[Feature Request]: Add SeLu activation in the OpenVino IR (TensorFlow Conversion)

[VimalWill]

Request Description

I quantized my TensorFlow model to INT-8 using NNCF and trying to convert to OpenVino IR for performance analysis , but I was struck with the error as I'm using SeLu activation in my model. Ig OpenVino IR don't have support for SeLu?

Feature Use Case

No response

Issue submission checklist

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

[vurusovs]

@VimalWill according to documentation, SeLu is supported - https://docs.openvino.ai/2023.0/openvino_docs_ops_activation_Selu_1.html or https://docs.openvino.ai/2023.0/classov_1_1op_1_1v0_1_1Selu.html

Could you provide an example of code you use and explain the error you receive?

[VimalWill]

Hi @vurusovs This is the issue and my version is (2023.0.1-11005-fa1c41994f3-releases/2023/0) OpConversionFailure: Check 'is_conversion_successful' failed at src/frontends/tensorflow/src/frontend.cpp:264: FrontEnd API failed with OpConversionFailure: : [TensorFlow Frontend] Internal error, no translator found for operation(s): Selu

[VimalWill]

ov_model = mo.convert_model(tf_model) ov_quantized_model = mo.convert_model(tf_quantized_model)

I followed this code from doc of nncf

[rkazants]

https://github.com/openvinotoolkit/openvino/pull/19528 with Selu support is under review.

[rkazants]

@VimalWill, PR was merged to master. If you have a chance, please build the master. And the operation is going to the next release 2023.2

Best regards, Roman

[VimalWill]

@rkazants thanks, let me try building it

[VimalWill]

can any one suggest me the steps to build from master, I'm still getting the issue and ig i'm not building it properly. Python ver: 3.10

[vurusovs]

@VimalWill could you provide more details what the problem you met? Please, take a look at build instructions in https://github.com/openvinotoolkit/openvino/blob/master/docs/dev/build.md

[VimalWill]

@vurusovs I followed the same steps only for building from source in linux machine, but the get the same error as i reported earlier when I try to convert my nncf based int 8 quantized tf model to openvino IR using convert_model()

[andrei-kochin]

@VimalWill do you have any good news on your trials?

[andrei-kochin]

Closing as operation was enabled on master. @VimalWill Feel free to re-open in case of any concerns