[pull] master from tensorflow:master - #8788

Merged

pull[bot] merged 11 commits into

Cache-Cloud:masterfrom

tensorflow:master

Aug 28, 2026

Commits on Aug 28, 2026

Allow ProfilerController to consume data in the kCollectData state.

sannidhyachauhan
authored and
tensorflower-gardener
committed
2c93352
[XLA:GPU] Add DegenerateDimensionRewriter to GPU pipeline.
tensorflower-gardener
committed
3b6d61b
[XLA:GPU] Use symmetric memory and peer address lookup for collective kernel thunks.

PatriosTheGreat
authored and
tensorflower-gardener
committed
9dd68c0
PR #47701: [xla] Extract a separate collective_rendezvous library

ezhulenev
authored and
tensorflower-gardener
committed
8601795
Fix schedule-aware collective CSE to advance candidates when distance threshold is exceeded.

jbspooner
authored and
tensorflower-gardener
committed
5b4ac04
[re-land][xla:cpu] add full msan support

cota
authored and
tensorflower-gardener
committed
e7e4855
Delete GPU Record stack trace as we don't need it anymore.

changm
authored and
tensorflower-gardener
committed
6bbfc2b
[XLA:SchedulingGroups] Do not drop the trivial groups that include ops for which "keep_trivial_sync_annotation" returns true. Delete the redundant code block which only checks groups with a single op.

seherellis
authored and
tensorflower-gardener
committed
1b3c59e
Prevent float overflow by scaling kExp input bounds with downstream reduction sizes.

bhatuzdaname
authored and
tensorflower-gardener
committed
2d48b97
Adapt xla aot test with new autotune cache format.

shyamliagrawal
authored and
tensorflower-gardener
committed
2f66a4e
Generalize PjRtStreamExecutorClient::Load to be a general function on

pschuh
authored and
tensorflower-gardener
committed
ec78c9a