[tmva][sofie] Status of the support of ONNX operators

[axmat]

This issue should track the status of the support of ONNX operators in TMVA/SOFIE.

Supported :

• Add
• AveragePool
• BatchNormalization
• Conv
• Gemm
• GlobalAveragePool
• GRU
• LSTM
• MaxPool
• Pool
• Relu
• Reshape
• RNN
• Selu
• Sigmoid
• Slice
• Transpose

Work in progress :

• Softmax, [TMVA][SOFIE] New developments in SOFIE for parsing ONNX files  #10315
• Identity, [TMVA][SOFIE] New developments in SOFIE for parsing ONNX files  #10315
• Concat, [TMVA][SOFIE] New developments in SOFIE for parsing ONNX files  #10315 and [TMVA][SOFIE] Add Concat Operator #9311
• InstanceNormalization, [tmva][sofie][gsoc] Add InstanceNormalization operator #8885

Missing :

• Abs
• Acos
• Acosh
• And
• ArgMax
• ArgMin
• Asin
• Asinh
• Atan
• Atanh
• BitShift
• Cast
• Ceil
• Clip
• Compress
• ConcatFromSequence
• Constant
• ConstantOfShape
• ConvInteger
• ConvTranspose
• Cos
• Cosh
• CumSum
• DepthToSpace
• DequantizeLinear
• Det
• Div
• Dropout
• Einsum
• Elu
• Equal
• Erf
• Exp
• Expand
• EyeLike
• Flatten
• Floor
• Gather
• GatherElements
• GatherND
• GlobalLpPool
• GlobalMaxPool
• Greater
• GridSample
• HardSigmoid
• Hardmax
• Identity
• If
• IsInf
• IsNaN
• LRN
• LeakyRelu
• Less
• Log
• Loop
• LpNormalization
• LpPool
• MatMul
• MatMulInteger
• Max
• MaxRoiPool
• MaxUnpool
• Mean
• Min
• Mod
• Mul
• Multinomial
• Neg
• NonMaxSuppression
• NonZero
• OneHot
• Optional
• OptionalGetElement
• OptionalHasElement
• Or
• PRelu
• Pad
• Pow
• QLinearConv
• QLinearMatMul
• QuantizeLinear
• RandomNormal
• RandomNormalLike
• RandomUniform
• RandomUniformLike
• Reciprocal
• ReduceL1
• ReduceL2
• ReduceLogSum
• ReduceLogSumExp
• ReduceMax
• ReduceMean
• ReduceMin
• ReduceSum
• ReduceSumSquare
• Resize
• ReverseSequence
• RoiAlign
• Round
• Scan
• ScatterElements
• ScatterND
• SequenceAt
• SequenceConstruct
• SequenceEmpty
• SequenceErase
• SequenceInsert
• SequenceLength
• Shape
• Shrink
• Sign
• Sin
• Sinh
• Size
• Softplus
• Softsign
• SpaceToDepth
• Split
• SplitToSequence
• Sqrt
• StringNormalizer
• Sub
• Sum
• Tan
• Tanh
• TfldfVectorizer
• ThresholdedRelu
• Tile
• TopK
• Trilu
• Unique
• Unsqueeze
• Where
• Xor

[youlxs]

Do we have the plan to support "MatMul" ?

[lmoneta]

MatMul is now supported in both master and 6.30, see

#12894

This above list needs to be updated with the new supported operator

[youlxs]

Do we have a plan to support "Split" ?

[lmoneta]

We can plan to support it if needed. It might take a little bit of time, since we need to support in SOFIE sequence of tensors which are the output of Split. For which type of model do you need Split?

[youlxs]

MatMul is now supported in both master and 6.30, see

#12894

This above list needs to be updated with the new supported operator

thanks

[youlxs]

We can plan to support it if needed. It might take a little bit of time, since we need to support in SOFIE sequence of tensors which are the output of Split. For which type of model do you need Split?

I want to convert my model to the raw code to improve the prediction speed. As I saw from your slide the raw code generated through tmva has better performance than the original onnx runtime.

[Shlok-Saxena]

Hi, I saw this issue about documenting ONNX operator support in SOFIE. I’ve been working with TMVA SOFIE parsers recently and added documentation that reflects the operators currently registered in the codebase.

I’ve opened a PR with these changes: #20811

[Shlok-Saxena]

Hi, I worked on documenting the supported and missing ONNX operators in SOFIE parsers and opened PR #20811 for it.

Initially, the operator list was added as a README.md in root/tmva/sofie_parsers. Since I’ve now opened a separate PR #20825 to introduce the main root/tmva/sofie_parsers/README.md, I’m going to move the operator list into a dedicated document ONNX_OPERATOR_STATUS.md under root/tmva/sofie directory and link it from the existing root/tmva/sofie/README.md — instead of keeping it in the parsers README.

This avoids overlapping changes and keeps the README focused.

Feedback on the preferred location and format for the operator list is welcome!