[pull] master from tensorflow:master#253

Open

pull[bot] wants to merge 10000 commits into

greedforgood:masterfrom

tensorflow:master

This pull request is big! We're only showing the most recent 250 commits

Commits on Jul 14, 2026

Lower approx_top_k to stable_sort+slice

apivovarov
authored and
tensorflower-gardener
committed

Commits on Jul 15, 2026

[PjRt-IFRT] Use HighwayHash for array hashing

hyeontaek
authored and
tensorflower-gardener
committed
[XLA:GPU] Use RAFT select_k for unstable TopK

apivovarov
authored and
tensorflower-gardener
committed
Implement CommonPjRtClient::Linearize for GPU/CPU.

pschuh
authored and
tensorflower-gardener
committed
Reverts 024bbc8f5be6808148437567deab4b719acfe58d

junwhanahn
authored and
tensorflower-gardener
committed
PR #45475: [xla:gpu] Remove allocation_granularity from GpuExecutableBufferAllocator

ezhulenev
authored and
tensorflower-gardener
committed
Add AutotuneCacheContext in local cache.

shyamliagrawal
authored and
tensorflower-gardener
committed
Automated Code Change
tensorflower-gardener
committed
Update `AddMandatoryConstraints` in `layout_assignment.cc` (`OpenXLA`) to enforce mandatory operand and result layout constraints across `kAsyncStart` and `kAsyncDone` chains.

amitsabne1
authored and
tensorflower-gardener
committed
Fix Col2im 3D version.
wangw-1991
committed
Conditionally include sysmacros.h only on Linux
tensorflower-gardener
committed
Automated Code Change
tensorflower-gardener
committed
Automated Code Change
tensorflower-gardener
committed
[XLA:GPU] Rollback: Refactor DotMerger to use DotOperandDims and ShapeTracker.

mooskagh
authored and
tensorflower-gardener
committed
[XLA:GPU] Refactor GemmFusionSwapOperands to use DotOperandDims and remove Triton emitter compatibility check.

mooskagh
authored and
tensorflower-gardener
committed
Automated Code Change
tensorflower-gardener
committed
PR #45691: [XLA:GPU][oneAPI] Skip PTX kernel tests on oneAPI backend

bhavani-subramanian
authored and
tensorflower-gardener
committed
Account for reduction dimension padding in GPU indexing performance model.

Moerafaat
authored and
tensorflower-gardener
committed
PR #45692: [XLA:GPU[oneAPI] Removed limit on stream creation per device for oneAPI backend

bhavani-subramanian
authored and
tensorflower-gardener
committed
PR #45687: [XLA:GPU][oneAPI] Skip Mosaic GPU tests on oneAPI backend

bhavani-subramanian
authored and
tensorflower-gardener
committed
[XLA:GPU] Add extra filter callback to ScaledDotRewriter on GPU

akuegel
authored and
tensorflower-gardener
committed
[xla:gpu] Improve dot fusion cost model by incorporating hardware launch waves

karupayun
authored and
tensorflower-gardener
committed
PR #45669: [ROCm] Add complex GEMM transpose unit tests

magaonka-amd
authored and
tensorflower-gardener
committed
[XLA:GPU] Disallow convert fusions for devices pre-Ampere (e.g. Volta).

derdrdirk
authored and
tensorflower-gardener
committed
PR #45646: [XLA:GPU] Restrict generic Triton FP4 codegen

and-ivanov
authored and
tensorflower-gardener
committed
Automated Code Change
tensorflower-gardener
committed
Automated Code Change
tensorflower-gardener
committed
[XLA:GPU] Refactor DotMerger to use DotOperandDims and ShapeTracker.

mooskagh
authored and
tensorflower-gardener
committed
Add end-to-end support for FP6 formats (float6_e2m3fn and float6_e3m2fn) across CPU and GPU.

hawkinsp
authored and
tensorflower-gardener
committed
[XLA:GPU] Add roots list to experimental::TiledHloRegion

Marcin Radomski
authored and
tensorflower-gardener
committed
PR #45421: [XLA:SE] Simplify VMM allocator record and pending state

shawnwang18
authored and
tensorflower-gardener
committed
PR #45567: [XLA:GPU] Split command buffer VA remapping into its own allocator class

shawnwang18
authored and
tensorflower-gardener
committed
PR #45422: [XLA:SE] Consolidate VMM allocator mapped allocation paths

shawnwang18
authored and
tensorflower-gardener
committed
XTile: Skip emitting no-op reduce-precision operations.

vwbaker
authored and
tensorflower-gardener
committed
Support reduction_op for unreduced_axes with HloShardingV3.

bixia1
authored and
tensorflower-gardener
committed
[PJRT:GPU] Return UnimplementedError from GetAllocatorStats when memory statistics are not tracked

hawkinsp
authored and
tensorflower-gardener
committed
Add custom handler to flatten selectively for while/conditionals.

ekayaaslan
authored and
tensorflower-gardener
committed
[PJRT] Add UndonatableCommonPjRtBuffer: a hold-free PjRtBuffer for non-donating workloads.

xiaoxlu
authored and
tensorflower-gardener
committed
break amd tests

pineapplejuice233
authored and
tensorflower-gardener
committed
Integrate LLVM at llvm/llvm-project@6e6272088e69

rupprecht
authored and
tensorflower-gardener
committed
Automated Code Change
tensorflower-gardener
committed
[XLA:TPU] DCE should not remove unused outputs that have side effects
tensorflower-gardener
committed
Add boundary checks to pad-like kernels

snnn
authored and
tensorflower-gardener
committed
[PJRT] Rollback UndonatableCommonPjRtBuffer

megrez-yliu
authored and
tensorflower-gardener
committed
Update execution flow for the wheel upload
tensorflower-gardener
committed

Commits on Jul 16, 2026

Commits on Jul 17, 2026

Commits on Jul 18, 2026

address PR review suggestion and fix build issue
gzmkl
committed
Fix host memory leak in tf.data.Dataset.from_generator (GitHub #123269)
Daksh Prajapati
committed
Simplify _internal_py_func graph walk per Gemini review
Daksh Prajapati
committed

Commits on Jul 19, 2026

Fix NaN gradient of tf.math.igamma at a=1, x=0
SEPURI-SAI-KRISHNA
committed

Commits on Jul 20, 2026

per review suggestion, more code change with cudnn impl., and tests
gzmkl
committed

Commits on Jul 21, 2026

Commits on Jul 22, 2026

Commits on Jul 23, 2026

Commits on Jul 24, 2026