Tian, Feng

I have made corresponding changes in local to pass test. your PR will be merged to next release. Thanks

you need update your machine to XEON server for faster speed.

> @ftian1 if accelerator other than CUDA want to support AutoTP WOQ, which set of OpBuilder/kernels needs to be implemented? Can you provide a link to kernel usage in the...

PS: after manually switch pydantic version from 2.0a2 to 1.10.7. the latest code works. I have a quick glance on that, it's because pydantic 2.0a2 has removed \_\_field\_\_ attributes in...

@microsoft-github-policy-service agree company="Intel"

@xiaoxiawu-microsoft sorry for the late response due to PRC holiday and thanks for your review. I have fixed the yapf scan issue. but in my local, the detection of destroyed...

@xiaoxiawu-microsoft Those pre-ci errors are not related with my changes, could you pls have a check?

@yaozhewei thanks for the valuable feedback. we are evaluating if we can remove or enhance callback like you suggested. will get back to you soon. as for post-training quantization support,...

@yaozhewei per our investigation, we found it's doable to remove those explicit callbacks. we are preparing PR for further review. will ping you when it's ready. as for calibration based...

>> For quantization proposal, the post-training quantization is in some sense already implemented. Users can use static activation quantization method with few batches inference to get the calibration. See here...