Free mem trigger with all2all for sync trigger eviction #5058
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
✅ Deploy Preview for pytorch-fbgemm-docs ready!
To edit notification comments on pull requests, go to your Netlify project configuration. |
EddyLXJ
added a commit
to EddyLXJ/torchrec
that referenced
this pull request
Oct 27, 2025
…#3442) Summary: X-link: pytorch/FBGEMM#5058 X-link: facebookresearch/FBGEMM#2067 Before KVZCH is using ID_COUNT and MEM_UTIL eviction trigger mode, both are very tricky and hard for model engineer to decide what num to use for the id count or mem util threshold. Besides that, the eviction start time is out of sync after some time in training, which can cause great qps drop during eviction. This diff is adding support for free memory trigger eviction. It will check how many free memory left every N batch in every rank and if free memory below the threshold, it will trigger eviction in all tbes of all ranks using all reduce. In this way, we can force the start time of eviction in all ranks. Reviewed By: emlin Differential Revision: D83896528
EddyLXJ
added a commit
to EddyLXJ/FBGEMM-1
that referenced
this pull request
Oct 27, 2025
Summary: X-link: facebookresearch/FBGEMM#2067 X-link: meta-pytorch/torchrec#3442 Before KVZCH is using ID_COUNT and MEM_UTIL eviction trigger mode, both are very tricky and hard for model engineer to decide what num to use for the id count or mem util threshold. Besides that, the eviction start time is out of sync after some time in training, which can cause great qps drop during eviction. This diff is adding support for free memory trigger eviction. It will check how many free memory left every N batch in every rank and if free memory below the threshold, it will trigger eviction in all tbes of all ranks using all reduce. In this way, we can force the start time of eviction in all ranks. Reviewed By: emlin Differential Revision: D83896528
EddyLXJ
added a commit
to EddyLXJ/torchrec
that referenced
this pull request
Oct 27, 2025
…#3442) Summary: X-link: pytorch/FBGEMM#5058 X-link: facebookresearch/FBGEMM#2067 Before KVZCH is using ID_COUNT and MEM_UTIL eviction trigger mode, both are very tricky and hard for model engineer to decide what num to use for the id count or mem util threshold. Besides that, the eviction start time is out of sync after some time in training, which can cause great qps drop during eviction. This diff is adding support for free memory trigger eviction. It will check how many free memory left every N batch in every rank and if free memory below the threshold, it will trigger eviction in all tbes of all ranks using all reduce. In this way, we can force the start time of eviction in all ranks. Reviewed By: emlin Differential Revision: D83896528
EddyLXJ
added a commit
to EddyLXJ/FBGEMM-1
that referenced
this pull request
Oct 27, 2025
Summary: X-link: facebookresearch/FBGEMM#2067 X-link: meta-pytorch/torchrec#3442 Before KVZCH is using ID_COUNT and MEM_UTIL eviction trigger mode, both are very tricky and hard for model engineer to decide what num to use for the id count or mem util threshold. Besides that, the eviction start time is out of sync after some time in training, which can cause great qps drop during eviction. This diff is adding support for free memory trigger eviction. It will check how many free memory left every N batch in every rank and if free memory below the threshold, it will trigger eviction in all tbes of all ranks using all reduce. In this way, we can force the start time of eviction in all ranks. Reviewed By: emlin Differential Revision: D83896528
EddyLXJ
force-pushed
the
export-D83896528
branch
from
e193fa6 to
a1b5ddd
Compare
EddyLXJ
added a commit
to EddyLXJ/FBGEMM-1
that referenced
this pull request
Oct 27, 2025
Summary: X-link: facebookresearch/FBGEMM#2067 X-link: meta-pytorch/torchrec#3442 Before KVZCH is using ID_COUNT and MEM_UTIL eviction trigger mode, both are very tricky and hard for model engineer to decide what num to use for the id count or mem util threshold. Besides that, the eviction start time is out of sync after some time in training, which can cause great qps drop during eviction. This diff is adding support for free memory trigger eviction. It will check how many free memory left every N batch in every rank and if free memory below the threshold, it will trigger eviction in all tbes of all ranks using all reduce. In this way, we can force the start time of eviction in all ranks. Differential Revision: D83896528
EddyLXJ
added a commit
to EddyLXJ/FBGEMM-1
that referenced
this pull request
Oct 27, 2025
Summary: X-link: facebookresearch/FBGEMM#2067 X-link: meta-pytorch/torchrec#3442 Before KVZCH is using ID_COUNT and MEM_UTIL eviction trigger mode, both are very tricky and hard for model engineer to decide what num to use for the id count or mem util threshold. Besides that, the eviction start time is out of sync after some time in training, which can cause great qps drop during eviction. This diff is adding support for free memory trigger eviction. It will check how many free memory left every N batch in every rank and if free memory below the threshold, it will trigger eviction in all tbes of all ranks using all reduce. In this way, we can force the start time of eviction in all ranks. Differential Revision: D83896528
EddyLXJ
force-pushed
the
export-D83896528
branch
from
a1b5ddd to
8ed19d4
Compare
EddyLXJ
added a commit
to EddyLXJ/FBGEMM-1
that referenced
this pull request
Oct 27, 2025
Summary: X-link: facebookresearch/FBGEMM#2067 X-link: meta-pytorch/torchrec#3442 Before KVZCH is using ID_COUNT and MEM_UTIL eviction trigger mode, both are very tricky and hard for model engineer to decide what num to use for the id count or mem util threshold. Besides that, the eviction start time is out of sync after some time in training, which can cause great qps drop during eviction. This diff is adding support for free memory trigger eviction. It will check how many free memory left every N batch in every rank and if free memory below the threshold, it will trigger eviction in all tbes of all ranks using all reduce. In this way, we can force the start time of eviction in all ranks. Differential Revision: D83896528
Summary: X-link: facebookresearch/FBGEMM#2067 See diff D85604160, this KVZCHEvictionTBEConfig is in FBGEMM and used in torchrec. Both FBGEEM and torchrec are open source in github. It is required to land first, otherwise torchrec github build will throw error {F1983027645} Reviewed By: emlin Differential Revision: D83896528
EddyLXJ
force-pushed
the
export-D83896528
branch
from
8ed19d4 to
8b9301e
Compare
|
This pull request has been merged in 0267804. |
Bernard-Liu
pushed a commit
to ROCm/FBGEMM
that referenced
this pull request
Oct 31, 2025
Summary: Pull Request resolved: pytorch#5058 X-link: https://github.com/facebookresearch/FBGEMM/pull/2067 See diff D85604160, this KVZCHEvictionTBEConfig is in FBGEMM and used in torchrec. Both FBGEEM and torchrec are open source in github. It is required to land first, otherwise torchrec github build will throw error {F1983027645} Reviewed By: emlin Differential Revision: D83896528 fbshipit-source-id: 7a8bacc3d0ee1f53a797dac6ba2647d372a15074
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Summary:
X-link: https://github.com/facebookresearch/FBGEMM/pull/2067
X-link: meta-pytorch/torchrec#3442
Before KVZCH is using ID_COUNT and MEM_UTIL eviction trigger mode, both are very tricky and hard for model engineer to decide what num to use for the id count or mem util threshold. Besides that, the eviction start time is out of sync after some time in training, which can cause great qps drop during eviction.
This diff is adding support for free memory trigger eviction. It will check how many free memory left every N batch in every rank and if free memory below the threshold, it will trigger eviction in all tbes of all ranks using all reduce. In this way, we can force the start time of eviction in all ranks.
Reviewed By: emlin
Differential Revision: D83896528