cee1 commented on PR #11155:
URL: https://github.com/apache/tvm/pull/11155#issuecomment-1118125802
> Why do we think the DEFAULT path is "accurate" here ?
DEFAULT path is (somewhat) consistent with QAT, specifically, the rounding
behavior within fake quantization.
For DEFAULT path, only one right shift is required, and the (only one)
rounding happens there.
For NEON path, the (possible) extra rounding behavior is introduced by NEON
instruction `sqrdmulh`, which has some side-effect (see
https://discuss.tvm.apache.org/t/quantization-aligning-result-of-tvm-to-torchs/12225/2?u=cee1
)
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
