Skip to content

Latest commit

 

History

History
341 lines (304 loc) · 20.8 KB

SUPPORT_STATUS.md

File metadata and controls

341 lines (304 loc) · 20.8 KB
Raw

Implement status

Total 63/228

Neural Network Layer

Count 8/18

Function Available float generic
Affine yes yes yes
RNN no - -
LSTM no - -
GRU no - -
Convolution yes yes yes
FusedConvolution no - -
DepthwiseConvolution yes yes yes
Deconvolution yes yes yes
DepthwiseDeconvolution no - -
DeformableConvolution no - -
AdaptiveSeparableConvolution no - -
MaxPooling yes yes yes
AveragePooling yes yes yes
GlobalAveragePooling no - -
SumPooling yes yes yes
Unpooling yes yes yes
Embed no - -
RoiAlign no - -

Neural Network Activation Functions

Count 11/22

Function Available float generic
Sigmoid yes yes yes
Swish yes yes yes
Tanh yes yes yes
ReLU yes yes yes
LeakyReLU yes yes yes
Softmax yes yes yes
LogSoftmax no - -
ELU yes yes yes
SELU yes yes yes
CReLU yes yes yes
CELU yes yes yes
PReLU yes yes yes
GELU no - -
Mish no - -
ReLU6 no - -
HardSigmoid no - -
HardTanh no - -
LogSigmoid no - -
SoftPlus no - -
SoftSign no - -
TanhShrink no - -
Sinc no - -

Normalization

Count 2/14

Function Available float generic
FusedBatchNormalization no - -
BatchNormalization yes yes yes
GroupNormalization no - -
InstanceNormalization no - -
LayerNormalization no - -
NormNormalization no - -
SyncBatchNormalization no - -
TensorNormalization no - -
WeightNormalization no - -
WeightStandardization no - -
SpectralNorm no - -
MeanSubtraction yes yes yes
ClipGradByValue no - -
ClipGradByNorm no - -

Reduction

Count 1/10

Function Available float generic
Sum yes yes yes
CumSum no - -
Mean no - -
Max no - -
Min no - -
Norm no - -
Prod no - -
CumProd no - -
ReduceSum no - -
ReduceMean no - -

Arithmetic

Count 11/14

Function Available float generic
Add2 yes yes yes
AddN no - -
BcAdd2 no - -
Sub2 yes yes yes
Mul2 yes yes yes
MulN no - -
Div2 yes yes yes
Pow2 yes yes yes
AddScalar yes yes yes
MulScalar yes yes yes
PowScalar yes yes yes
RSubScalar yes yes yes
RDivScalar yes yes yes
RPowScalar yes yes yes

Logical

Count 5/30

Function Available float generic
Sign yes yes yes
Minimum2 yes yes yes
Maximum2 yes yes yes
MinimumScalar yes yes yes
MaximumScalar yes yes yes
LogicalAnd no - -
LogicalOr no - -
LogicalXor no - -
Equal no - -
NotEqual no - -
GreaterEqual no - -
Greater no - -
LessEqual no - -
Less no - -
SearchSorted no - -
LogicalAndScalar no - -
LogicalOrScalar no - -
LogicalXorScalar no - -
EqualScalar no - -
NotEqualScalar no - -
GreaterEqualScalar no - -
GreaterScalar no - -
LessEqualScalar no - -
LessScalar no - -
LogicalNot no - -
IsNaN no - -
IsInf no - -
ResetNaN no - -
ResetInf no - -
Where no - -

Math

Count 6/24

Function Available float generic
Constant no - -
Arange no - -
Linspace no - -
Abs yes yes yes
Exp yes yes yes
Log yes yes yes
Identity yes yes yes
BatchMatmul yes yes yes
Round yes yes yes
Ceil no - -
Floor no - -
Sin no - -
Cos no - -
Tan no - -
Sinh no - -
Cosh no - -
ASin no - -
ACos no - -
ATan no - -
ATan2 no - -
ASinh no - -
ACosh no - -
ATanh no - -
Erf no - -

Array Manipulation

Count 11/34

Function Available float generic
Concatenate yes yes yes
Split yes yes yes
Stack yes yes yes
Slice yes yes yes
Pad yes yes yes
Transpose yes yes yes
Broadcast no - -
BroadcastTo no - -
Tile no - -
OneHot no - -
Flip yes yes yes
Shift yes yes yes
Sort no - -
Reshape yes yes yes
Shape no - -
MatrixDiag yes yes yes
MatrixDiagPart yes yes yes
Trilu no - -
Meshgrid no - -
BatchDet no - -
BatchInv no - -
BatchLogdet no - -
BatchCholesky no - -
Assign no - -
Gather no - -
GatherNd no - -
BoolGather no - -
ScatterNd no - -
ScatterAdd no - -
BoolScatter no - -
BoolFill no - -
PackPaddedSequence no - -
PadPackedSequence no - -
NonZero no - -

Signal Processing

Count 0/6

Function Available float generic
Interpolate no - -
ONNXResize no - -
FFT no - -
IFFT no - -
STFT no - -
ISTFT no - -

Stochasticity

Count 1/15

Function Available float generic
Dropout yes yes yes
TopKData no - -
TopKGrad no - -
Rand no - -
Randint no - -
Randn no - -
RandBinomial no - -
RandBeta no - -
RandGamma no - -
RandomChoice no - -
RandomCrop no - -
RandomFlip no - -
RandomShift no - -
RandomErase no - -
ImageAugmentation no - -

Loss Functions

Count 0/9

Function Available float generic
SigmoidCrossEntropy no - -
BinaryCrossEntropy no - -
SoftmaxCrossEntropy no - -
CategoricalCrossEntropy no - -
SquaredError no - -
AbsoluteError no - -
HuberLoss no - -
EpsilonInsensitiveLoss no - -
KLMultinomial no - -

Geometric Neural Network Layers

Count 0/3

Function Available float generic
AffineGrid no - -
WarpByGrid no - -
WarpByFlow no - -

Quantization Neural Network Layers

Count 6/14

Function Available float generic
BinarySigmoid yes yes yes
BinaryTanh yes yes yes
BinaryConnectAffine yes yes yes
BinaryConnectConvolution yes yes yes
BinaryWeightAffine yes yes yes
BinaryWeightConvolution yes yes yes
INQAffine no - -
INQConvolution no - -
FixedPointQuantize no - -
MinMaxQuantize no - -
Pow2Quantize no - -
Prune no - -
QuantizeLinear no - -
DequantizeLinear no - -

Validation

Count 0/3

Function Available float generic
TopNError no - -
BinaryError no - -
ConfusionMatrix no - -

Unsupported, Special Use

Count 1/12

Function Available float generic
VATNoise no - -
Unlink no - -
Sink no - -
NmsDetection2d no - -
ONNXNonMaxSuppression no - -
MaxPoolingBackward no - -
PatchCorrelation no - -
Unique no - -
EyeLike no - -
Mod2 no - -
BitShift no - -
Einsum yes ? ?