Develop upstream sync 250427#2939

Open

linchen1-robot wants to merge 1344 commits intodevelop-upstreamfrom

develop-upstream-sync-250427

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

Commits on Apr 22, 2025

Commits on Apr 23, 2025

Commits on Apr 24, 2025

No need to really create device buffers for an error PjRt Buffer.
tensorflower-gardener
committed
#sdy Support multiple src/tgt pairs in sdy_all_to_all
tensorflower-gardener
committed
Add debugging logs in tfrt_gpu_client.cc
tensorflower-gardener
committed
Fix bug where `DeleteDimension()` doesn't adjust dimension attributes for dense layouts.
tensorflower-gardener
committed
create a default GpuExecutableRunOptions for TfrtGpuClient
tensorflower-gardener
committed
Implement SubByteTransfer in TfrtGpuClient

hhb
authored and
tensorflower-gardener
committed
Migrate :legalize_tf used by AddLegalizeTFToStablehloPasses to tensorflow/compiler/mlir/stablehlo

ecalubaquib
authored and
tensorflower-gardener
committed
Automated Code Change
tensorflower-gardener
committed
[XLA:GPU] Move collectives registration from collective_thunk to a separate target.
tensorflower-gardener
committed
Automated Code Change
tensorflower-gardener
committed
Integrate LLVM at llvm/llvm-project@f9c01b59e3d5
tensorflower-gardener
committed
Automated Code Change
tensorflower-gardener
committed
compat: Update forward compatibility horizon to 2025-04-24
tensorflower-gardener
committed
Update GraphDef version to 2207.
tensorflower-gardener
committed
Update platform_version() method to be consistent with the current PJRT GPU client.

changhuilin
authored and
tensorflower-gardener
committed
temporarily remove references to `sdy.all_to_all` in lit tests because of the assembly format change.

tomnatan30
authored and
tensorflower-gardener
committed
[XLA:GPU] Remove the `FP8ToFP8EndToEnd` from fusion_emitter_device_test.

Moerafaat
authored and
tensorflower-gardener
committed
[XLA:GPU] Remove TODOs related to "support more types" through Triton. This bug doesn't serve an active purpose. Support should be added on-need basis.

Moerafaat
authored and
tensorflower-gardener
committed
Allow disabling nvshmem for build targets.
tensorflower-gardener
committed
Make PrecisionTestsForTriton less implementation-dependent and more readable

gflegar
authored and
tensorflower-gardener
committed
[XLA:GPU] Expose `FindOutputTileSizesForEpilogue` via nest_gemm_fusion.h.

golechwierowicz
authored and
tensorflower-gardener
committed
Reverts 763ff8abced374c8effd14c156ad2a1086e50093

beckerhe
authored and
tensorflower-gardener
committed
[XLA:GPU][Emitters] Enable s4 reductions.

pifon2a
authored and
tensorflower-gardener
committed
Clean up test HLOs.

chsigg
authored and
tensorflower-gardener
committed
Remove the special handling for Quant OPs in TAC
tensorflower-gardener
committed
Adds and populates additional fields in the SolverParams structure.
tensorflower-gardener
committed
Added `WorkerThread` abstraction.

mwhittaker
authored and
tensorflower-gardener
committed
Allow `ResourceVariable::AssignFrom` to not initialize the newly allocated memory
tensorflower-gardener
committed
[xla:ffi] Use std::string_view in FFI API headers

ezhulenev
authored and
tensorflower-gardener
committed
[xla:cpu] Resolve custom call target at construction time

ezhulenev
authored and
tensorflower-gardener
committed
Traverse computations DFS (using post-ordered/scheduled instructions) to better approximate the order of the first execution of each instruction (and the resulting value definition order).
tensorflower-gardener
committed
Add support on checking duplicated donation and error inputs.
tensorflower-gardener
committed
[XLA:GPU] run transpose dimension folding before nesting gemms

metaflow
authored and
tensorflower-gardener
committed
Remove upper bound for Protobuf compatibility.
tensorflower-gardener
committed
Fixing in lowering to LLIR:

Moerafaat
authored and
tensorflower-gardener
committed
Add --platforms flags to .bazelrc for Android, iOS, Macos and elinux.
tensorflower-gardener
committed
[xla:ffi] Add benchmark for enum attribute decoding

ezhulenev
authored and
tensorflower-gardener
committed
[XLA:GPU] copy shape with layout when hoisting bitcasts in nested_gemm_fusion pass

metaflow
authored and
tensorflower-gardener
committed
Re-enable CompiledProgramsCount Test
tensorflower-gardener
committed
Add PjRtDeviceEventOrPromise which represents a promise of a future DeviceEvent

pschuh
authored and
tensorflower-gardener
committed
[PJRT] Add link to the PJRT plugin tutorial slides and recording in the doc.

GleasonK
authored and
tensorflower-gardener
committed
Add more descriptive errors in hlo-translate.

GleasonK
authored and
tensorflower-gardener
committed
Internal changes for cleanup

junwhanahn
authored and
tensorflower-gardener
committed
Internal cleanup of unused libraries.
tensorflower-gardener
committed
#mu add ArrayAttr and DictAttr

chunnienc
authored and
tensorflower-gardener
committed
[XLA:Python] Add nanobind_pywrap_extension (from JAX) and add _ops and _profiler Python wrapper modules.

hawkinsp
authored and
tensorflower-gardener
committed
[XLA:benchmark] Add TargetMetric to BenchmarkConfig and update registry files to set up OSS microbenchmarking

juliagmt-google
authored and
tensorflower-gardener
committed
fork post_quantize.cc to tf_post_quantize.cc without lite dependencies

ecalubaquib
authored and
tensorflower-gardener
committed

Commits on Apr 25, 2025

Commits on Apr 26, 2025

Commits on Apr 27, 2025