[Feature Request]: Implement Onnx operators

openvinotoolkit / openvino

OpenVINO™ is an open-source toolkit for optimizing and deploying AI inference

https://docs.openvino.ai

Apache License 2.0

6.86k stars 2.19k forks source link

[Feature Request]: Implement Onnx operators #26700

Open avidiyal opened 3 hours ago

avidiyal commented 3 hours ago

Request Description

ML.Net models which use ONNX in the backend has some operators which are not supported by OpenVINO

Like "ai.onnx.ml.LabelEncoder, ai.onnx.ml.OneHotEncoder, ai.onnx.ml.Imputer, ai.onnx.ml.FeatureVectorizer, ai.onnx.ml.TreeEnsembleRegressor"

MLnet is becoming more predominantly being used for client usecases.

These are part of ONNX operator support https://onnx.ai/onnx/operators/onnx_aionnxml_LabelEncoder.html

Feature Use Case

No response

Issue submission checklist

[X] The feature request or improvement must be related to OpenVINO

rkazants commented 2 hours ago

Hi @avidiyal, thanks for the ticket.

@gkrivor, we can support some of these operations in ONNX FE directly (like ai.onnx.ml.FeatureVectorizer, ai.onnx.ml.OneHotEncoder). ai.onnx.ml.LabelEncoder is a sort of Lookup operation already implemented in openvino-tokenizers for TF FE and we can re-use it. ai.onnx.ml.TreeEnsembleRegressor is more for ML domain rather than DL so I have doubt about adding this support.

Best regards, Roman