Rebase Huawei serving commits onto new-serving-2.20#51

Closed

Copilot wants to merge 7663 commits intonew-serving-2.20from

copilot/squash-commits-for-ease-of-review

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

Commits on Nov 28, 2025

[xla:gpu] Fix clangtidy warnings (except `misc-include-cleaner`).

chsigg
authored and
tensorflower-gardener
committed
Make tf_profiler_pybind_cc_library_wrapper generate layering check compliant targets

beckerhe
authored and
tensorflower-gardener
committed
[XLA:GPU] Fix usages of deprecated tsl::errors::Unimplemented in xla

loislo
authored and
tensorflower-gardener
committed
Re-enable CollectiveOpsTestE2EShardedUnsharded test.
tensorflower-gardener
committed
[XLA:GPU] Streamline the order of operations in `EmitCollectiveThunk`.

olegshyshkov
authored and
tensorflower-gardener
committed
[XLA][codegen] Move emitter helpers and dot algorithms to xtile namespace

basioli-k
authored and
tensorflower-gardener
committed
Fix output bytes for cuBLAS custom calls in GPU HLO Cost Model.
tensorflower-gardener
committed
[XLA:GPU] Fix usages of deprecated tsl::errors::Internal in xla

loislo
authored and
tensorflower-gardener
committed
Delete unused `ReshapeToTileDimension`, `ContainsTileSharding` functions.

KanishAnand
authored and
tensorflower-gardener
committed
[XLA:GPU] Add ExecuteReplicated override that returns executable, optimized module and literals.

olegshyshkov
authored and
tensorflower-gardener
committed
[XLA:GPU] call AdjustDebugOptionsForAutotuning from autotuner util

metaflow
authored and
tensorflower-gardener
committed
[XLA:GPU] Fix usages of deprecated tsl::errors::OutOfRange in xla

loislo
authored and
tensorflower-gardener
committed
Verify that input arrays are on the expected devices before execution

junwhanahn
authored and
tensorflower-gardener
committed
[xla:gpu] Remove redundant Thunk::GetGpuCollectives

ezhulenev
authored and
tensorflower-gardener
committed
Fix ExactInterpolatorKey equality and enable sol_latency_estimator_test.

chsigg
authored and
tensorflower-gardener
committed
[XLA:GPU] Fix usages of deprecated tsl::errors::PermissionDenied in xla

loislo
authored and
tensorflower-gardener
committed
[XLA:GPU] Fix usages of deprecated tsl::errors::DataLoss and tsl::errors::InvalidArgument in xla

loislo
authored and
tensorflower-gardener
committed

Commits on Nov 29, 2025

Commits on Nov 30, 2025

Commits on Dec 1, 2025

Commits on Dec 2, 2025

Commits on Dec 3, 2025

Introduce a variant of `ToProto` that takes the proto as an output parameter

junwhanahn
authored and
tensorflower-gardener
committed
This change replaces uses of std::iota with absl::c_iota, which operate directly on containers. This simplifies the code by removing the need to pass begin() and end() iterators.

apivovarov
authored and
tensorflower-gardener
committed
Avoid copying references in `AbstractTrackedDeviceBuffer::raw_buffer()` on read-only access

junwhanahn
authored and
tensorflower-gardener
committed
Add an IsEmpty() API to AttributeMap

krishnaharidasan
authored and
tensorflower-gardener
committed
Port more BufferUse users to provide shape

ermilovmaxim
authored and
tensorflower-gardener
committed
Add a variant of `Shape:ToProto`/`Layout::ToProto` that takes a proto to be serialized into

junwhanahn
authored and
tensorflower-gardener
committed
Update `rules_ml_toolchain` to disable `--@cuda_driver//:include_cuda_umd_libs` by default.

ybaturina
authored and
tensorflower-gardener
committed
Automated Code Change
tensorflower-gardener
committed
[xla:collectives] Move ReductionKind to a separate target to break dependency on HLO IR

ezhulenev
authored and
tensorflower-gardener
committed
Remove unused `GetMostOccurringDevice` function.

ZixuanJiang
authored and
tensorflower-gardener
committed
Automated Code Change
tensorflower-gardener
committed
compat: Update forward compatibility horizon to 2025-12-03
tensorflower-gardener
committed
Update GraphDef version to 2430.
tensorflower-gardener
committed
Add "hermetic_cuda_umd" to TF build config in XLA CI

beckerhe
authored and
tensorflower-gardener
committed
[xla:ffi] Do not depend on GPU backend from default FFI implementation

ezhulenev
authored and
tensorflower-gardener
committed
[XLA:GPU] Add support for S32 keys for SortPairs.

akuegel
authored and
tensorflower-gardener
committed
[JAX] Split _jax extension into _xla and _jax extensions.

hawkinsp
authored and
tensorflower-gardener
committed
[XLA] Remove TF_ prefix from TF_{ASSERT,EXPECT}_OK

Marcin Radomski
authored and
tensorflower-gardener
committed
[xla:cpu] Remove unused runtime_single_threaded_matmul for f32

ezhulenev
authored and
tensorflower-gardener
committed
Avoid zero-initializing memory allocated for sub-byte packing

junwhanahn
authored and
tensorflower-gardener
committed
[xla:ffi] Remove internal context decoding code duplication

ezhulenev
authored and
tensorflower-gardener
committed
PR #32378: [XLA:CPU][oneDNN][BugFix] Fix unsafe subtraction from unsigned values

akhilgoe
authored and
tensorflower-gardener
committed
Add serialization and deserialization of IFRT IR MPMD executables.
tensorflower-gardener
committed
[XLA:GPU] Disable debug thunk passes in autotuning

Marcin Radomski
authored and
tensorflower-gardener
committed
Reverts daaf047d90ab7fc50e6890a961e22efbdb517307

hawkinsp
authored and
tensorflower-gardener
committed
Don't call HeapSummary() by default on windows as it's expensive

protobird-git
authored and
tensorflower-gardener
committed
Add session_id support to ProfilerSessionWrapper and ExportToTensorBoard.

subhamsoni-google
authored and
tensorflower-gardener
committed
Added buildable copy of coordination service.

mwhittaker
authored and
tensorflower-gardener
committed
Internal cleanup. Nothing to see here.
tensorflower-gardener
committed
Update hermetic CUDA UMD version to match CUDA driver version installed on RBE machines.

ybaturina
authored and
tensorflower-gardener
committed
[xla:ffi] Relax the check for number of args and rets if handler doesn't bind any of them

ezhulenev
authored and
tensorflower-gardener
committed
#HLODiff Add comment explaining HloInstructionNodeProps::subgraph_fingerprint.
tensorflower-gardener
committed
[XLA:GPU][XLA:CPU] Add simplifications for HloOpcode::kConditional.

thcmbs
authored and
tensorflower-gardener
committed
[XLA] Add an option to control whether the collective pipeliner uniquifies channel IDs.

mkuperst
authored and
tensorflower-gardener
committed
#HLODiff Add option to skip precomputing instruction dependencies in hlo diff.
tensorflower-gardener
committed
[XLA:GPU] Do not create an additional LLVM module for Triton.

pifon2a
authored and
tensorflower-gardener
committed
override Thunk::buffer_uses where needed. Part 2

ermilovmaxim
authored and
tensorflower-gardener
committed
[tsl]Migrate from std::algorithms to absl::c_

apivovarov
authored and
tensorflower-gardener
committed

Commits on Dec 4, 2025

Commits on Mar 20, 2026