Nadav Elyahu

@loadams Sure, thanks :)

Halting this PR till habana version 1.16.0 will be released. CI issue should be fixed there.

> > Halting this PR till habana version 1.16.0 will be released. CI issue should be fixed there. > > Thanks @nelyahu - when it is released, go ahead and...

Hi @jeffra :) can you please help reviewing this change?

Tagging original author of the PR @Yejing-Lai CC: @tjruwase

Hi @tjruwase, PR https://github.com/microsoft/DeepSpeed/pull/4697 was merged while back, without concrete performance motivation. In addition it increases in un-even balancing of weight sharding, as it enforces blocks of 64 elements, under...

Hi @lekurile , can you please run ds-chat coverage on this PR? i reproduced the issue was reproted in https://github.com/nelyahu/DeepSpeed/commit/bc48371c5e1fb8fd70fc79285e66201dbb65679b and fixed it. would like to get a pre-commit validation...

> > Hi @lekurile , can you please run ds-chat coverage on this PR? i reproduced the issue was reproted in [nelyahu@bc48371](https://github.com/nelyahu/DeepSpeed/commit/bc48371c5e1fb8fd70fc79285e66201dbb65679b) and fixed it. would like to get a...

@loadams @mrwyattii Can you help taking this PR further? @deepcharm handled all review comments.

@loadams @mrwyattii @lekurile can this change be merged?