Skip to content

Pull requests: openxla/shardy

New pull request New
34 Open 816 Closed
34 Open 816 Closed
Author
Filter by author
Loading
Label
Filter by label
Loading
Use alt + click/return to exclude labels
or ⇧ + click/return for logical OR
Projects
Filter by project
Loading
Milestones
Filter by milestone
Loading
Reviews
Filter by reviews
No reviews Review required Approved review Changes requested
Assignee
Filter by who’s assigned
Assigned to nobody Loading
Sort
Sort by
Newest Oldest Most commented Least commented Recently updated Least recently updated Best match
Most reactions
👍 👎 😄 🎉 😕 ❤️ 🚀 👀

Pull requests list

[mpmd] Use attributes to persist MPMD control dependencies across passes
#862 opened Oct 30, 2025 by copybara-service bot Loading…
test on shard along reduction axes: tpu enable shardy partitioner.
#854 opened Oct 20, 2025 by copybara-service bot Loading…
test: tpu enable shardy partitioner.
#852 opened Oct 17, 2025 by copybara-service bot Loading…
Add sharding rule for stablehlo.batch_norm_inference op
#847 opened Oct 15, 2025 by sshonTT Loading…
3
Shard along reduction axes only when the output tensor sizes are smaller than the source input tensor sizes, to reflect the trade off between all-reduce and all-gather.
#842 opened Oct 14, 2025 by copybara-service bot Loading…
test: dedupFunctionsFully enabled.
#839 opened Oct 13, 2025 by copybara-service bot Loading…
Support unpinned host as a type of on host
#827 opened Oct 8, 2025 by copybara-service bot Loading…
Update schedule in op_priority_propagation.cc.
#814 opened Oct 4, 2025 by copybara-service bot Loading…
@ZixuanJiang
[mpmd] Add custom pipeline API for user-defined scheduling and merging
#798 opened Sep 30, 2025 by copybara-service bot Loading…
[mpmd] Use stage id instead of mesh id in pipeline_schedule.cc
#797 opened Sep 30, 2025 by copybara-service bot Loading…
#sdy never hard fail in getSharding, only in setSharding.
#726 opened Sep 9, 2025 by copybara-service bot Loading…
@tomnatan30
Apply add/remove has_side_effect in MPMD pipelines
#708 opened Sep 4, 2025 by copybara-service bot Loading…
Avoid CSE on stablehlo.custom_call marked with no_cse attribute
#707 opened Sep 4, 2025 by copybara-service bot Loading…
[mpmd] Use global mesh.
#700 opened Sep 2, 2025 by copybara-service bot Loading…
Verify that the shrdings inside the sdy.manual_computation shares the same mesh.
#624 opened Jul 23, 2025 by copybara-service bot Loading…
@ZixuanJiang
Add numerics check for unreduced e2e.
#571 opened Jun 6, 2025 by copybara-service bot Loading…
#sdy Mark unreduced ops with frontend attributes.
#567 opened Jun 5, 2025 by copybara-service bot Loading…
#sdy Add all-to-all fusion canonicalization pattern.
#516 opened May 7, 2025 by copybara-service bot Loading…
Reverts 417dcc0d08153de3d533390db5258035e5869b54
#510 opened May 7, 2025 by copybara-service bot Loading…
@beckerhe
Test commit to ensure I haven't broken the config
#427 opened Mar 19, 2025 by copybara-service bot Loading…
@ddunl
#sdy add temporary pass in XLA SDY import to not allow meshes with different axis shapes.
#415 opened Mar 11, 2025 by copybara-service bot Loading…
@bartchr808
Don't allow sideways propagation between operands/results when propagation direction is forward/backwards.
#410 opened Mar 6, 2025 by copybara-service bot Loading…
@tomnatan30
Create dump directory if not exist.
#372 opened Feb 7, 2025 by yliu120 Loading…
Replace std::bind with std::bind_front for PropagationDirectionAlongFactor.
#363 opened Feb 5, 2025 by copybara-service bot Loading…
@ZixuanJiang
#sdy add sharding rules for ragged_all_to_all custom call
#338 opened Jan 23, 2025 by copybara-service bot Loading…
@ghpvnist
ProTip! Follow long discussions with comments:>50.