Function-Level Support Status

ONNX Support Status

Note:In this document, the numbers in the header of all tables represent the version of onnx opset.
ONNX Version:1.6.0

Import

  • ✓: onnx specification defined, and supported.
  • X: onnx specification defined, but not support yet.
  • Empty: Not defined (Support status follows latest).

Total: 93/155

ONNX Operator 1 2 3 4 5 6 7 8 9 10 11 NNabla Func Description
Abs                   Abs  
Acos                     ACos  
Acosh                     ACosh  
Add                 Reshape, Add2  
And                 LogicalAnd, Reshape  
ArgMax               X Max  
ArgMin               X Min  
Asin                     ASin  
Asinh                     ASinh  
Atan                     ATan  
Atanh                     ATanh  
AveragePool             X X Pad, AveragePooling Not all features are verified, since some features are not supported by caffe2. Those features can be verified by ONNXRuntime when opset > 6. Some feature is not supported by Nnabla such as Pad’s edge mode. if opset >= 10, the ceil_mode is not supported.
BatchNormalization X         X X       BatchNormalization  
BitShift                     X   Not yet implemented.
Cast             X     Abs, Log  
Ceil                   Ceil  
Clip                 MinimumScalar, MaximumScalar, Identity  
Compress                 X   X   Not yet implemented.
Concat               X Concatenate  
ConcatFromSequence                     X   Not yet implemented.
Constant             X   X Identity  
ConstantOfShape                     Constant  
Conv                 X Convolution  
ConvInteger                   X     Not yet implemented.
ConvTranspose                 X Deconvolution, Pad  
Cos                     Cos  
Cosh                     Cosh  
CumSum                     X   Not yet implemented.
DepthToSpace                 Reshape, Transpose  
DequantizeLinear                   X     Not yet implemented.
Det                     X   Not yet implemented.
Div                 Div2, Reshape  
Dropout X         X     X   Identity  
DynamicQuantizeLinear                     X   Not yet implemented.
Elu                   ELU  
Equal               X Reshape, Equal  
Erf                 X       Not yet implemented.
Exp                   Exp  
Expand                   Broadcast, Reshape  
EyeLike                 X       Not yet implemented.
Flatten               Reshape  
Floor                   Floor  
GRU X   X       X           Not yet implemented.
Gather                 Slice, Concatenate  
GatherElements                     X   Not yet implemented.
GatherND                     X   Not yet implemented.
Gemm             BatchMatmul, Reshape, Add2, MulScalar  
GlobalAveragePool                   GlobalAveragePooling  
GlobalLpPool X X                     Not yet implemented.
GlobalMaxPool X                       Not yet implemented.
Greater               Reshape, Greater  
HardSigmoid                   MaximumScalar, MulScalar, AddScalar, HardSigmoid, MinimumScalar  
Hardmax                 Reshape, OneHot, Max  
Identity                   Identity  
If X                       Not yet implemented.
InstanceNormalization                   Split, Reshape, BatchNormalization, Concatenate  
IsInf                     IsInf  
IsNaN                     IsNaN  
LRN                   Transpose, MulScalar, PowScalar, AddScalar, Div2, SumPooling  
LSTM X           X           Not yet implemented.
LeakyRelu                   LeakyReLU  
Less               Reshape, Less  
Log                   Log  
LogSoftmax                 Reshape, Max, Exp, Log, Sub2, Sum, Add2  
Loop X                   X   Not yet implemented.
LpNormalization X                       Not yet implemented.
LpPool X X                 X   Not yet implemented.
MatMul                 BatchMatmul  
MatMulInteger                   X     Not yet implemented.
Max               Maximum2  
MaxPool           X   X X Pad, MaxPooling Not all features are verified, since some features are not supported by caffe2. Those features can be verified by ONNXRuntime. if opset >= 10, the ceil_mode is not supported, dilations is not equal to 1 is not supported.
MaxRoiPool X                       Not yet implemented.
MaxUnpool                 X   X   Not yet implemented.
Mean               Broadcast, Stack, Mean  
MeanVarianceNormalization                 X       Not yet implemented.
Min               Minimum2  
Mod                   X     Not yet implemented.
Mul                 Reshape, Mul2  
Multinomial             X           Not yet implemented.
Neg                   MulScalar  
NonMaxSuppression                   X X   Not yet implemented.
NonZero                 X       Not yet implemented.
Not                   LogicalNot  
OneHot                 X   X   Not yet implemented.
Or                 Reshape, LogicalOr  
PRelu         X   X     PReLU  
Pad               Pad Onnx required to support “edge” mode, while nnabla does not support it.
Pow                 Reshape, Pow2  
QLinearConv                   X     Not yet implemented.
QLinearMatMul                   X     Not yet implemented.
QuantizeLinear                   X     Not yet implemented.
RNN X           X           Not yet implemented.
RandomNormal X                       Not yet implemented.
RandomNormalLike X                       Not yet implemented.
RandomUniform X                       Not yet implemented.
RandomUniformLike X                       Not yet implemented.
Range                     X   Not yet implemented.
Reciprocal                   RDivScalar  
ReduceL1 X                   X   Not yet implemented.
ReduceL2 X                   X   Not yet implemented.
ReduceLogSum X                   X   Not yet implemented.
ReduceLogSumExp X                   X   Not yet implemented.
ReduceMax                 Max  
ReduceMean                 Mean  
ReduceMin                 Min  
ReduceProd                 Prod  
ReduceSum                 Sum  
ReduceSumSquare                 PowScalar, Sum  
Relu                   ReLU  
Reshape                 Reshape  
Resize                   X X   Not yet implemented.
ReverseSequence                   X     Not yet implemented.
RoiAlign                   X     Not yet implemented.
Round                     Round  
Scan               X X   X   Not yet implemented.
Scatter                 X   X   Not yet implemented.
ScatterElements                     X   Not yet implemented.
ScatterND                     X   Not yet implemented.
Selu                   SELU  
SequenceAt                     X   Not yet implemented.
SequenceConstruct                     X   Not yet implemented.
SequenceErase                     X   Not yet implemented.
SequenceInsert                     X   Not yet implemented.
SequenceLength                     X   Not yet implemented.
Shape X                       Not yet implemented.
Shrink                 X       Not yet implemented.
Sigmoid                   Sigmoid  
Sign                     Sign  
Sin                     Sin  
Sinh                     Sinh  
Size X                       Not yet implemented.
Slice               X Slice  
Softmax                 Reshape, Exp, Div2, Sub2, Sum, Max  
Softplus                   SoftPlus  
Softsign                   SoftSign  
SpaceToDepth                   Reshape, Transpose  
Split               Stack, Split  
SplitToSequence                     X   Not yet implemented.
Sqrt                   PowScalar  
Squeeze                 Reshape  
StringNormalizer                   X     Not yet implemented.
Sub                 Sub2, Reshape  
Sum               Add2  
Tan                     Tan  
Tanh                   Tanh  
TfIdfVectorizer                 X       Not yet implemented.
ThresholdedRelu                     GreaterScalar, Where, Constant  
Tile                   Tile  
TopK X                 X X   Not yet implemented.
Transpose                   Transpose  
Unique                     X   Not yet implemented.
Unsqueeze                 Reshape  
Upsample           X   Unpooling  
Where                     Where  
Xor                 Reshape, LogicalXor  

Export

  • ✓: Support to export this opset.
  • △: Partially support to export this opset (e.g. some cases cannot be supported, or not completely tested).
  • X: Supported, but test failed.
  • Empty: Not support corresponding opset version.

Total: 119/173

Neural Network Layer

Count 11/14

NNabla Function 6 7 9 10 11 ONNX Op Description
Affine Gemm, Reshape  
RNN             Not yet implemented.
LSTM             Not yet implemented.
GRU             Not yet implemented.
Convolution Reshape, Conv  
DepthwiseConvolution Reshape, Conv  
Deconvolution ConvTranspose, Reshape Caffe2 and onnxruntime do not support dilations != 1.
DepthwiseDeconvolution ConvTranspose, Reshape Caffe2 and onnxruntime do not support dilations != 1.
MaxPooling X Reshape, Pad, MaxPool  
AveragePooling X Reshape, Pad, AveragePool Currently only supports the cases where both ignore_border and including_pad are True.
GlobalAveragePooling GlobalAveragePool  
SumPooling X X Reshape, Pad, Constant, Mul, AveragePool  
Unpooling Resize The kernel only supports 2d on opset 6.
Embed Gather  

Neural Network Activation Functions

Count 21/21

NNabla Function 6 7 9 10 11 ONNX Op Description
Sigmoid Sigmoid  
Swish Mul, Sigmoid  
Tanh Tanh  
ReLU Relu  
LeakyReLU LeakyRelu  
Softmax Exp, ReduceMax, Sub, ReduceSum, Div ONNX Add, Sub operator does not support multidirectional broadcasting on opset 6.
LogSoftmax Exp, ReduceMax, Sub, Log, ReduceSum  
ELU Elu  
SELU Selu  
CReLU Concat, Neg, Relu  
CELU Elu, Concat, Neg  
PReLU Reshape, PRelu  
GELU Sqrt, Constant, Pow, Tanh, Mul, Add, Div  
ReLU6 Min, Relu, Constant  
HardSigmoid HardSigmoid  
HardTanh Neg, Min, Max, Constant  
LogSigmoid Sigmoid, Log  
SoftPlus Softplus  
SoftSign Softsign  
TanhShrink Sub, Tanh  
Sinc X X Constant, Where, Sin, Equal, Div  

Normalization

Count 1/6

NNabla Function 6 7 9 10 11 ONNX Op Description
FusedBatchNormalization X X X X X Relu, Constant, ReduceMean, BatchNormalization, Mul, Sub, ReduceSum, Div Not yet implemented.
BatchNormalization Reshape, Constant, ReduceMean, BatchNormalization, Mul, Sub, ReduceSum, Div In inferring stage, caffe2 mistmatch onnx 1.4.x’s implementation, “in-place” feature cannot be applied.
SyncBatchNormalization             Not yet implemented.
MeanSubtraction             Not yet implemented.
ClipGradByValue             Not yet implemented.
ClipGradByNorm             Not yet implemented.

Reduction

Count 5/7

NNabla Function 6 7 9 10 11 ONNX Op Description
Sum ReduceSum  
Mean ReduceMean  
Max ReduceMax  
Min ReduceMin  
Prod ReduceProd  
ReduceSum             Not yet implemented.
ReduceMean             Not yet implemented.

Arithmetic

Count 11/12

NNabla Function 6 7 9 10 11 ONNX Op Description
Add2 Add ONNX Add operator does not support multidirectional broadcasting on opset 6.
BcAdd2             Not yet implemented.
Sub2 Sub ONNX Sub operator does not support multidirectional broadcasting on opset 6.
Mul2 Mul ONNX Mul operator does not support multidirectional broadcasting on opset 6.
Div2 Div ONNX Div operator does not support multidirectional broadcasting on opset 6.
Pow2 Pow ONNX Pow operator does not support multidirectional broadcasting on opset 6.
AddScalar Add, Constant  
MulScalar Mul, Constant  
PowScalar Pow, Constant  
RSubScalar Sub, Constant  
RDivScalar Constant, Div  
RPowScalar Pow, Constant  

Logical

Count 29/29

NNabla Function 6 7 9 10 11 ONNX Op Description
Sign X X Sign  
Minimum2 Min, Add, Constant ONNX Add operator does not support multidirectional broadcasting on opset 6.
Maximum2 Add, Max, Constant ONNX Add operator does not support multidirectional broadcasting on opset 6.
MinimumScalar Min, Add, Constant  
MaximumScalar Add, Max, Constant  
LogicalAnd And  
LogicalOr Or  
LogicalXor Xor  
Equal Equal  
NotEqual Equal, Not  
GreaterEqual Less, Not  
Greater Greater  
LessEqual Greater, Not  
Less Less  
LogicalAndScalar And, Constant  
LogicalOrScalar Or, Constant  
LogicalXorScalar Xor, Constant  
EqualScalar Equal, Constant  
NotEqualScalar Equal, Not, Constant  
GreaterEqualScalar Less, Not, Constant  
GreaterScalar Greater, Constant  
LessEqualScalar Greater, Not, Constant  
LessScalar Less, Constant  
LogicalNot Not  
IsNaN X X IsNaN  
IsInf X X X IsInf  
ResetNaN X X Where, IsNaN, Constant  
ResetInf X X X Where, IsInf, Constant  
Where X X Where  

Math

Count 22/22

NNabla Function 6 7 9 10 11 ONNX Op Description
Constant Identity, Constant  
Arange Identity, Constant  
Abs Abs  
Exp Exp  
Log Log  
Identity Identity  
BatchMatmul Reshape, MatMul, Transpose  
Round X X X X Round  
Ceil Ceil  
Floor Floor  
Sin X Sin  
Cos X Cos  
Tan X Tan  
Sinh X X Sinh  
Cosh X X Cosh  
ASin X Asin  
ACos X Acos  
ATan X Atan  
ATan2 X Div, Atan  
ASinh X X Asinh  
ACosh X X Acosh  
ATanh X X Atanh  

Array Manipulation

Count 12/19

NNabla Function 6 7 9 10 11 ONNX Op Description
Concatenate Concat  
Split Split, Squeeze  
Stack Concat, Unsqueeze  
Slice Slice, Constant ONNX slice cannot support step != 1 on opset < 10.
Pad Pad, Constant When the mode of the pad is reflect, if the size of the pad exceeds the input size, caffe2 and onnxruntime cannot handle it.
Transpose Transpose  
Broadcast X X    
BroadcastTo    
Tile Reshape, Tile, Constant  
OneHot Gather, Reshape, Flatten  
Flip Gather, Transpose, Identity  
Shift             Not yet implemented.
Sort             Not yet implemented.
Reshape Reshape, Constant  
MatrixDiag             Not yet implemented.
MatrixDiagPart             Not yet implemented.
Assign             Not yet implemented.
GatherNd             Not yet implemented.
ScatterNd             Not yet implemented.

Signal Processing

Count 1/3

NNabla Function 6 7 9 10 11 ONNX Op Description
Interpolate X X X X Reshape, Resize  
FFT             Not yet implemented.
IFFT             Not yet implemented.

Stochasticity

Count 0/11

NNabla Function 6 7 9 10 11 ONNX Op Description
Dropout X X X X X Dropout The Dropout in nnabla has no test mode and contains random parameters, so the test result is not the same as onnx.
TopKData             Not yet implemented.
TopKGrad             Not yet implemented.
Rand             Not yet implemented.
Randint             Not yet implemented.
Randn             Not yet implemented.
RandomChoice             Not yet implemented.
RandomCrop             Not yet implemented.
RandomFlip             Not yet implemented.
RandomShift             Not yet implemented.
ImageAugmentation             Not yet implemented.

Loss Functions

Count 0/9

NNabla Function 6 7 9 10 11 ONNX Op Description
SigmoidCrossEntropy             Not yet implemented.
BinaryCrossEntropy             Not yet implemented.
SoftmaxCrossEntropy             Not yet implemented.
CategoricalCrossEntropy             Not yet implemented.
SquaredError             Not yet implemented.
AbsoluteError             Not yet implemented.
HuberLoss             Not yet implemented.
EpsilonInsensitiveLoss             Not yet implemented.
KLMultinomial             Not yet implemented.

Quantization Neural Network Layers

Count 6/12

NNabla Function 6 7 9 10 11 ONNX Op Description
BinarySigmoid X X Where, Greater, Constant  
BinaryTanh X X Where, Greater, Constant  
BinaryConnectAffine Gemm, Reshape  
BinaryConnectConvolution Reshape, Conv  
BinaryWeightAffine Mul, Reshape, Add, MatMul  
BinaryWeightConvolution Mul, Reshape, Conv, Add  
INQAffine             Not yet implemented.
INQConvolution             Not yet implemented.
FixedPointQuantize             Not yet implemented.
MinMaxQuantize             Not yet implemented.
Pow2Quantize             Not yet implemented.
Prune             Not yet implemented.

Validation

Count 0/3

NNabla Function 6 7 9 10 11 ONNX Op Description
TopNError             Not yet implemented.
BinaryError             Not yet implemented.
ConfusionMatrix             Not yet implemented.

Unsupported, Special Use

Count 0/5

NNabla Function 6 7 9 10 11 ONNX Op Description
VATNoise             Not yet implemented.
Unlink             Not yet implemented.
Sink             Not yet implemented.
NmsDetection2d             Not yet implemented.
MaxPoolingBackward             Not yet implemented.

Tensorflow Support Status

Note:In this document, the numbers in the header of all tables represent the version of onnx opset.

Import

  • ✓: Supported
  • △: Partially supported
  • X: Supported, but test failed.
  • Empty: Not support yet.

Total: 86/120

Tensorflow support status
Tensorflow Function Status NNabla Func Description
Abs Abs  
Acos ACos  
Acosh ACosh  
Add Add2  
AddN Add2  
All     Not yet implemented.
Any     Not yet implemented.
ArgMax Max  
ArgMin Min  
Asin ASin  
Asinh ASinh  
Atan ATan  
Atan2     Not yet implemented.
Atanh ATanh  
AvgPool Pad, Transpose, AveragePooling Some feature is not supported by Nnabla such as Pad’s edge mode.
AvgPool3D     Not yet implemented.
BatchMatMul BatchMatmul, Transpose  
BiasAdd Reshape, Add2  
Cast     Not yet implemented.
Ceil Ceil  
ConcatV2 Concatenate  
Const Add2  
Conv2D Pad, Convolution, Transpose Tensorflow require GPU to perform related test cases. This issue is recorded only for memo.
Conv2DBackpropFilter     Not yet implemented.
Conv2DBackpropInput Deconvolution, Transpose Tensorflow require GPU to perform related test cases. This issue is recorded only for memo.
Conv3D     Not yet implemented.
Conv3DBackpropFilterV2     Not yet implemented.
Conv3DBackpropInputV2     Not yet implemented.
Cos Cos  
Cosh Cosh  
DepthToSpace Reshape, Transpose Tensorflow require GPU to perform related test cases. This issue is recorded only for memo.
DepthwiseConv2dNative     Not yet implemented.
DepthwiseConv2dNativeBackpropFilter     Not yet implemented.
DepthwiseConv2dNativeBackpropInput     Not yet implemented.
Div Div2  
Elu ELU  
Equal Equal  
Erf     Not yet implemented.
Erfc     Not yet implemented.
Exp Exp  
ExpandDims Reshape  
Fill     Not yet implemented.
Flatten Reshape  
Floor Floor  
FloorDiv Div2, Floor  
FloorMod Div2, Sub2, Mul2, Floor  
FusedBatchNorm BatchNormalization, Transpose It did not pass testing for training mode.
GatherNd     Not yet implemented.
GatherV2     Not yet implemented.
Greater Greater  
GreaterEqual Less, LogicalNot  
Identity Identity  
IsInf     Not yet implemented.
IsNan IsNaN  
LeakyRelu LeakyReLU  
Less Less  
LessEqual Greater, LogicalNot  
Log Log  
LogSoftmax     Not yet implemented.
LogicalAnd LogicalAnd  
LogicalNot LogicalNot  
LogicalOr LogicalOr  
LogicalXor LogicalAnd, LogicalOr, LogicalNot  
MatrixBandPart     Not yet implemented.
Max Max  
MaxPool Transpose, Pad, MaxPooling  
MaxPool3D     Not yet implemented.
MaxPoolWithArgmax     Not yet implemented.
Maximum Maximum2  
Mean Mean  
Min Min  
Minimum Minimum2  
Mul Mul2  
Neg MulScalar  
NotEqual Equal, LogicalNot  
OneHot     Not yet implemented.
Pack Reshape, Concatenate  
Pad Pad  
Pow Pow2  
Prod Prod  
RandomShuffle     Not yet implemented.
RandomStandardNormal     Not yet implemented.
RandomUniform     Not yet implemented.
RealDiv Div2  
Reciprocal RDivScalar  
Relu ReLU  
Relu6 MinimumScalar, MaximumScalar  
Reshape Reshape Some test cases failed for some nnabla’s implementation limitation (e.g. -1 is regarded as batch_size).
ReverseSequence     Not yet implemented.
Rsqrt PowScalar, RDivScalar  
Select     Not yet implemented.
Selu SELU  
Shape     Not yet implemented.
Sigmoid Sigmoid  
Sign Sign  
Sin Sin  
Sinh Sinh  
Size     Not yet implemented.
Slice Slice  
Softmax     Not yet implemented.
Softplus SoftPlus  
Softsign SoftSign  
SpaceToDepth Reshape, Transpose Tensorflow require GPU to perform related test cases. This issue is recorded only for memo.
SplitV Stack, Split  
Sqrt PowScalar  
Square Mul2  
SquaredDifference Sub2, Mul2  
Squeeze Reshape  
StopGradient Identity  
StridedSlice Slice  
Sub Sub2  
Sum Sum  
Tan Tan  
Tanh Tanh  
Tile Tile  
TopKV2     Not yet implemented.
Transpose Transpose  
TruncateDiv     Not yet implemented.
TruncateMod     Not yet implemented.
Unpack Reshape, Stack, Split, Concatenate  

Export

  • ✓: Supported
  • △: Partially supported
  • X: Supported, but test failed.
  • Empty: Not support yet.

Total: 115/173

Neural Network Layer

Count 11/14

NNabla Function Status TF Op Description
Affine Reshape, Placeholder, MatMul, Const, Mul, Add  
RNN     Not yet implemented.
LSTM     Not yet implemented.
GRU     Not yet implemented.
Convolution Reshape, SpaceToBatchND, Pad, Placeholder, Transpose, Const, Add, Split, Identity, ConcatV2, Conv2D, BatchToSpaceND The cases dilations and strides larger than 1 are not supported by tensorflow.
DepthwiseConvolution Conv2D, Reshape, SpaceToBatchND, Pad, Placeholder, Transpose, Const, Split, ConcatV2, Add, BatchToSpaceND The cases dilations and strides larger than 1 are not supported by tensorflow.
Deconvolution Reshape, Slice, Placeholder, Transpose, Const, Split, Identity, ConcatV2, Conv2DBackpropInput, Add The cases dilations larger than 1 are not supported by tensorflow.
DepthwiseDeconvolution Reshape, Slice, Placeholder, Transpose, Const, Split, ConcatV2, Conv2DBackpropInput, Add The cases dilations larger than 1 are not supported by tensorflow.
MaxPooling Reshape, PadV2, Placeholder, Transpose, Const, MaxPool, MaxPool3D  
AveragePooling AvgPool3D, Reshape, Pad, Placeholder, Transpose, Const, AvgPool Currently only supports the cases both ignore_border and including_pad are True.
GlobalAveragePooling Const, Pack, Sub, SplitV, Range, Mean  
SumPooling AvgPool3D, Reshape, Pad, Placeholder, Transpose, Const, Mul, AvgPool  
Unpooling Reshape, Placeholder, Transpose, ResizeNearestNeighbor, Const, LogicalAnd, Merge, Switch, Identity, Assert, Mul, Equal, StridedSlice, NoOp, Cast The kernel only supports 2d.
Embed GatherV2, Placeholder, Const  

Neural Network Activation Functions

Count 21/21

NNabla Function Status TF Op Description
Sigmoid Sigmoid, Placeholder  
Swish Mul, Sigmoid, Placeholder  
Tanh Placeholder, Tanh  
ReLU Relu, Placeholder  
LeakyReLU LeakyRelu, Placeholder  
Softmax Placeholder, Const, Exp, Sub, RealDiv, Sum, Max  
LogSoftmax Placeholder, Const, Exp, Sub, Log, Sum, Max  
ELU Less, Placeholder, Const, Elu, Exp, Mul, GreaterEqual, Sub, Cast, Add  
SELU Minimum, Placeholder, Const, Exp, Mul, Sub, Maximum, Add  
CReLU Relu, Placeholder, Const, ConcatV2, Neg  
CELU Less, Placeholder, Const, Elu, Exp, ConcatV2, Mul, GreaterEqual, Sub, Neg, Cast, Add  
PReLU Reshape, Relu, Placeholder, Const, Mul, Sub, Abs, Add  
GELU Sqrt, Placeholder, Const, Pow, Tanh, Mul, RealDiv, Add  
ReLU6 Relu, Placeholder, Const, Pack, Min  
HardSigmoid Minimum, Placeholder, Const, Mul, Maximum, Add  
HardTanh Placeholder, Const, Pack, Neg, Min, Max  
LogSigmoid Sigmoid, Log, Placeholder  
SoftPlus Softplus, Placeholder  
SoftSign Placeholder, Softsign  
TanhShrink Sub, Placeholder, Tanh  
Sinc Placeholder, Select, Const, Sin, Equal, RealDiv  

Normalization

Count 1/6

NNabla Function Status TF Op Description
FusedBatchNormalization X Reshape, Relu, Placeholder, Const, Mul, Rsqrt, Sub, Mean, RealDiv, Sum, Add Not yet implemented.
BatchNormalization Reshape, Placeholder, Const, Mul, Rsqrt, Sub, Mean, RealDiv, Sum, Add In inferring stage, caffe2 mistmatch onnx 1.4.x’s implementation, “in-place” feature cannot be applied.
SyncBatchNormalization     Not yet implemented.
MeanSubtraction     Not yet implemented.
ClipGradByValue     Not yet implemented.
ClipGradByNorm     Not yet implemented.

Reduction

Count 5/7

NNabla Function Status TF Op Description
Sum Sum, Placeholder, Const  
Mean Mean, Placeholder, Const  
Max Placeholder, Max, Const  
Min Min, Placeholder, Const  
Prod Prod, Placeholder, Const  
ReduceSum     Not yet implemented.
ReduceMean     Not yet implemented.

Arithmetic

Count 11/12

NNabla Function Status TF Op Description
Add2 Placeholder, Add  
BcAdd2     Not yet implemented.
Sub2 Sub, Placeholder  
Mul2 Mul, Placeholder  
Div2 RealDiv, Placeholder  
Pow2 Pow, Placeholder  
AddScalar Placeholder, Add, Const  
MulScalar Mul, Placeholder, Const  
PowScalar Pow, Placeholder, Const  
RSubScalar Sub, Placeholder, Const  
RDivScalar RealDiv, Placeholder, Const  
RPowScalar Pow, Placeholder, Const  

Logical

Count 27/29

NNabla Function Status TF Op Description
Sign Placeholder, Sign  
Minimum2 Placeholder, Const, Pack, Min, Add  
Maximum2 Placeholder, Add, Const, Pack, Max  
MinimumScalar Placeholder, Const, Pack, Min, Add  
MaximumScalar Placeholder, Add, Const, Pack, Max  
LogicalAnd LogicalAnd, Placeholder  
LogicalOr LogicalOr, Placeholder  
LogicalXor LogicalAnd, LogicalOr, Placeholder, LogicalNot  
Equal Equal, Placeholder  
NotEqual LogicalNot, Equal, Placeholder  
GreaterEqual Less, Placeholder, LogicalNot  
Greater Greater, Placeholder  
LessEqual Greater, Placeholder, LogicalNot  
Less Less, Placeholder  
LogicalAndScalar LogicalAnd, Placeholder, Const  
LogicalOrScalar LogicalOr, Placeholder, Const  
LogicalXorScalar Placeholder, Const, LogicalAnd, LogicalOr, LogicalNot  
EqualScalar Equal, Placeholder, Const  
NotEqualScalar LogicalNot, Equal, Placeholder, Const  
GreaterEqualScalar Less, Placeholder, LogicalNot, Const  
GreaterScalar Greater, Placeholder, Const  
LessEqualScalar Greater, Placeholder, LogicalNot, Const  
LessScalar Less, Placeholder, Const  
LogicalNot Placeholder, LogicalNot  
IsNaN IsNan, Placeholder  
IsInf X   Not yet implemented.
ResetNaN IsNan, Placeholder, Select, Const  
ResetInf X   Not yet implemented.
Where Placeholder, Select  

Math

Count 21/22

NNabla Function Status TF Op Description
Constant Identity, Const  
Arange Identity, Const  
Abs Abs, Placeholder  
Exp Exp, Placeholder  
Log Log, Placeholder  
Identity Placeholder, Identity  
BatchMatmul Reshape, Placeholder, Transpose, Const, BatchMatMulV2  
Round X   Not yet implemented.
Ceil Ceil, Placeholder  
Floor Floor, Placeholder  
Sin Sin, Placeholder  
Cos Cos, Placeholder  
Tan Tan, Placeholder  
Sinh Sinh, Placeholder  
Cosh Cosh, Placeholder  
ASin Asin, Placeholder  
ACos Acos, Placeholder  
ATan Placeholder, Atan  
ATan2 RealDiv, Placeholder, Atan  
ASinh Asinh, Placeholder  
ACosh Acosh, Placeholder  
ATanh Atanh, Placeholder  

Array Manipulation

Count 12/19

NNabla Function Status TF Op Description
Concatenate ConcatV2, Placeholder, Const  
Split Squeeze, SplitV, Placeholder, Const  
Stack ConcatV2, ExpandDims, Placeholder, Const  
Slice Slice, Placeholder, Const step != 1” exceed the scope of onnx opset 9, not supported.
Pad MirrorPad, PadV2, Placeholder, Const When the mode of the pad is reflect, if the size of the pad exceeds the input size, tensorflow cannot handle it.
Transpose Placeholder, Transpose, Const  
Broadcast    
BroadcastTo    
Tile Reshape, Tile, Placeholder, Const  
OneHot Reshape, GatherV2, Placeholder, Const  
Flip Placeholder, Transpose, Const, Identity, GatherV2  
Shift     Not yet implemented.
Sort     Not yet implemented.
Reshape Reshape, Placeholder, Const  
MatrixDiag     Not yet implemented.
MatrixDiagPart     Not yet implemented.
Assign     Not yet implemented.
GatherNd     Not yet implemented.
ScatterNd     Not yet implemented.

Signal Processing

Count 0/3

NNabla Function Status TF Op Description
Interpolate X   Not yet implemented.
FFT     Not yet implemented.
IFFT     Not yet implemented.

Stochasticity

Count 0/11

NNabla Function Status TF Op Description
Dropout X Placeholder The Dropout in nnabla has no test mode and contains random parameters, so the test result is not the same as tensorflow.
TopKData     Not yet implemented.
TopKGrad     Not yet implemented.
Rand     Not yet implemented.
Randint     Not yet implemented.
Randn     Not yet implemented.
RandomChoice     Not yet implemented.
RandomCrop     Not yet implemented.
RandomFlip     Not yet implemented.
RandomShift     Not yet implemented.
ImageAugmentation     Not yet implemented.

Loss Functions

Count 0/9

NNabla Function Status TF Op Description
SigmoidCrossEntropy     Not yet implemented.
BinaryCrossEntropy     Not yet implemented.
SoftmaxCrossEntropy     Not yet implemented.
CategoricalCrossEntropy     Not yet implemented.
SquaredError     Not yet implemented.
AbsoluteError     Not yet implemented.
HuberLoss     Not yet implemented.
EpsilonInsensitiveLoss     Not yet implemented.
KLMultinomial     Not yet implemented.

Quantization Neural Network Layers

Count 6/12

NNabla Function Status TF Op Description
BinarySigmoid Greater, Placeholder, Select, Const  
BinaryTanh Greater, Placeholder, Select, Const  
BinaryConnectAffine Reshape, Placeholder, MatMul, Const, Mul, Add  
BinaryConnectConvolution Conv2D, Reshape, Pad, Placeholder, Transpose, Const, Split, Identity, ConcatV2, Add The cases dilations and strides larger than 1 are not supported by tensorflow.
BinaryWeightAffine Reshape, Placeholder, MatMul, Const, Mul, Add  
BinaryWeightConvolution Conv2D, Reshape, Pad, Placeholder, Transpose, Const, Split, Identity, Mul, ConcatV2, Add The cases dilations and strides larger than 1 are not supported by tensorflow.
INQAffine     Not yet implemented.
INQConvolution     Not yet implemented.
FixedPointQuantize     Not yet implemented.
MinMaxQuantize     Not yet implemented.
Pow2Quantize     Not yet implemented.
Prune     Not yet implemented.

Validation

Count 0/3

NNabla Function Status TF Op Description
TopNError     Not yet implemented.
BinaryError     Not yet implemented.
ConfusionMatrix     Not yet implemented.

Unsupported, Special Use

Count 0/5

NNabla Function Status TF Op Description
VATNoise     Not yet implemented.
Unlink     Not yet implemented.
Sink     Not yet implemented.
NmsDetection2d     Not yet implemented.
MaxPoolingBackward     Not yet implemented.

NNabla C Runtime Support Status

nnabla version: 1.0.21

  • ✓: Supported
  • △: Partially supported
  • X: Supported, but test failed or no test data.
  • Empty: Not support yet.

Export

Total: 56/173

Neural Network Layer

Count 8/14

NNabla Function Status Description
Affine  
RNN    
LSTM    
GRU    
Convolution  
DepthwiseConvolution  
Deconvolution  
DepthwiseDeconvolution    
MaxPooling  
AveragePooling  
GlobalAveragePooling    
SumPooling  
Unpooling  
Embed    

Neural Network Activation Functions

Count 11/21

NNabla Function Status Description
Sigmoid  
Swish  
Tanh  
ReLU  
LeakyReLU  
Softmax  
LogSoftmax    
ELU  
SELU  
CReLU  
CELU  
PReLU  
GELU    
ReLU6    
HardSigmoid    
HardTanh    
LogSigmoid    
SoftPlus    
SoftSign    
TanhShrink    
Sinc    

Normalization

Count 1/6

NNabla Function Status Description
FusedBatchNormalization    
BatchNormalization  
SyncBatchNormalization    
MeanSubtraction X  
ClipGradByValue    
ClipGradByNorm    

Reduction

Count 1/7

NNabla Function Status Description
Sum  
Mean    
Max    
Min    
Prod    
ReduceSum    
ReduceMean    

Arithmetic

Count 11/12

NNabla Function Status Description
Add2  
BcAdd2    
Sub2  
Mul2  
Div2  
Pow2  
AddScalar  
MulScalar  
PowScalar  
RSubScalar  
RDivScalar  
RPowScalar  

Logical

Count 5/29

NNabla Function Status Description
Sign  
Minimum2  
Maximum2  
MinimumScalar  
MaximumScalar  
LogicalAnd    
LogicalOr    
LogicalXor    
Equal    
NotEqual    
GreaterEqual    
Greater    
LessEqual    
Less    
LogicalAndScalar    
LogicalOrScalar    
LogicalXorScalar    
EqualScalar    
NotEqualScalar    
GreaterEqualScalar    
GreaterScalar    
LessEqualScalar    
LessScalar    
LogicalNot    
IsNaN    
IsInf    
ResetNaN    
ResetInf    
Where    

Math

Count 6/22

NNabla Function Status Description
Constant    
Arange    
Abs  
Exp  
Log  
Identity  
BatchMatmul  
Round  
Ceil    
Floor    
Sin    
Cos    
Tan    
Sinh    
Cosh    
ASin    
ACos    
ATan    
ATan2    
ASinh    
ACosh    
ATanh    

Array Manipulation

Count 7/19

NNabla Function Status Description
Concatenate  
Split  
Stack  
Slice  
Pad    
Transpose  
Broadcast    
BroadcastTo    
Tile    
OneHot    
Flip  
Shift X  
Sort    
Reshape  
MatrixDiag X  
MatrixDiagPart X  
Assign    
GatherNd    
ScatterNd    

Signal Processing

Count 0/3

NNabla Function Status Description
Interpolate    
FFT    
IFFT    

Stochasticity

Count 0/11

NNabla Function Status Description
Dropout X  
TopKData    
TopKGrad    
Rand    
Randint    
Randn    
RandomChoice    
RandomCrop    
RandomFlip    
RandomShift    
ImageAugmentation    

Loss Functions

Count 0/9

NNabla Function Status Description
SigmoidCrossEntropy    
BinaryCrossEntropy    
SoftmaxCrossEntropy    
CategoricalCrossEntropy    
SquaredError    
AbsoluteError    
HuberLoss    
EpsilonInsensitiveLoss    
KLMultinomial    

Quantization Neural Network Layers

Count 6/12

NNabla Function Status Description
BinarySigmoid  
BinaryTanh  
BinaryConnectAffine  
BinaryConnectConvolution  
BinaryWeightAffine  
BinaryWeightConvolution  
INQAffine    
INQConvolution    
FixedPointQuantize    
MinMaxQuantize    
Pow2Quantize    
Prune    

Validation

Count 0/3

NNabla Function Status Description
TopNError    
BinaryError    
ConfusionMatrix    

Unsupported, Special Use

Count 0/5

NNabla Function Status Description
VATNoise    
Unlink    
Sink    
NmsDetection2d    
MaxPoolingBackward