| PR |
0.70 |
Add BUILD_svcbtc file |
|
Unusual phrasing, emojis, and lack of te |
2026-05-15 |
| PR |
0.21 |
Raise explicit error for remote DTensor meshes in to_numpy |
|
Formal summary, still domain-specific an |
2026-05-16 |
| COMMIT |
0.20 |
Automated Code Change |
|
Generic 'Automated Code Change' phrasing |
2026-05-16 |
| COMMIT |
0.20 |
Automated Code Change |
|
Generic 'Automated Code Change' phrasing |
2026-05-16 |
| COMMIT |
0.20 |
Automated Code Change |
|
Generic 'Automated Code Change' phrasing |
2026-05-16 |
| COMMIT |
0.10 |
Automated Code Change |
|
Generic automation message, but plausibl |
2026-05-15 |
| COMMIT |
0.10 |
Automated Internal Change |
|
Generic automation, but no explicit AI s |
2026-05-15 |
| COMMIT |
0.10 |
Automated Code Change |
|
Generic 'Automated Code Change' may impl |
2026-05-15 |
| COMMIT |
0.10 |
Automated Code Change |
|
Generic 'Automated Code Change' may impl |
2026-05-15 |
| COMMIT |
0.10 |
Automated Code Change |
|
Generic 'Automated Code Change' may impl |
2026-05-15 |
| COMMIT |
0.10 |
Automated Code Change |
|
Generic 'Automated Code Change' may impl |
2026-05-15 |
| COMMIT |
0.00 |
Integrate LLVM at llvm/llvm-project@88afb5af47a9 |
|
Concise commit summary, no AI traits. |
2026-05-16 |
| COMMIT |
0.00 |
Enable `xla_enable_rgv3_materialization` in the default HLO |
|
Dense technical detail and domain-specif |
2026-05-16 |
| COMMIT |
0.00 |
[XLA:GPU] fusion_block_level_rewriter: do not force multi-ou |
|
Brief, uses code-specific shorthand, hum |
2026-05-16 |
| COMMIT |
0.00 |
PR #41038: [ROCm] Use hermetic llvm to compile xla |
|
Some template structure, but change summ |
2026-05-16 |
| COMMIT |
0.00 |
Skip `NoReshardOnBroadcastDims` for V3 and write a V3 test f |
|
Technical discussion, minor grammatical |
2026-05-16 |
| COMMIT |
0.00 |
Add GetMemorySpaceKindIds() to PjRtTopologyDescription. |
|
Technical and purpose-specific commit, h |
2026-05-16 |
| COMMIT |
0.00 |
Add TfLiteIntArrayCreate() and TfLiteIntArrayFree() to TfLit |
|
Explains platform-specific issue using n |
2026-05-16 |
| COMMIT |
0.00 |
Make TensorResponse::InitPartial return absl::Status. |
|
Brief and technical commit message; no A |
2026-05-16 |
| COMMIT |
0.00 |
Optimize ArgMinMax (Reduce) when fed by scalar constant via |
|
Technical summary with domain specifics; |
2026-05-16 |
| COMMIT |
0.00 |
PR #42464: [XLA:GPU][oneAPI] Part4 of oneAPI plugin: SE GPU |
|
Template import with human-written commi |
2026-05-15 |
| COMMIT |
0.00 |
Replace gtl::linked_hash_map with absl::flat_hash_map. |
|
Very concise, domain-typical commit mess |
2026-05-15 |
| COMMIT |
0.00 |
PR #34219: Document "--xla_syntax_sugar_async_ops" flag |
|
Slightly more formal, but markdown and e |
2026-05-15 |
| COMMIT |
0.00 |
[GPU] Migrate RequiresCollectiveSymmetricMemorySpace to unif |
|
Highly domain-specific technical descrip |
2026-05-15 |
| COMMIT |
0.00 |
PR #40593: [XLA:GPU] Ensure that the host and target platfor |
|
Typical PR template with detailed human |
2026-05-15 |
| COMMIT |
0.00 |
Support `PartialReplication` in `HloSharding::NumTiles` |
|
Standard technical PR summary; uses doma |
2026-05-15 |
| COMMIT |
0.00 |
tensorflow: Add direct dependency on xxd. |
|
Concise commit message with context, sta |
2026-05-15 |
| COMMIT |
0.00 |
Refactor `UniqueTiledDim` to re-use `HloSharding::TiledDataR |
|
Refactoring commit message; terse and te |
2026-05-15 |
| COMMIT |
0.00 |
Implement MakeCanonicalShapeForMemorySpace for GPU and c-api |
|
Terse, technical commit message without |
2026-05-15 |
| COMMIT |
0.00 |
make compilation of llvm ir to ptx async for Triton |
|
Terse and domain-specific language, no A |
2026-05-15 |
| COMMIT |
0.00 |
PR #37936: [NVIDIA GPU] Reduce rendezvous frequency in gpu e |
|
Contains human-like summary and edit his |
2026-05-15 |
| COMMIT |
0.00 |
Integrate LLVM at llvm/llvm-project@4f52e220a69f |
|
Standard LLVM integration message, no AI |
2026-05-15 |
| COMMIT |
0.00 |
The fusion_constant_sinking pass is updated to support sinki |
|
Precise technical language and test focu |
2026-05-15 |
| COMMIT |
0.00 |
Fix flaky test distributed_save_ft_test. |
|
Brief, task-focused message with no AI i |
2026-05-15 |
| COMMIT |
0.00 |
CUDA 13 Release build changes |
|
Release note in standard format; only te |
2026-05-15 |
| COMMIT |
0.00 |
strip_memory_placement_annotations_test uses HloTestBase, bu |
|
Technical note about test deprecation, t |
2026-05-15 |
| COMMIT |
0.00 |
collect spilled register stats for CustomKernelThunks |
|
Domain-specific debugging message, human |
2026-05-15 |
| COMMIT |
0.00 |
Return output path in Splitter::Write() with rollforward fix |
|
Technical revert note, precise and speci |
2026-05-15 |
| COMMIT |
0.00 |
Fix `AnalyzeTileAssignment` in `tile_assignment.cc` to const |
|
Terse and technical commit message with |
2026-05-15 |
| COMMIT |
0.00 |
[PART 2] Externalize HLO isolation test into a new open-sour |
|
Contains domain jargon and informal tone |
2026-05-15 |
| COMMIT |
0.00 |
Make it possible for users of unsafe_client_pointer to fetch |
|
Technical language and specific referenc |
2026-05-15 |
| COMMIT |
0.00 |
delete unmaintained code. |
|
Terse, informal, lacks AI-typical phrasi |
2026-05-15 |
| COMMIT |
0.00 |
Revert zstd.patch update due to source of changes being reve |
|
Standard structured revert message, typi |
2026-05-15 |
| COMMIT |
0.00 |
Add documentations regarding the expected lock acquisition b |
|
Human typo in 'documentations', brief te |
2026-05-15 |
| COMMIT |
0.00 |
[XLA:GPU] Handle zero tile sizes in XTile emitters |
|
Technical title, informal bracketed pref |
2026-05-15 |
| COMMIT |
0.00 |
Integrate LLVM at llvm/llvm-project@adb5802b494c |
|
Structured technical update with referen |
2026-05-15 |
| COMMIT |
0.00 |
Add checksum verification for downloaded binaries in ml_buil |
|
Brief technical description, domain-typi |
2026-05-15 |
| COMMIT |
0.00 |
Support `HloShardingV3` in `MakeTiledPartitionOrdinals` wher |
|
Highly technical, uses domain-specific l |
2026-05-15 |
| COMMIT |
0.00 |
[XLA:CPU] Implement StripMemoryPlacementAnnotations pass and |
|
Technical jargon and changelog format; h |
2026-05-15 |
| COMMIT |
0.00 |
Re-enable CuDNN fusion tests and relax workspace size check. |
|
Technical, succinct changes, no AI phras |
2026-05-15 |
| COMMIT |
0.00 |
PR #41817: [XLA:GPU] Migrate RaggedAllToAll command buffer |
|
Imported PR with structured human-contri |
2026-05-15 |
| COMMIT |
0.00 |
Fix data races with GetAttrToInputsMapHoldingGIL and GetAttr |
|
Technical wording, explicit detail, huma |
2026-05-15 |
| COMMIT |
0.00 |
[XLA:GPU] Change the shard count for fusion_emitter_parametr |
|
Brief, direct statement, human-typical c |
2026-05-15 |
| COMMIT |
0.00 |
PR #41997: [ROCm] Stop rocprofiler context in RocmTracer::Di |
|
PR import; content includes domain detai |
2026-05-15 |
| COMMIT |
0.00 |
PR #42562: Update BANT_VERSION to 0.2.8 |
|
Project update with direct language, typ |
2026-05-15 |
| COMMIT |
0.00 |
nvml is absolutily threadsafe |
|
Informal, contains typo ('absolutily'), |
2026-05-15 |
| COMMIT |
0.00 |
Adjust XLA buffer assignment: use 1.5 GiB safety margin for |
|
Technical, concise, uses domain-specific |
2026-05-15 |
| COMMIT |
0.00 |
[XLA:GPU] Delete GpuCodegenTest. |
|
Brief, direct, domain-specific commit me |
2026-05-15 |
| COMMIT |
0.00 |
Support subgroup partial replication instead of Last tile di |
|
Concise, technical, specific to function |
2026-05-15 |
| COMMIT |
0.00 |
[XLA:GPU] Update softmax rewriter to use new tiling analysis |
|
Technical update, domain-specific termin |
2026-05-15 |
| COMMIT |
0.00 |
PR #35868: Fix crashes due to kernel cache confusion. (#3322 |
|
Detailed PR summary with technical conte |
2026-05-15 |
| COMMIT |
0.00 |
Update toolchain patch for LLVM. |
|
Technical and specific; no AI stylistic |
2026-05-15 |
| COMMIT |
0.00 |
Integrate LLVM at llvm/llvm-project@9a4faee1068c |
|
LLVM integration message is standard and |
2026-05-15 |
| COMMIT |
0.00 |
util: Replace embed_files with cc_embed_data and move xxd to |
|
Uses domain-specific jargon and informal |
2026-05-15 |
| COMMIT |
0.00 |
Add MakeCanonicalShapeForMemorySpace to allow fetching the c |
|
Succinct, technical, no AI-like tone or |
2026-05-15 |
| COMMIT |
0.00 |
[IFRT] Precompute the tile information of `xla::HloSharding` |
|
Somewhat formal, but mostly technical wi |
2026-05-15 |
| COMMIT |
0.00 |
[XLA] Initial sparse convolution support for HloEvaluator. |
|
Technical description, but slight formal |
2026-05-15 |
| COMMIT |
0.00 |
nvml is absolutily threadsafe |
|
Colloquial and includes a typo ('absolut |
2026-05-14 |
| COMMIT |
0.00 |
Fix a bug where `StreamExecutorGpuClient::CopyToMemorySpace` |
|
Direct bug summary, technical, no AI mar |
2026-05-14 |
| COMMIT |
0.00 |
[GPU] Zero-Copy one-shot Ragged-All-To-All optimization |
|
Technical changelog, domain detail, info |
2026-05-14 |
| COMMIT |
0.00 |
Implement lazy cancellation filtering in PriorityTaskQueue |
|
Terse technical phrasing, no AI signals. |
2026-05-14 |
| COMMIT |
0.00 |
bounds check buffer ids in StringVectorPullTableEntry |
|
Short, lacks AI style, typical human com |
2026-05-14 |
| COMMIT |
0.00 |
Fix clang-tidy diff test failure on ARM64 CI. |
|
Short, technical summary with no AI hall |
2026-05-14 |
| COMMIT |
0.00 |
PR #40921: [NVIDIA GPU ] Move barrier out of loop in collect |
|
Informal tone and domain-specific refere |
2026-05-14 |
| COMMIT |
0.00 |
PR #42104: [GPU] Fix cuBLAS FP8 GEMM dimension handling. |
|
Technical content, concise, with no sign |
2026-05-14 |
| COMMIT |
0.00 |
util: Replace embed_files with cc_embed_data and move xxd to |
|
Domain-specific, informal style; lacks A |
2026-05-14 |
| COMMIT |
0.00 |
Change value_index type to SPLITS_TYPE in RaggedRangeOp |
|
Very terse, technical; no indication of |
2026-05-14 |
| COMMIT |
0.00 |
dlclose pjrt plugins on error |
|
Brief, imperative and technical; does no |
2026-05-14 |
| COMMIT |
0.00 |
PR #42364: [ROCm] Add TDM (Tensor Descriptor Memory) support |
|
Technical, specific; phrasing is not AI- |
2026-05-14 |
| COMMIT |
0.00 |
check return value of PyArray_SetBaseObject |
|
Short, imperative style; no AI signals p |
2026-05-14 |
| COMMIT |
0.00 |
add default case in ConvertFromPjRtBuffer |
|
Short and technical; lacks typical AI te |
2026-05-14 |
| COMMIT |
0.00 |
Roll back https://github.com/openxla/xla/pull/42346 due to O |
|
Concise, technical rollback message; cle |
2026-05-14 |
| COMMIT |
0.00 |
PR #42511: Bump github/codeql-action from 4.35.2 to 4.35.3 |
|
— |
2026-05-14 |
| COMMIT |
0.00 |
PR #42537: [xla:gpu] Do not resolve leaf dependencies throug |
|
Technical and concise commit/pr message, |
2026-05-14 |
| COMMIT |
0.00 |
PR #42545: [ROCm] Disable multigpu tests and use singlegpu r |
|
Slightly formal, but uses template; free |
2026-05-14 |
| COMMIT |
0.00 |
Add explicit check for `!IsTuple` in `HloSharding::tile_assi |
|
Direct, technical, and domain-specific p |
2026-05-14 |
| COMMIT |
0.00 |
Handle "control_dep" custom calls in XLA. |
|
Concise technical description, no AI mar |
2026-05-14 |
| COMMIT |
0.00 |
Optimize HloValue::IsRootOf by directly comparing against th |
|
Detailed technical commit, domain langua |
2026-05-14 |
| COMMIT |
0.00 |
avoid unnecessary copies in xplane_to_profile_instructions |
|
Informal, technical phrasing, not AI-lik |
2026-05-14 |
| COMMIT |
0.00 |
avoid CHECK in SocketServer |
|
Direct, domain-specific commit message. |
2026-05-14 |
| COMMIT |
0.00 |
remove std::move in lambda capture |
|
Short, terse commit message, no AI trait |
2026-05-14 |
| COMMIT |
0.00 |
Add out-of-memory (OOM) status payload helpers to PJRT error |
|
Technical description, clear domain cont |
2026-05-14 |
| COMMIT |
0.00 |
[XLA] Add HloSharding reconstruction utils |
|
Technical PR, uses domain terms and deta |
2026-05-14 |
| COMMIT |
0.00 |
PR #41809: [XLA:GPU] Migrate CollectiveBroadcast command b |
|
Technical, concise, domain-specific; nor |
2026-05-14 |
| COMMIT |
0.00 |
Fix HLO path for default_registry.yml. |
|
Very terse, domain-specific commit messa |
2026-05-14 |
| COMMIT |
0.00 |
Add overflow checking to dimension multiplication in broadca |
|
Short, technical language, domain-specif |
2026-05-14 |
| COMMIT |
0.00 |
PR #42523: [XLA/TSL] Three macOS arm64 build fixes |
|
Detailed, technical description, domain |
2026-05-14 |
| COMMIT |
0.00 |
[PJRT] Disambiguate temporary dump directory names to elimin |
|
Terse, technical, typographical errors s |
2026-05-14 |
| COMMIT |
0.00 |
PR #42577: [oneAPI] Fix the Postsubmit failures for Intel GP |
|
Technical explanation, minor verbosity, |
2026-05-14 |
| COMMIT |
0.00 |
Add sharding conversion in the code paths where Python callb |
|
Brief, domain-specific, structured, typi |
2026-05-14 |
| COMMIT |
0.00 |
Run shardy/google/integrate_latest.sh to integrate. |
|
Terse, procedural language; characterist |
2026-05-14 |
| COMMIT |
0.00 |
Flatten only the children for the first visit for always-clo |
|
Technical, informal explanation; fits hu |
2026-05-14 |
| PR |
0.00 |
Fix potential integer overflow and improve error handling in |
|
— |
2026-05-15 |
| PR |
0.00 |
Add bounds checks and handle memory allocation failures in E |
|
— |
2026-05-13 |
| PR |
0.00 |
Set default hidden visibility for the hwloc build. |
|
— |
2026-05-15 |
| PR |
0.00 |
Add bounds check for element length in MemmappedFileSystem. |
|
— |
2026-05-13 |
| PR |
0.00 |
Fix potential integer overflows in TFLite kernel offset calc |
|
— |
2026-05-13 |
| PR |
0.00 |
Enhance EagerService op validation and type checking. |
|
— |
2026-05-16 |
| PR |
0.00 |
Add nullptr check before setting shardy shardings from parse |
|
— |
2026-05-15 |
| PR |
0.00 |
Support virtual device stream priority in PluggableDevice |
|
— |
2026-05-08 |
| PR |
0.00 |
Automated Code Change |
|
— |
2026-05-16 |
| PR |
0.00 |
Integrate LLVM at llvm/llvm-project@88afb5af47a9 |
|
— |
2026-05-16 |
| PR |
0.00 |
Enable `xla_enable_rgv3_materialization` in the default HLO |
|
— |
2026-05-14 |
| PR |
0.00 |
[XLA:GPU] fusion_block_level_rewriter: do not force multi-ou |
|
— |
2026-05-13 |
| PR |
0.00 |
PR #41038: [ROCm] Use hermetic llvm to compile xla |
|
— |
2026-05-14 |
| PR |
0.00 |
Skip `NoReshardOnBroadcastDims` for V3 and write a V3 test f |
|
— |
2026-05-13 |
| PR |
0.00 |
Prototype enabling `HloShardingV3` |
|
— |
2026-03-03 |
| PR |
0.00 |
[XLA:CPU] Fix double-rounding numerical bug in float8 conver |
|
— |
2026-05-16 |
| PR |
0.00 |
Automated Code Change |
|
— |
2026-05-16 |
| PR |
0.00 |
Reverts 90dbe1c419bb07a7ab219175349fc1b7cb3f59b4 |
|
— |
2026-05-15 |
| PR |
0.00 |
Automated Code Change |
|
— |
2026-05-16 |
| PR |
0.00 |
Automated Code Change |
|
— |
2026-05-16 |
| PR |
0.00 |
Automated Code Change |
|
— |
2026-05-16 |
| PR |
0.00 |
Automated Code Change |
|
— |
2026-05-16 |
| PR |
0.00 |
Automated Code Change |
|
— |
2026-05-16 |
| PR |
0.00 |
Automated Code Change |
|
— |
2026-05-16 |
| PR |
0.00 |
Automated Code Change |
|
— |
2026-05-16 |
| PR |
0.00 |
Automated Code Change |
|
— |
2026-05-16 |
| PR |
0.00 |
Automated Code Change |
|
— |
2026-05-16 |
| PR |
0.00 |
Automated Code Change |
|
— |
2026-05-16 |
| PR |
0.00 |
Automated Code Change |
|
— |
2026-05-16 |
| PR |
0.00 |
Automated Code Change |
|
— |
2026-05-16 |
| PR |
0.00 |
Add native C++ to Wasm translation boundaries and TypeScript |
|
— |
2026-05-16 |
| PR |
0.00 |
Automated Code Change |
|
— |
2026-05-16 |
| PR |
0.00 |
Automated Code Change |
|
— |
2026-05-16 |
| PR |
0.00 |
Automated Code Change |
|
— |
2026-05-16 |
| PR |
0.00 |
Automated Code Change |
|
— |
2026-05-16 |
| PR |
0.00 |
Automated Code Change |
|
— |
2026-05-16 |
| PR |
0.00 |
Automated Code Change |
|
— |
2026-05-12 |
| PR |
0.00 |
Adds parameterized tests for ragged-all-to-all memory space |
|
— |
2026-05-15 |
| PR |
0.00 |
Automated Code Change |
|
— |
2026-05-11 |
| PR |
0.00 |
Update reshard_impl_test_lib to properly canonicalize layout |
|
— |
2026-05-16 |
| PR |
0.00 |
Automated Code Change |
|
— |
2026-05-11 |
| PR |
0.00 |
Automated Code Change |
|
— |
2026-05-16 |
| PR |
0.00 |
Add GetMemorySpaceKindIds() to PjRtTopologyDescription. |
|
— |
2026-05-16 |
| PR |
0.00 |
Automated Code Change |
|
— |
2026-05-14 |
| PR |
0.00 |
Add GetMemorySpaceKindIds() to PjRtTopologyDescription. |
|
— |
2026-05-16 |
| PR |
0.00 |
Add F6E3M2FN and F6E2M3FN to xla::PrimitiveType. |
|
— |
2026-05-16 |
| PR |
0.00 |
Add TfLiteIntArrayCreate() and TfLiteIntArrayFree() to TfLit |
|
— |
2026-05-12 |
| PR |
0.00 |
Make TensorResponse::InitPartial return absl::Status. |
|
— |
2026-05-15 |
| PR |
0.00 |
Optimize ArgMinMax (Reduce) when fed by scalar constant via |
|
— |
2026-05-14 |
| PR |
0.00 |
[XLA][HLO Value Tracking] Add original value attribute in MH |
|
— |
2026-05-15 |
| PR |
0.00 |
[XLA] HLO Isolator: Add test for miscompare file dumping and |
|
— |
2026-05-15 |
| PR |
0.00 |
PR #42464: [XLA:GPU][oneAPI] Part4 of oneAPI plugin: SE GPU |
|
— |
2026-05-15 |
| PR |
0.00 |
Replace gtl::linked_hash_map with absl::flat_hash_map. |
|
— |
2026-05-15 |
| PR |
0.00 |
Remove null checks for `get_ptr` in SparseCore continuation |
|
— |
2026-05-15 |
| PR |
0.00 |
PR #34219: Document "--xla_syntax_sugar_async_ops" flag |
|
— |
2026-05-11 |
| PR |
0.00 |
[XLA] Fix continuation queue to support late binding |
|
— |
2026-05-15 |
| PR |
0.00 |
[GPU] Migrate RequiresCollectiveSymmetricMemorySpace to unif |
|
— |
2026-05-15 |
| PR |
0.00 |
PR #32521: [XLA:CPU][oneDNN] Update dot plus elementwise op |
|
— |
2026-05-12 |
| PR |
0.00 |
util: Replace embed_files with cc_embed_data and drop xxd de |
|
— |
2026-05-15 |
| PR |
0.00 |
PR #40593: [XLA:GPU] Ensure that the host and target platfor |
|
— |
2026-05-11 |
| PR |
0.00 |
Update visibility to pathway OSS. |
|
— |
2026-05-15 |
| PR |
0.00 |
Support `PartialReplication` in `HloSharding::NumTiles` |
|
— |
2026-05-15 |
| PR |
0.00 |
tensorflow: Add direct dependency on xxd. |
|
— |
2026-05-15 |
| PR |
0.00 |
PR #42374: Lift CrossHostTransferBuffers into CommonPjRtClie |
|
— |
2026-05-14 |
| PR |
0.00 |
Refactor `UniqueTiledDim` to re-use `HloSharding::TiledDataR |
|
— |
2026-05-15 |
| PR |
0.00 |
Added logic to securely initialize `Indvar` at the start of |
|
— |
2026-05-15 |
| PR |
0.00 |
Corrected the size calculation for host memory allocation in |
|
— |
2026-05-15 |
| PR |
0.00 |
Update HDF5 plugin path handling in Keras. |
|
— |
2026-05-15 |
| PR |
0.00 |
Add structured cost fields to QueryCost and RequestCost APIs |
|
— |
2026-05-15 |
| PR |
0.00 |
Implement MakeCanonicalShapeForMemorySpace for GPU and c-api |
|
— |
2026-05-15 |
| PR |
0.00 |
make compilation of llvm ir to ptx async for Triton |
|
— |
2026-05-11 |
| PR |
0.00 |
Ensure thread-safe concurrent initialization in PjRtCompiler |
|
— |
2026-05-15 |
| PR |
0.00 |
PR #42248: [ROCm] Add Triton autotuning configs for MI300 an |
|
— |
2026-05-11 |
| PR |
0.00 |
PR #37936: [NVIDIA GPU] Reduce rendezvous frequency in gpu e |
|
— |
2026-05-15 |
| PR |
0.00 |
[ReplicaGroupV3][JAX+stablehlo][8/n] Renaming mesh symbols i |
|
— |
2026-05-14 |
| PR |
0.00 |
Update XNNPACK in XLA |
|
— |
2026-05-15 |
| PR |
0.00 |
Integrate LLVM at llvm/llvm-project@4f52e220a69f |
|
— |
2026-05-15 |
| PR |
0.00 |
The fusion_constant_sinking pass is updated to support sinki |
|
— |
2026-05-13 |
| PR |
0.00 |
Fix flaky test distributed_save_ft_test. |
|
— |
2026-05-15 |
| PR |
0.00 |
Enable lazy imports for //third_party/tensorflow/python/trai |
|
— |
2026-05-15 |
| PR |
0.00 |
[XLA:TPU] Refine CallInliner override policy in HloHostDevic |
|
— |
2026-05-15 |
| PR |
0.00 |
Add validation for remote eager operations and packed handle |
|
— |
2026-05-15 |
| PR |
0.00 |
CUDA 13 Release build changes |
|
— |
2026-05-09 |
| PR |
0.00 |
[SDY][ExplicitReduceScatter] Export SDY reduce-scatter into |
|
— |
2026-05-06 |
| PR |
0.00 |
strip_memory_placement_annotations_test uses HloTestBase, bu |
|
— |
2026-05-15 |
| PR |
0.00 |
Add support for log fp16 in XNNPACK delegate. |
|
— |
2026-05-15 |
| PR |
0.00 |
collect spilled register stats for CustomKernelThunks |
|
— |
2026-05-14 |
| PR |
0.00 |
Migrate cuLaunchHostFunc to cuLaunchHostFunc_v2 with SPINWAI |
|
— |
2026-05-15 |
| PR |
0.00 |
Return output path in Splitter::Write() with rollforward fix |
|
— |
2026-05-13 |
| PR |
0.00 |
Prevent HLO printer from generating invalid output with cust |
|
— |
2026-05-15 |
| PR |
0.00 |
Fix `AnalyzeTileAssignment` in `tile_assignment.cc` to const |
|
— |
2026-05-13 |
| PR |
0.00 |
Enable lazy imports for //third_party/tensorflow/python/ops: |
|
— |
2026-05-15 |
| PR |
0.00 |
Move targets depending on legacy HLO runner to restricted su |
|
— |
2026-04-21 |
| PR |
0.00 |
[PART 2] Externalize HLO isolation test into a new open-sour |
|
— |
2026-05-14 |
| PR |
0.00 |
Add offending input shapes and cached shapes to IFRT compila |
|
— |
2026-05-15 |
| PR |
0.00 |
Make it possible for users of unsafe_client_pointer to fetch |
|
— |
2026-05-14 |
| PR |
0.00 |
Fix Shardy partitioner to by-pass optimization barriers when |
|
— |
2026-05-15 |
| PR |
0.00 |
Update tensorflow Copybara directives |
|
— |
2026-05-14 |
