[Autoparallel] Add local_map variant of DSv3 and 2D mesh AP #2129
+313
−2
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
xmfan
added a commit
that referenced
this pull request
Dec 9, 2025
stack-info: PR: #2129, branch: xmfan/stack/7
This was referenced Dec 9, 2025
meta-cla
bot
added
the
CLA Signed
xmfan
requested review from
fegin,
tianyu-l,
wconstab and
wwwjn
as code owners
xmfan
added a commit
that referenced
this pull request
Dec 9, 2025
stack-info: PR: #2129, branch: xmfan/stack/7
xmfan
added a commit
that referenced
this pull request
Dec 9, 2025
stack-info: PR: #2129, branch: xmfan/stack/7
xmfan
added a commit
that referenced
this pull request
Dec 9, 2025
stack-info: PR: #2129, branch: xmfan/stack/7
xmfan
requested review from
allenwang28,
ebsmothers,
joecummings and
pbontrager
as code owners
xmfan
added a commit
that referenced
this pull request
Dec 12, 2025
stack-info: PR: #2129, branch: xmfan/stack/7
xmfan
added a commit
that referenced
this pull request
Dec 12, 2025
stack-info: PR: #2129, branch: xmfan/stack/7
| for layer in model.layers.values(): | ||
| if layer.moe_enabled: | ||
| layer.moe.mesh = world_mesh | ||
| layer.moe.axis_name = "dp_shard_in_ep" |
Contributor
There was a problem hiding this comment.
This seems okay for now, but when we enable TP, this should change so just add a comment that modify this when enabling TP.
Member
Author
There was a problem hiding this comment.
I added an assert to check for dp2ep specifically, because i don't think we're handling the mesh setup well...
| parallel_dims.tp_enabled | ||
| and not job_config.parallelism.disable_loss_parallel | ||
| ) | ||
| assert not loss_parallel_enabled |
Contributor
There was a problem hiding this comment.
Instead of failing here why not disable loss parallel and give a warning?
Member
Author
There was a problem hiding this comment.
warnings are easy to miss, you won't know loss parallel is disable if you don't spot it before titan starts dumping its logs
sanketpurandare
approved these changes
Dec 12, 2025
Contributor
sanketpurandare
left a comment
sanketpurandare
left a comment
There was a problem hiding this comment.
Just some minor comments that can be addressed optionally
stack-info: PR: #2129, branch: xmfan/stack/7
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
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.
Stacked PRs:
[Autoparallel] Add local_map variant of DSv3 and 2D mesh AP
Currently, the AP experiment monkey patches Titan's main DSv3 implementation. But this is prone to breakage from both model definition changes in titan and from HOP/partitioner related changes in core. When these breaks happen, people are usually blocked until I find the root cause.
I'm going on PTO for the rest of the year, so I'm adding an integration to AP's DSv3 model in an attempt to make the development more stable for the upcoming PP integration.
Test: https://gist.github.com/xmfan/db15fda1e1bc1df7cd523005fe0baf33