[pull] master from tensorflow:master#2374

Open

pull[bot] wants to merge 10000 commits into

xiaohanhuang:masterfrom

tensorflow:master

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

Commits on Jun 4, 2026

This is an automatic update to a device compatibility allowlist.
tensorflower-gardener
committed
Automated Code Change
tensorflower-gardener
committed
Fix type confusion vulnerability in TfLiteSettingsJsonParser by enforcing root_type and verifying flatbuffer structure before accessing root.

dmiltr3
authored and
tensorflower-gardener
committed
[XLA:GPU] Make BufferView public and add intersection/packing capabilities.

mooskagh
authored and
tensorflower-gardener
committed
PR #43402: [XLA:GPU][oneAPI] Skip block_scaled_dot tests in collectives e2e test suite

akhilgoe
authored and
tensorflower-gardener
committed
[XLA:GPU] Add ShapeTracker::Narrow()

mooskagh
authored and
tensorflower-gardener
committed
[XLA:GPU] Respect PrecisionConfig::HIGHEST in DotAlgorithmRewriter and xtile codegen.

loislo
authored and
tensorflower-gardener
committed
[XLA:GPU] Add ShapeTracker::Zip method.

mooskagh
authored and
tensorflower-gardener
committed
Validate bin index in HistogramFixedWidth op.

ratgr
authored and
tensorflower-gardener
committed
PR #43122: [xla:gpu] Add support for collective group launch on multiple comms

ezhulenev
authored and
tensorflower-gardener
committed
Fix uninitialized stack memory disclosure and crashes in generated C++ Eager Ops.

patnotz
authored and
tensorflower-gardener
committed
Use ValidateCreateOptions in PJRT_Client_Create for CPU.

mkuperst
authored and
tensorflower-gardener
committed
[XLA:CPU] Compile and select JIT Eigen helper library with multiple vector widths

seantalts
authored and
tensorflower-gardener
committed
[xla] Avoid a divide-by-zero bug in the XLA shape partitioner in the case of scalar tensors.
tensorflower-gardener
committed
Allow mixed precision verification in ApplyAxlTransforms pass.
tensorflower-gardener
committed
Use input/output operand for loop counter in 4-bit fully connected kernels.

lrdxgm
authored and
tensorflower-gardener
committed
Remove unused MKL build configurations from bazelrc files

yuriivcs
authored and
tensorflower-gardener
committed
Add validation check ensuring allocated content size matches expected shape dimensions in SVDF const node merging.

dmiltr3
authored and
tensorflower-gardener
committed
Add explanation about clobber registers v25-v31 to 4-bit NEON fully connected kernels on AArch64

dmiltr3
authored and
tensorflower-gardener
committed
Replace backend specific versions of PjRtDeviceEventSet with a version based on

pschuh
authored and
tensorflower-gardener
committed
Fix out-of-bounds read and write in TOCO CopyArrayData by adding bounds checks.

dmiltr3
authored and
tensorflower-gardener
committed
Add constant folder for TFLite LogisticOp.
tensorflower-gardener
committed
Rollback of device compatibility allowlist update to unblock downstream test failures.
tensorflower-gardener
committed
Verify that Resource and Variant tensors are not constant or raw user inputs before dereferencing their internal buffer pointer.

dmiltr3
authored and
tensorflower-gardener
committed
Explicitly deprecate the Java NNAPI Delegate.

fergushenderson
authored and
tensorflower-gardener
committed
[XLA] Serialize the details of IotaTileAssignment in MeshProto
tensorflower-gardener
committed
Remove unnecessary `volatile`.

lrdxgm
authored and
tensorflower-gardener
committed
Fix heap buffer overflow by correctly allocating buffer size and unpacking kTfLiteInt2 tensors in GetTensor.

dmiltr3
authored and
tensorflower-gardener
committed
Fix out-of-bounds access vulnerabilities in signature_def_util.cc by adding bounds checks.

sgarciagoogle
authored and
tensorflower-gardener
committed
Rename `tsl::RCReference<xla::PjRtDeviceEventPromise>` to `xla::PjRtDeviceEventPromiseRef` to prepare for adding a c-api for promises.

pschuh
authored and
tensorflower-gardener
committed
Support bf16 `TRANSPOSE` and `DYNAMIC_UPDATE_SLICE`.
tensorflower-gardener
committed
Add PjRtStagingBuffer to allow unifying how pjrt-client implementations handle

pschuh
authored and
tensorflower-gardener
committed
PR #43764: [xla:gpu] When rewriting for dynamic-memcpy check that all slicing instructions are compatible

ezhulenev
authored and
tensorflower-gardener
committed

Commits on Jun 5, 2026

Commits on Jun 6, 2026

Commits on Jun 7, 2026

Automated Code Change
tensorflower-gardener
committed
[XLA:GPU] gpu performance timings

metaflow
authored and
tensorflower-gardener
committed
fix accidentally removed code in tpu_initialize_util

Peter Gavin
authored and
tensorflower-gardener
committed
Enables the optimization to fold trailing transposes into scatter operations for the GPU backend.
tensorflower-gardener
committed

Commits on Jun 8, 2026

add fp16 support for TFL_AddOp

sirakiin
authored and
tensorflower-gardener
committed
add fp16 support for TFL_MulOp

sirakiin
authored and
tensorflower-gardener
committed
Wrap OpenGL interop functions with CL_DELEGATE_NO_GL in TFLite GPU delegate
tensorflower-gardener
committed
add fp16 support for TFL_SinOp

sirakiin
authored and
tensorflower-gardener
committed
add fp16 support for TFL_CosOp

sirakiin
authored and
tensorflower-gardener
committed
add fp16 support for TFL_ConcatenationOp

sirakiin
authored and
tensorflower-gardener
committed
Automated Code Change
tensorflower-gardener
committed
Use BufferAllocations when loading legacy AOT result

EusebioDM
authored and
tensorflower-gardener
committed
Automated Code Change
tensorflower-gardener
committed
PR #43822: [xla:cuda] Delete legacy cuda device allocator

ezhulenev
authored and
tensorflower-gardener
committed
PR #43878: [ROCm] Add bes keywords for xla PR checks

alekstheod
authored and
tensorflower-gardener
committed
Pre compute Buffer Assignment summary

EusebioDM
authored and
tensorflower-gardener
committed
Automated Code Change
tensorflower-gardener
committed
[XLA:GPU] Skip reshape contiguity check if no concrete tile sizes assigned yet

vickyliu-go4it
authored and
tensorflower-gardener
committed
Use correct integer type for stride.

mkuperst
authored and
tensorflower-gardener
committed
Fix `ConvertUnreducedSharding` to set sharding in case of tuple named shardings

KanishAnand
authored and
tensorflower-gardener
committed
Remove `use_named_sharding` args from HloSharding::Replicate and SingleDevice.

KanishAnand
authored and
tensorflower-gardener
committed
Reverts 3bf2b8004374b4fcf22e4bb55cebfc07a5d73733

EusebioDM
authored and
tensorflower-gardener
committed
Add support for passing additional flags to ptxas.

beckerhe
authored and
tensorflower-gardener
committed
Unit Test for BufferAssignment Proto

Becker-ZH
authored and
tensorflower-gardener
committed
implement DeferRelocatableCompilationCompilationProvider GetLatestPtxIsaVersion

ermilovmaxim
authored and
tensorflower-gardener
committed
Fix integer overflow in MMAPAllocation bounds check, address C++ 64-bit portability/style review findings, and preserve class validity invariant.

dmiltr3
authored and
tensorflower-gardener
committed
Allow the option to not update the layout of the entry computation
tensorflower-gardener
committed
Avoid constant's content serialization if serializing HloModule

ermilovmaxim
authored and
tensorflower-gardener
committed
Fix integer truncation in TfLite GatherStrings.

sgarciagoogle
authored and
tensorflower-gardener
committed
add fp16 support for TFL_SliceOp

sirakiin
authored and
tensorflower-gardener
committed
Add "Interpreting TPU States" section to megascale debugging doc.

bbadawi1
authored and
tensorflower-gardener
committed
Allow high-rank SelectV2 broadcasts
tensorflower-gardener
committed
Update Softmax kernels to handle large dimensions.

ratgr
authored and
tensorflower-gardener
committed
Improve robustness of protocol buffer handling in StreamExecutor TPU utilities by gracefully supporting empty payloads.

ratgr
authored and
tensorflower-gardener
committed
PR #43729: [xla:gpu] Split command buffer warmup from initialization and add warmup for custom_call thunk

shawnwang18
authored and
tensorflower-gardener
committed
GetFusionEmitter does not need mlir_context arg

ermilovmaxim
authored and
tensorflower-gardener
committed
PR #43897: [ROCm] Don't use async deallocaton in MIOpen autotuner backend

draganmladjenovic
authored and
tensorflower-gardener
committed

Commits on Jun 9, 2026

Squashing commits after sync tool breakage

Sumu004
authored and
tensorflower-gardener
committed
Reverts changelist 929108895

gcarranza-1
authored and
tensorflower-gardener
committed
Fix dot simplification check under `HloShardingV3` for unreduced sharding.

KanishAnand
authored and
tensorflower-gardener
committed
Export `__new__` in TensorFlow API goldens for subclasses of `tuple`.

QEDady
authored and
tensorflower-gardener
committed
PR #7925: Support get size of tflite model larger than 2GB.

taste1981
authored and
tensorflower-gardener
committed
[XLA:GPU] use experimental tiling in gemm_fusion pass

metaflow
authored and
tensorflower-gardener
committed
[XLA:TPU] InstructionFusion - split ShouldFuseIntoMultiOutput

georgepaw
authored and
tensorflower-gardener
committed
Refactor: Use `set_sharding_may_convert_to_v2` in dot_handler.cc

Kingston Mandisodza
authored and
tensorflower-gardener
committed
Add conditional compilation for GL interop in OpenCL delegate.

gcarranza-1
authored and
tensorflower-gardener
committed
[XLA:CPU] Bump OSS Eigen version to ea13a98decd497a8c5588fb5de71b57bcf10d864

seantalts
authored and
tensorflower-gardener
committed
Preserve host memory space when automatic layout assignment is used

kho
authored and
tensorflower-gardener
committed
Add `experimental_disable_delegate_node_fusion` option to interpreter

marialyu
authored and
tensorflower-gardener
committed

Commits on Jun 10, 2026

Squashing commits after sync tool breakage

zoranjovanovic-ns
authored and
tensorflower-gardener
committed
PR #42306: Avoid normalizing batched gathers with overlapping batching dims and start_index_map

zeph1912
authored and
tensorflower-gardener
committed
Automated Code Change
tensorflower-gardener
committed
[XLA:GPU] explicitly set sofmax_rewriter parameters

metaflow
authored and
tensorflower-gardener
committed
Fix non-determinism in de-duped backend configs

EusebioDM
authored and
tensorflower-gardener
committed
PR #43478: [XLA:GPU] DeviceAddressVmmAllocator supports reservation-backed VMM allocations

shawnwang18
authored and
tensorflower-gardener
committed
[XLA] Intercept metadata payload from frontend_attributes

thcmbs
authored and
tensorflower-gardener
committed
Remove pass to lower chlo.scan to stablehlo.reduce_window.

chsigg
authored and
tensorflower-gardener
committed
Fix out-of-bounds read in TFLite GPU unaligned concat shader generation.

dmiltr3
authored and
tensorflower-gardener
committed
[XLA:GPU] Remove dynamic dimension handling from matmul_indexing_utils.

mooskagh
authored and
tensorflower-gardener
committed

Commits on Jun 12, 2026

Commits on Jun 13, 2026

Commits on Jun 15, 2026