|
}
|
|
inputFrameworkOpName: "Pad"
|
|
}
|
|
rule {
|
|
ruleName: "argdescriptorconstant"
|
|
functionName: "argdescriptorconstant"
|
|
inputFloatName: "padValue"
|
|
ruleType: "attribute"
|
|
transformerArgs {
|
|
key: "value"
|
|
transformerArgs {
|
|
name: "padValue"
|
|
argType: DOUBLE
|
|
}
|
|
}
|
|
inputFrameworkOpName: "Pad"
|
|
}
|
|
}
|
|
mappings {
|
|
frameworkName: "onnx"
|
|
opName: "conv2d"
|
|
inputFrameworkOpName: "Conv"
|
|
rule {
|
|
ruleName: "ndarraymapping"
|
|
functionName: "ndarraymapping"
|
|
inputTensorName: "X"
|
|
inputTensorName: "W"
|
|
inputTensorName: "B"
|
|
outputTensorName: "input"
|
|
outputTensorName: "weights"
|
|
outputTensorName: "bias"
|
|
inputToOutput {
|
|
key: "input"
|
|
value: "X"
|
|
}
|
|
inputToOutput {
|
|
key: "weights"
|
|
value: "W"
|
|
}
|
|
inputToOutput {
|
|
key: "bias"
|
|
value: "B"
|
|
}
|
|
ruleType: "tensor"
|
|
inputFrameworkOpName: "Conv"
|
|
}
|
|
rule {
|
|
ruleName: "argdescriptorconstant"
|
|
functionName: "argdescriptorconstant"
|
|
inputIntName: "isNCHW"
|
|
ruleType: "attribute"
|
|
transformerArgs {
|
|
key: "value"
|
|
transformerArgs {
|
|
name: "isNCHW"
|
|
argType: INT64
|
|
argIndex: 9
|
|
}
|
|
}
|
|
inputFrameworkOpName: "Conv"
|
|
}
|
|
rule {
|
|
ruleName: "argdescriptorconstant"
|
|
functionName: "argdescriptorconstant"
|
|
inputIntName: "wFormat"
|
|
ruleType: "attribute"
|
|
transformerArgs {
|
|
key: "value"
|
|
transformerArgs {
|
|
name: "wFormat"
|
|
int64Value: 1
|
|
argType: INT64
|
|
argIndex: 10
|
|
}
|
|
}
|
|
inputFrameworkOpName: "Conv"
|
|
}
|
|
rule {
|
|
ruleName: "stringequals"
|
|
functionName: "stringequals"
|
|
inputStringAttrName: "auto_pad"
|
|
outputIntName: "isSameMode"
|
|
inputFloatName: "auto_pad"
|
|
inputToOutput {
|
|
key: "isSameMode"
|
|
value: "auto_pad"
|
|
}
|
|
ruleType: "attribute"
|
|
transformerArgs {
|
|
key: "isSameMode"
|
|
transformerArgs {
|
|
name: "auto_pad"
|
|
argIndex: 8
|
|
stringValue: "SAME"
|
|
}
|
|
}
|
|
inputFrameworkOpName: "Conv"
|
|
}
|
|
rule {
|
|
ruleName: "listattributevaluelookuptoindex"
|
|
functionName: "listattributevaluelookuptoindex"
|
|
inputIntName: "dH"
|
|
outputIntName: "dH"
|
|
inputFloatName: "dilations"
|
|
inputToOutput {
|
|
key: "dH"
|
|
value: "dilations"
|
|
}
|
|
ruleType: "attribute"
|
|
transformerArgs {
|
|
key: "dH"
|
|
transformerArgs {
|
|
name: "dilations"
|
|
argIndex: 6
|
|
}
|
|
transformerArgs {
|
|
name: "dH"
|
|
int64Value: 1
|
|
argType: INT64
|
|
argIndex: 6
|
|
}
|
|
}
|
|
transformerArgs {
|
|
key: "dH"
|
|
transformerArgs {
|
|
name: "dilations"
|
|
argIndex: 6
|
|
}
|
|
transformerArgs {
|
|
name: "dH"
|
|
int64Value: 1
|
|
argType: INT64
|
|
argIndex: 6
|
|
}
|
|
}
|
|
inputFrameworkOpName: "Conv"
|
|
}
|
|
rule {
|
|
ruleName: "listattributevaluelookuptoindex"
|
|
functionName: "listattributevaluelookuptoindex"
|
|
inputIntName: "dW"
|
|
outputIntName: "dW"
|
|
inputFloatName: "dilations"
|
|
inputToOutput {
|
|
key: "dW"
|
|
value: "dilations"
|
|
}
|
|
ruleType: "attribute"
|
|
transformerArgs {
|
|
key: "dW"
|
|
transformerArgs {
|
|
name: "dilations"
|
|
int64Value: 1
|
|
argIndex: 7
|
|
}
|
|
transformerArgs {
|
|
name: "dW"
|
|
int64Value: 1
|
|
argType: INT64
|
|
argIndex: 7
|
|
}
|
|
}
|
|
transformerArgs {
|
|
key: "dW"
|
|
transformerArgs {
|
|
name: "dilations"
|
|
int64Value: 1
|
|
argIndex: 7
|
|
}
|
|
transformerArgs {
|
|
name: "dW"
|
|
int64Value: 1
|
|
argType: INT64
|
|
argIndex: 7
|
|
}
|
|
}
|
|
inputFrameworkOpName: "Conv"
|
|
}
|
|
rule {
|
|
ruleName: "listattributevaluelookuptoindex"
|
|
functionName: "listattributevaluelookuptoindex"
|
|
outputIntName: "pH"
|
|
inputFloatName: "pads"
|
|
inputToOutput {
|
|
key: "pH"
|
|
value: "pads"
|
|
}
|
|
ruleType: "attribute"
|
|
transformerArgs {
|
|
key: "pH"
|
|
transformerArgs {
|
|
name: "pads"
|
|
argIndex: 4
|
|
}
|
|
}
|
|
inputFrameworkOpName: "Conv"
|
|
}
|
|
rule {
|
|
ruleName: "listattributevaluelookuptoindex"
|
|
functionName: "listattributevaluelookuptoindex"
|
|
outputIntName: "pW"
|
|
inputFloatName: "pads"
|
|
inputToOutput {
|
|
key: "pW"
|
|
value: "pads"
|
|
}
|
|
ruleType: "attribute"
|
|
transformerArgs {
|
|
key: "pW"
|
|
transformerArgs {
|
|
name: "pads"
|
|
int64Value: 1
|
|
argIndex: 5
|
|
}
|
|
}
|
|
inputFrameworkOpName: "Conv"
|
|
}
|
|
rule {
|
|
ruleName: "listattributevaluelookuptoindex"
|
|
functionName: "listattributevaluelookuptoindex"
|
|
inputIntName: "strides"
|
|
outputIntName: "sH"
|
|
inputFloatName: "strides"
|
|
inputToOutput {
|
|
key: "sH"
|
|
value: "strides"
|
|
}
|
|
ruleType: "attribute"
|
|
transformerArgs {
|
|
key: "sH"
|
|
transformerArgs {
|
|
name: "strides"
|
|
argIndex: 2
|
|
}
|
|
transformerArgs {
|
|
name: "strides"
|
|
int64Value: 1
|
|
argType: INT64
|
|
argIndex: 2
|
|
}
|
|
}
|
|
transformerArgs {
|
|
key: "sH"
|
|
transformerArgs {
|
|
name: "strides"
|
|
argIndex: 2
|
|
}
|
|
transformerArgs {
|
|
name: "strides"
|
|
int64Value: 1
|
|
argType: INT64
|
|
argIndex: 2
|
|
}
|
|
}
|
|
inputFrameworkOpName: "Conv"
|
|
}
|
|
rule {
|
|
ruleName: "listattributevaluelookuptoindex"
|
|
functionName: "listattributevaluelookuptoindex"
|
|
inputIntName: "strides"
|
|
outputIntName: "sW"
|
|
inputFloatName: "strides"
|
|
inputToOutput {
|
|
key: "sW"
|
|
value: "strides"
|
|
}
|
|
ruleType: "attribute"
|
|
transformerArgs {
|
|
key: "sW"
|
|
transformerArgs {
|
|
name: "strides"
|
|
int64Value: 1
|
|
argIndex: 3
|
|
}
|
|
transformerArgs {
|
|
name: "strides"
|
|
int64Value: 1
|
|
argType: INT64
|
|
argIndex: 3
|
|
}
|
|
}
|
|
transformerArgs {
|
|
key: "sW"
|
|
transformerArgs {
|
|
name: "strides"
|
|
int64Value: 1
|
|
argIndex: 3
|
|
}
|
|
transformerArgs {
|
|
name: "strides"
|
|
int64Value: 1
|
|
argType: INT64
|
|
argIndex: 3
|
|
}
|
|
}
|
|
inputFrameworkOpName: "Conv"
|
|
}
|
|
rule {
|
|
ruleName: "listattributevaluelookuptoindex"
|
|
functionName: "listattributevaluelookuptoindex"
|
|
outputIntName: "kW"
|
|
inputFloatName: "kernel_shape"
|
|
inputToOutput {
|
|
key: "kW"
|
|
value: "kernel_shape"
|
|
}
|
|
ruleType: "attribute"
|
|
transformerArgs {
|
|
key: "kW"
|
|
transformerArgs {
|
|
name: "kernel_shape"
|
|
int64Value: 1
|
|
}
|
|
}
|
|
inputFrameworkOpName: "Conv"
|
|
}
|
|
rule {
|
|
ruleName: "listattributevaluelookuptoindex"
|
|
functionName: "listattributevaluelookuptoindex"
|
|
outputIntName: "kH"
|
|
inputFloatName: "kernel_shape"
|
|
inputToOutput {
|
|
key: "kH"
|
|
value: "kernel_shape"
|
|
}
|
|
ruleType: "attribute"
|
|
transformerArgs {
|
|
key: "kH"
|
|
transformerArgs {
|
|
name: "kernel_shape"
|
|
argIndex: 1
|
|
}
|
|
}
|
|
inputFrameworkOpName: "Conv"
|
|
}
|
|
}
|
|
mappings {
|
|
frameworkName: "onnx"
|
|
opName: "greater"
|
|
inputFrameworkOpName: "Greater"
|
|
rule {
|
|
ruleName: "ndarraymapping"
|
|
functionName: "ndarraymapping"
|
|
inputTensorName: "A"
|
|
inputTensorName: "B"
|
|
outputTensorName: "input"
|
|
outputTensorName: "y"
|
|
inputToOutput {
|
|
key: "input"
|
|
value: "A"
|
|
}
|
|
inputToOutput {
|
|
key: "y"
|
|
value: "B"
|
|
}
|
|
ruleType: "tensor"
|
|
inputFrameworkOpName: "Greater"
|
|
}
|
|
rule {
|
|
ruleName: "argdescriptorconstant"
|
|
functionName: "argdescriptorconstant"
|
|
inputBooleanName: "inPlace"
|
|
ruleType: "attribute"
|
|
transformerArgs {
|
|
key: "value"
|
|
transformerArgs {
|
|
name: "inPlace"
|
|
argType: BOOL
|
|
}
|
|
}
|
|
inputFrameworkOpName: "Greater"
|
|
}
|
|
}
|
|
mappings {
|
|
frameworkName: "onnx"
|
|
opName: "sign"
|
|
inputFrameworkOpName: "Sign"
|
|
rule {
|
|
ruleName: "ndarraymapping"
|
|
functionName: "ndarraymapping"
|
|
inputTensorName: "input"
|
|
outputTensorName: "input"
|
|
inputToOutput {
|
|
key: "input"
|
|
value: "input"
|
|
}
|
|
ruleType: "tensor"
|
|
inputFrameworkOpName: "Sign"
|
|
}
|
|
rule {
|
|
ruleName: "argdescriptorconstant"
|
|
functionName: "argdescriptorconstant"
|
|
inputBooleanName: "inPlace"
|
|
ruleType: "attribute"
|
|
transformerArgs {
|
|
key: "value"
|
|
transformerArgs {
|
|
name: "inPlace"
|
|
argType: BOOL
|
|
}
|
|
}
|
|
inputFrameworkOpName: "Sign"
|
|
}
|
|
}
|
|
mappings {
|
|
frameworkName: "onnx"
|
|
opName: "softsign"
|
|
inputFrameworkOpName: "Softsign"
|
|
rule {
|
|
ruleName: "ndarraymapping"
|
|
functionName: "ndarraymapping"
|
|
inputTensorName: "input"
|
|
outputTensorName: "input"
|
|
inputToOutput {
|
|
key: "input"
|
|
value: "input"
|
|
}
|
|
ruleType: "tensor"
|
|
inputFrameworkOpName: "Softsign"
|
|
}
|
|
rule {
|
|
ruleName: "argdescriptorconstant"
|
|
functionName: "argdescriptorconstant"
|
|
inputBooleanName: "inPlace"
|
|
ruleType: "attribute"
|
|
transformerArgs {
|
|
key: "value"
|
|
transformerArgs {
|
|
name: "inPlace"
|
|
argType: BOOL
|
|
}
|
|
}
|
|
inputFrameworkOpName: "Softsign"
|
|
}
|
|
}
|
|
mappings {
|
|
frameworkName: "onnx"
|
|
opName: "exp"
|
|
inputFrameworkOpName: "Exp"
|
|
rule {
|
|
ruleName: "ndarraymapping"
|
|
functionName: "ndarraymapping"
|
|
inputTensorName: "input"
|
|
outputTensorName: "input"
|
|
inputToOutput {
|
|
key: "input"
|
|
value: "input"
|
|
}
|
|
ruleType: "tensor"
|
|
inputFrameworkOpName: "Exp"
|
|
}
|
|
rule {
|
|
ruleName: "argdescriptorconstant"
|
|
functionName: "argdescriptorconstant"
|
|
inputBooleanName: "inPlace"
|
|
ruleType: "attribute"
|
|
transformerArgs {
|
|
key: "value"
|
|
transformerArgs {
|
|
name: "inPlace"
|
|
argType: BOOL
|
|
}
|
|
}
|
|
inputFrameworkOpName: "Exp"
|
|
}
|
|
}
|