Questions about int4 quantization

[gaikwadrahul8]

Hi, all.
Since it is an inquiry rather than an issue, I will not write a template.

Looking at the kernel side code of tflite, I saw that int4 for filter is supported in several op kernels; conv2d, depthwise-conv2d, fully-connected.

Could you tell me if there are plans to support int4 quantization in the tflite converter or support int4 for each op's input as well as filter, and if so, what milestones do you have?

Thank you :)

[gaikwadrahul8]

This issue originally reported by @0-chan-kor has been moved to this dedicated repository for LiteRT to enhance issue tracking and prioritization. To ensure continuity, we have created this new issue on your behalf.

We appreciate your understanding and look forward to your continued involvement.

[pkgoogle]

Original Issue: tensorflow/tensorflow#60125