Page tree
Skip to end of metadata
Go to start of metadata

Introduction:

This page tracks the current status and development  to support ONNX  on Mxnet.  We currently support import of ONNX models into Mxnet.


ONNX-MXNET

We have two repositories for the import module:

  1. onnx-mxnet repository inside onnx https://github.com/onnx/onnx-mxnet  ( soon to be deprecated)
  2. A newly created and refactored module inside Mxnet -> contrib. https://github.com/apache/incubator-mxnet/tree/master/python/mxnet/contrib/onnx

 

Operator Coverage:

This table keeps track of the status of all ONNX operators supported by Mxnet.
ONNX backend test script reports the coverage on the operators and attributes.

OperatorTest CoverageMxnet
AbsYesOK
AddYesOK
And 

 

ArgMax OK
ArgMin OK
AveragePoolYesOK
BatchNormalizationYesOK
CastYesOK
Ceil OK
ClipYesNOT OK, under development
ConcatYesOK
ConstantYesNOT OK
ConvYesOK
ConvTranspose OK
DepthToSpace NOT OK, under development
DivYesOK
Dropout OK
EluYesOK
EqualYesNOT OK
ExpYesOK
FlattenYesPARTIAL, only supports default axis=1
Floor OK
GRU NOT OK , under development
GatherYesNOT OK, under development
GemmYesOK
GlobalAveragePoolYesOK
GlobalLpPool NOT OK, under development
GlobalMaxPoolYesOK
Greater NOT OK, under development
HardSigmoid NOT OK, under development
Hardmax NOT OK, under development
InstanceNormalization NOT OK, under development
LRN OK
LSTM NOT OK, under development
LeakyReluYesOK
Less NOT OK, under development
LogYesOK
LogSoftMax NOT OK, under development
LpNormalization NOT OK, under development
LpPool NOT OK, under development
LpPoolYesOK
MaxYesOK
MaxPoolYesOK
MaxRoiPool NOT OK, under development
Mean NOT OK, under development
MinYesOK
MulYesOK
NegYesOK
Not NOT OK, under development
Or NOT OK, under development
PReluYesOK
PadYesOK
PowYesPartial, only supports default axis=1
RNN NOT OK, under development
RandomNormal OK
RandomNormalLike OK
RandomUniform OK
RandomUniformLike OK
Reciprocal OK
ReduceL1 NOT OK, under development
ReduceL2 NOT OK, under development
ReduceLogSum NOT OK, under development
ReduceLogSumExp NOT OK, under development
ReduceMaxYesOK
ReduceMeanYesOK
ReduceMinYesOK
ReduceProdYesOK
ReduceSumYesOK
ReduceSumSquare NOT OK, under development
ReluYesOK
ReshapeYesOK
SeluYesNOT OK, under development
SigmoidYesOK
SliceYesNOT OK, under development
SoftmaxYesPARTIAL, only supports default axis=1
SoftplusYesNOT OK, under development
Softsign NOT OK, under development
SpaceToDepth NOT OK, under development
SplitYesNOT OK, under development
Sqrt OK
SqueezeYesOK
SubYesOK
SumYesOK
TanhYesOK
Tile NOT OK, under development
TransposeYesOK
Xor NOT OK, under development
experimental ATen NOT OK, under development
experimental Affine NOT OK, under development
experimental ConstantFill NOT OK, under development
experimental Crop NOT OK, under development
experimental FC NOT OK, under development
experimental GRUUnit NOT OK, under development
experimental GivenTensorFill NOT OK, under development
experimental Identity NOT OK, under development
experimental ImageScaler NOT OK, under development
experimental MeanVarianceNormalization NOT OK, under development
experimental ParametricSoftplus NOT OK, under development
experimental Scale NOT OK, under development
experimental ScaledTanh NOT OK, under development
experimental ThresholdedRelu NOT OK, under development
experimental Upsample NOT OK, under development

 

ONNX models coverage:

ONNX backend tests include below models. The table lists current status on Mxnet:

 

ONNX ModelsMxnet Support
bvlc_alexnetOK
densenet121OK
inception_v1NOT OK, precision difference due to issue
inception_v2NOT OK, precision difference due to issue
resnet50OK
shufflenetNOT OK, precision difference due to issue
squeezenetOK
vgg16OK
vgg19OK
bvlc_googlenetOK
bvlc_caffenetOK
bvlc-rcnn-ilsvrc13OK

 

 

 

 

 

 

 

 

 

 

 

  • No labels