In many cloud services provided by cloud vendors, many people are still using TensorFlow 1.15 for many reasons, and they do not upgrade TensorFlow versions easily for service stability.
However, for many operators, such as Intel MKL operator, the latest version update is in TensorFlow version 2.x instead of TensorFlow 1.15, so if you want to use the latest version of the operator in TensorFlow 1.15, you need to solve the operator replacement and compilation problems.
I would like to provide a compiled example (TensorFlow 1.15 / TensorFlow 2.6.0 mixed compilation) and detailed step-by-step documentation to developers who encounter such problems, to help them use the latest Intel MKL operators for better inference performance on CPU while keeping TensorFlow 1.15 version unchanged.
So I want to push the code and documentation to the official community to help people with this need, how and where is it appropriate to post it?
Thank you!