We are glad to share a new
RFC for adding quantization support for CPU in OpenXLA. The RFC proposes the leverage of Intel® Neural Compressor (
INC) tool to generate Tensorflow quantized model and illustrates how it would be executed by the XLA backend.
We are planning to discuss this RFC in the upcoming OpenXLA community
meeting on Tuesday, July 16th. We look forward for the valuable feedback from the community on this proposal.