Li Li
Li Li
> Hi @liligwu Thanks for your update. If we could use `4 * kWarpSize` (see above), that would be easier to maintain this parameter for both AMD and NVIDIA GPUs....
> > > Hi @liligwu Thanks for your update. If we could use `4 * kWarpSize` (see above), that would be easier to maintain this parameter for both AMD and...
> Thanks @liligwu for your investigation! We can revisit this issue on refactoring. Would you mind rebasing this PR and making sure all tests pass? I'll update the branch soon....
Hi @shintaro-iwasaki , the branch updated. Tests passed on my side locally. You can wait for the CI. Thank you!
> @liligwu Merged. Thanks for your contribution! @shintaro-iwasaki , I appreciate your help.
Mixed-dimension embeddingBag tables are enabled in the most recent commit. "test_cache_pipeline" is re-enabled accordingly.
> @liligwu has you resolved your issue? Yes, thanks @thakkarV's help.
@jeffra, would you please look at this issue? Thank you.
Hi @loadams , It has been a while. Please give me some time to confirm if the issue persists.
> Thanks for the contributions! It'd be nice if you could add explanation a little bit :) > > Judging from the changes, ROCm 5.7 has headers at both old...