Renato Barros Arantes

Bug: Missing function register_module_for_export

4

### What happened? I'm having an issue with TF v2.16 and [tensorflow_hub](https://github.com/tensorflow/hub/releases) v0.16, which is throwing an error about a missing function `register_module_for_export`. If I downgrade to tensorflow_hub v0.15, it...

test_shuffle fails on aarch64 when BF16 data type is enabled.

3

# Summary By enabling BF16 data type on aarch64 [here](https://github.com/renato-arantes/oneDNN/blob/2136e1597be32bc56cd8244ec282a7908d8c9a9f/src/cpu/platform.cpp#L121), `test_shuffle` fails [here](https://github.com/oneapi-src/oneDNN/blob/561d996658ea151cbbf38385a837615e75662939/src/cpu/aarch64/shuffle/jit_uni_shuffle_kernel.cpp#L71). # Version main, commit 0c2ea8b by @kawakami-k # Steps to reproduce The code [here](https://github.com/renato-arantes/oneDNN/blob/2136e1597be32bc56cd8244ec282a7908d8c9a9f/src/cpu/platform.cpp#L121) checks if [ACL](https://github.com/ARM-software/ComputeLibrary)...

src:cpu:aarch64 restricts the conv operation to just one post-op when…

8

# Description When using fp16 on AArch64 there are multiple failed conv tests due to a precision problem in oneDNN's fp16 operations. OneDNN conducts all fp16 post-op operations in fp32...

Pytorch quantization bias is not quantised on aarch64

4

Hi all, I created a simple Pytorch image classification example that correctly classifies a sample image with `76%` accuracy. I then applied static quantization to the model and it continued...

Enables static quantization for aarch64

2

This pull request enables static quantization for aarch64, supporting convolution in signed 8-bit (s8s8s8) and unsigned 8-bit (u8s8u8) formats. cc @jgong5 @mingfeima @XiaobingSuper @sanchitintel @ashokei @jingxu10