| commit |
1.00 |
Use single reshape instead of expand_dims chain in regressio |
|
Explicitly credits gemini-code-assist fo |
2026-08-13 |
| pr |
1.00 |
Fix bessel_i0 intermediate overflow for large finite |x| |
|
Explicitly discloses use of AI coding to |
2026-08-07 |
| pr |
0.78 |
Add init to MLModelConfiguration |
|
Overly formal explanatory prose and spec |
2026-08-19 |
| commit |
0.70 |
PR #47082: Fix infinite backpedal loop in BFCAllocator::Exte |
|
Explicitly states it was generated with |
2026-08-14 |
| commit |
0.70 |
PR #47083: GPU: emit memzero for collective-permute with no |
|
Explicitly states it was generated with |
2026-08-14 |
| commit |
0.70 |
PR #47079: Fix argument-evaluation-order UB in CommonPjRtCli |
|
Explicitly states it was generated with |
2026-08-14 |
| pr |
0.32 |
Fix conv1d eager/graph mode inconsistency for zero-sized out |
|
Highly polished and exhaustive, but grou |
2026-03-18 |
| pr |
0.30 |
lite: fix memory leak in TFLiteSavedModelConverterV2.convert |
|
Detailed and polished, though the concre |
2026-07-23 |
| pr |
0.28 |
Build: Synchronize TensorFlow and XNNPACK tags for ARM build |
|
Polished technical explanation with conc |
2026-04-26 |
| pr |
0.27 |
[tf.data] Validate buffer_size in FixedLengthRecordDataset t |
|
Thorough technical report with concrete |
2026-07-29 |
| pr |
0.24 |
Make checkpoint reader safe for free-threaded Python |
|
Detailed, domain-specific engineering pr |
2026-08-19 |
| pr |
0.22 |
Make TFLite conversion bindings safe for free-threaded Pytho |
|
Highly polished and formulaic, but techn |
2026-08-19 |
| pr |
0.22 |
Make TFLite Python runtime bindings safe for free-threaded P |
|
Very polished structured prose, though d |
2026-08-19 |
| pr |
0.08 |
Add variance_epsilon > 0 validation to batch_normalization |
|
Concise technical rationale with specifi |
2026-07-05 |
| pr |
0.07 |
Fix ELU gradient precision for negative inputs |
|
Detailed numerical and validation specif |
2026-08-07 |
| pr |
0.06 |
fix(rnn): bounds check seq_len_max in BlockLSTM ops to preve |
|
Specific kernel details and issue refere |
2026-08-15 |
| pr |
0.06 |
Fix integer truncation and SIGSEGV in BlockLSTMGrad kernel ( |
|
Specific failure mechanism and concrete |
2026-08-11 |
| pr |
0.03 |
Fix: Validate tensor size to prevent OOM DoS |
|
Terse, issue-specific wording suggests h |
2026-08-19 |
| pr |
0.03 |
Add tensor dimension validation to WriteAudioSummary. Resolv |
|
Terse, issue-focused wording and informa |
2026-01-16 |
| commit |
0.02 |
Clamp dynamic strided_slice runtime size to zero to prevent |
|
Highly specific domain details and refer |
2026-08-13 |
| commit |
0.02 |
Guard TF_TensorToPyArray against tensors above NPY_MAXDIMS |
|
Uses concrete TensorFlow internals, test |
2026-08-13 |
| pr |
0.02 |
Fix XLA autoclustering silently dropping while_loop output |
|
Concise, domain-specific explanation wit |
2026-08-19 |
| pr |
0.02 |
Docs/clarify where gradienttape unconnected |
|
Personal, informal description with natu |
2026-04-05 |
| commit |
0.01 |
Fix stack buffer overflow in TFLite Select reference kernel |
|
Detailed reproduction and implementation |
2026-08-13 |
| commit |
0.01 |
Add inline pylint disables to lines shifted in merge |
|
Terse, repository-specific merge note wi |
2026-08-13 |
| commit |
0.01 |
Resolve merge conflicts with origin/master |
|
Brief conventional commit message descri |
2026-08-13 |
| pr |
0.01 |
fix: TensorScatterUpdate aborts on malformed indices shape |
|
Contains a typo and highly specific repr |
2026-08-19 |
| pr |
0.01 |
Fix tf.raw_ops.Digamma to return -inf for input 0 |
|
Only a terse issue-fix title with no AI- |
2026-03-09 |
| commit |
0.00 |
Fix flakiness in PjRtCopyToMemorySpaceTest.CopyWithDonatedRa |
|
Terse, repository-specific fix message w |
2026-08-14 |
| commit |
0.00 |
Align flag parsing behavior. |
|
Brief, conventional engineering commit m |
2026-08-14 |
| commit |
0.00 |
Rollforward of "Add PjRtRawClient implementation for CpuClie |
|
Concise repository-specific rollforward |
2026-08-14 |
| commit |
0.00 |
Fix integer overflow in ParameterizedTruncatedNormalOp. |
|
Technical explanation uses specific type |
2026-08-14 |
| commit |
0.00 |
Handle self-donated buffers and raw buffer aliases in `Commo |
|
Terse, domain-specific commit title with |
2026-08-14 |
| commit |
0.00 |
Add xla_disable_xsymbol_upload toggle to XLA DebugOptions |
|
Concise technical toggle addition with a |
2026-08-14 |
| commit |
0.00 |
Support reflect and replicate padding in PyTorch convolution |
|
Specific framework and dialect terminolo |
2026-08-14 |
| commit |
0.00 |
Provide native convolution support (`pltpu.conv` and `lax.co |
|
Detailed domain-specific changelist desc |
2026-08-14 |
| commit |
0.00 |
Implement dynamic dimension propagation for `HloOpcode::kOpt |
|
Specific opcode and dynamic-shape contex |
2026-08-14 |
| commit |
0.00 |
Fix type confusion in RpcGetValueOp and memory leak in RpcCl |
|
Terse bug-fix message naming specific op |
2026-08-14 |
| commit |
0.00 |
Fix integer overflow DoS in ExtractImagePatches and ExtractV |
|
Terse, domain-specific fix summary with |
2026-08-14 |
| commit |
0.00 |
Update Shardy to 2832731619ffb4bcc718faa4aa8054214a68c969. |
|
Brief dependency update written in typic |
2026-08-14 |
| commit |
0.00 |
[XLA:GPU] Support integer rounding tolerance in literal comp |
|
Detailed technical summary with domain-s |
2026-08-14 |
| commit |
0.00 |
Delete the floating point optimization level and memory fitt |
|
Short, informal rationale typical of a h |
2026-08-14 |
| commit |
0.00 |
[XLA:CPU] Route non-Eigen convolutions to the MLIR emitter |
|
Highly specific implementation details i |
2026-08-14 |
| commit |
0.00 |
Make XLA LatencyHidingScheduler thread-safe for parallel sim |
|
Concise technical summary with concrete |
2026-08-14 |
| commit |
0.00 |
PR #47225: [ROCm] Widen correlation ids to uint64_t to match |
|
Detailed domain-specific description, de |
2026-08-14 |
| commit |
0.00 |
[XLA:CPU] Enable PermutationSortExpander on CPU |
|
Concise, domain-specific commit descript |
2026-08-14 |
| commit |
0.00 |
Add PjRtRawClient implementation for CpuClient. |
|
Terse commit message with a specific rev |
2026-08-14 |
| commit |
0.00 |
PR #46831: [ROCm] Skip HostMemoryAllocateTest.Numa |
|
Template-driven PR with brief, informal |
2026-08-14 |
| commit |
0.00 |
Use BFC allocators for multi-device GPU autotuning. |
|
Technical description uses specific allo |
2026-08-14 |
| commit |
0.00 |
PR #47238: [ROCm] Add ROCM_PATH env to use correct header pa |
|
Template content is terse and includes c |
2026-08-14 |
| commit |
0.00 |
Remove (now) unused xla/tsl/platform/status_macros.h |
|
Extremely terse removal commit with a sp |
2026-08-14 |
| commit |
0.00 |
Update default OpenXLA CUDA version to 13.2 |
|
Brief, practical release-update notes wi |
2026-08-14 |
| commit |
0.00 |
Fix integer overflow in StringLengthOp loop counters. |
|
Detailed technical rationale references |
2026-08-14 |
| commit |
0.00 |
Adds `element_type` to `RCReference(Wrapper)`. |
|
Short API-change commit message with spe |
2026-08-14 |
| commit |
0.00 |
PR #47218: [xla:gpu] Fix DUS Fusion V2 descriptors when doub |
|
Detailed domain-specific bug explanation |
2026-08-14 |
| commit |
0.00 |
Add unit tests to show dataflow propagation across call boun |
|
Terse, domain-specific commit message wi |
2026-08-14 |
| commit |
0.00 |
Reverts ea98d9d1e05dfe3310fd7179bf1f57daa8b462ac |
|
Standard terse revert message with an in |
2026-08-14 |
| commit |
0.00 |
Add premap buffer size guidance relative to checkpoint size |
|
Concise, repository-specific documentati |
2026-08-14 |
| commit |
0.00 |
Migrate call to xla::cpu::internal::SortInplace away from th |
|
Terse technical migration message using |
2026-08-14 |
| commit |
0.00 |
Properly set the AoT compilation flag in cpu_compiler if it |
|
Concise technical fix naming the exact c |
2026-08-14 |
| commit |
0.00 |
Use self.assertAllClose in convert_to_constants_test.py |
|
Specific testing rationale with float32 |
2026-08-14 |
| commit |
0.00 |
Update and refactor XLA GitHub benchmark workflows. |
|
Detailed but conventional engineering ch |
2026-08-14 |
| commit |
0.00 |
[Mosaic] Add tpu.reduce op, which is similar to vector.multi |
|
Concise domain-specific feature descript |
2026-08-14 |
| commit |
0.00 |
PR #47160: Add an XLA flag to enable/disable cupti multi-sub |
|
Template-like PR text, but its free text |
2026-08-14 |
| commit |
0.00 |
Refactor to prepare for jax.debug.callback() in MPMD. |
|
Terse technical refactoring message nami |
2026-08-14 |
| commit |
0.00 |
Add PjRtRawClient implementation for CpuClient. |
|
Terse, project-specific commit message w |
2026-08-14 |
| commit |
0.00 |
Unify all GPU compliation using PjRtCompile so that ideally |
|
Technical and concise wording with speci |
2026-08-14 |
| commit |
0.00 |
PR #47195: [ROCm] Revert to use ROCm-7.2.4 in CI |
|
Template-structured PR text with specifi |
2026-08-14 |
| commit |
0.00 |
[IFRT] Enforce `MemoryKind` default to be "device" at constr |
|
Domain-specific explanation of IFRT comp |
2026-08-14 |
| commit |
0.00 |
[Mosaic-SC][Test][NFC] Fix clang-tidy modernize-use-bool-lit |
|
Brief, conventional commit message with |
2026-08-13 |
| commit |
0.00 |
Add `DifferentSourceLayout` test to IFRT `reshard_impl_test_ |
|
Concise test-related commit message usin |
2026-08-13 |
| commit |
0.00 |
Replace usages of `GpuTargetConfig.dnn_version_info` with `d |
|
Short, highly specific refactoring commi |
2026-08-13 |
| commit |
0.00 |
Fix ConcreteSharding::HasSamePartitioning ignoring index_dom |
|
Specific bug description and implementat |
2026-08-13 |
| commit |
0.00 |
PR #47149: [XLA] Fix bitcast-convert of literals to and from |
|
Thorough technical PR description, but i |
2026-08-13 |
| commit |
0.00 |
PR #46636: [ROCM] Add MORI inter-process eager initializatio |
|
Detailed domain-specific PR with informa |
2026-08-13 |
| commit |
0.00 |
PR #47191: [XLA:GPU] Demote and rate-limit the delay kernel |
|
Specific technical rationale and referen |
2026-08-13 |
| commit |
0.00 |
Reverts 2a435b40d1922821273bca94729556a2b38384f6 |
|
A terse revert message with a standard p |
2026-08-13 |
| commit |
0.00 |
PR #45996: [GPU][Triton][RaggedDot] Tile propagation for kRa |
|
Detailed domain-specific discussion and |
2026-08-13 |
| commit |
0.00 |
Support sub-instruction mismatch extraction in HLO isolation |
|
Concise, project-specific change descrip |
2026-08-13 |
| commit |
0.00 |
Avoid heap allocation in TypedKernel::Launch |
|
Natural technical explanation includes c |
2026-08-13 |
| commit |
0.00 |
fix heap_use_after_free issue. |
|
Brief informal commit message with a spe |
2026-08-13 |
| commit |
0.00 |
PR #47180: [ROCm] Remove xla bes keyword from jax CI |
|
Template-filled PR text is terse and con |
2026-08-13 |
| commit |
0.00 |
Rollback of PR #46353 |
|
Standard rollback message with a specifi |
2026-08-13 |
| commit |
0.00 |
[XLA:GPU] Update DotMerger to use a (rudimentary) cost-model |
|
Concise technical commit title with a sp |
2026-08-13 |
| commit |
0.00 |
Track dynamic shared memory limit globally per CUfunction. |
|
Specific CUDA implementation details and |
2026-08-13 |
| commit |
0.00 |
[XLA:CPU] Specialize inlined sort for BF16 and F16 |
|
Specific compiler changes, benchmarks, a |
2026-08-13 |
| commit |
0.00 |
Tune Ampere small tile FLOPS derate in GPU dot fusion cost m |
|
Contains specific Ampere performance dat |
2026-08-13 |
| commit |
0.00 |
[XLA:GPU] Add execution unit descriptions to A100 target con |
|
Highly domain-specific hardware referenc |
2026-08-13 |
| commit |
0.00 |
Add missing ABSL_CONST_INIT/TF_CONST_INIT to variable defini |
|
Terse, technical commit message with no |
2026-08-13 |
| commit |
0.00 |
[StreamExecutor] Add Ampere (CC 8.0) execution unit throughp |
|
Detailed, repository-specific implementa |
2026-08-13 |
| commit |
0.00 |
[xla:gpu] Allow converting async collectives to sync in sync |
|
Concise technical explanation tied to sp |
2026-08-13 |
| commit |
0.00 |
Add mlir namespace to mlir casts/isa |
|
Brief error-focused commit message is ch |
2026-08-13 |
| commit |
0.00 |
PR #46765: [XLA:CPU] Fix log1p approximation range |
|
Template structure is expected; free tex |
2026-08-13 |
| commit |
0.00 |
[XLA:CPU] Wrap missing elementwise opcodes in FusionWrapper |
|
Terse change description names exact opc |
2026-08-13 |
| commit |
0.00 |
PR #47127: [ROCm] Fix collective_ops_sharded_unsharded_e2e_t |
|
Template structure is expected; justific |
2026-08-13 |
| commit |
0.00 |
Emit the graph hover CSS in HloDotDumper::Header pre-URI-enc |
|
Terse internal commit wording with a rev |
2026-08-13 |
| commit |
0.00 |
Consolidate the various knobs for enabling op span exposure |
|
Brief project-specific change summary an |
2026-08-13 |
| commit |
0.00 |
[XLA:CPU] Parallelize sort over non-sorting dimensions using |
|
Domain-specific implementation details a |
2026-08-13 |
| commit |
0.00 |
Defer per-pass HLO html graph wrapping and file writes to a |
|
Concise internal engineering summary wit |
2026-08-13 |
| commit |
0.00 |
Cache compiled dump-filter regexes and the hostname/pid dump |
|
Specific HLO dumping terminology and ter |
2026-08-13 |
| commit |
0.00 |
Initial publishing of microbenchmarks for dense and subchann |
|
Short, informal project-specific commit |
2026-08-13 |
| commit |
0.00 |
Move GetDefaultDeviceAssignment to PjRtTopologyDescription. |
|
Concise technical commit with a specific |
2026-08-13 |
| commit |
0.00 |
Propagate dataflow constraints across exponential, iota, and |
|
Detailed domain-specific rationale and t |
2026-08-13 |
| commit |
0.00 |
Export Abseil status symbols in Windows custom op DLLs. |
|
Specific Windows export symbols and file |
2026-08-13 |
| commit |
0.00 |
[IFRT IR] Bump IFRT IR 12 weeks old version to 0.4.0 |
|
Short version-bump commit with a standar |
2026-08-13 |
| commit |
0.00 |
PR #45995: [GPU][Triton][RaggedDot] Tiling space and schedul |
|
Template structure is evident; free text |
2026-08-13 |
| commit |
0.00 |
Integrate StableHLO at openxla/stablehlo@7b1b1578 |
|
Minimal integration commit naming the ex |
2026-08-13 |
| commit |
0.00 |
Produce the output forms of one HLO dump point as parallel t |
|
Terse implementation summary using speci |
2026-08-13 |
| commit |
0.00 |
Implement Highway dynamic dispatch for `PackIntN` and `Unpac |
|
Polished but highly technical descriptio |
2026-08-13 |
| commit |
0.00 |
Exclude elapsed time of instructions in non-main asynchronou |
|
Technical commit message with specific i |
2026-08-13 |
| commit |
0.00 |
source_hash: Introduce a rule to compute the hash of a C++ s |
|
Concise, domain-specific description of |
2026-08-12 |
| commit |
0.00 |
Move deserialize into StreamExecutorExecutable. |
|
Terse, conventional commit message. |
2026-08-12 |
| commit |
0.00 |
PR #46910: [XLA:FFI] Add backend-agnostic collectives FFI ex |
|
Template-driven PR text with specific te |
2026-08-12 |
| commit |
0.00 |
Add optional donated dests for CrossHostReceiveBuffers. |
|
Concise domain-specific change descripti |
2026-08-12 |
| commit |
0.00 |
Configure Bazel downloader mirror rewrites and fix repo mapp |
|
Terse technical commit message naming sp |
2026-08-12 |
| commit |
0.00 |
Apply `internal_visibility` to more targets |
|
Brief conventional commit message. |
2026-08-12 |
| commit |
0.00 |
Allow tflite_extra_gles_deps usage for all Linux variants |
|
Concise technical change with a specific |
2026-08-12 |
| commit |
0.00 |
Internal changes only. |
|
Terse conventional commit message. |
2026-08-12 |
| commit |
0.00 |
Support kTfLiteFloat16 for dynamically quantized fully conne |
|
Specific operator and delegate terminolo |
2026-08-12 |
| pr |
0.00 |
Integrate LLVM at llvm/llvm-project@af6d788c7a01 |
|
— |
2026-08-20 |
| pr |
0.00 |
Automated Code Change |
|
— |
2026-07-18 |
| pr |
0.00 |
PR #121412: Fix out-of-bounds write in `TensorListScatterInt |
|
— |
2026-08-20 |
| pr |
0.00 |
PR #47509: Fix byte-vs-bit shift bug in GPU PadToStatic dyna |
|
— |
2026-08-19 |
| pr |
0.00 |
[XLA:GPU] Rename multimem parameters mosaic attribute to sym |
|
— |
2026-08-18 |
| pr |
0.00 |
Add LLO annotations to Pallas TPU |
|
— |
2026-08-19 |
| pr |
0.00 |
Fix tf.raw_ops.Digamma to return -inf for input 0 |
|
— |
2026-08-19 |
| pr |
0.00 |
Return an error when computing fingerprints for SavedModels |
|
— |
2026-08-19 |
| pr |
0.00 |
Add SplitProtoCompression options and Any reader support for |
|
— |
2026-08-13 |
| pr |
0.00 |
[XLA:GPU] Default enable enable new CuDNN convolution fusion |
|
— |
2026-08-20 |
| pr |
0.00 |
[XLA] Have Buffer Assignment and Heap Simulator respect exec |
|
— |
2026-06-24 |
| pr |
0.00 |
[XLA:GPU] Introduce `DegenerateDimensionRewriter` pass to re |
|
— |
2026-08-20 |
| pr |
0.00 |
[XLA:GPU] Disable expensive Scope Range Tracking during Aggr |
|
— |
2026-08-14 |
| pr |
0.00 |
Add unit tests for the non-flat support on LegalizeSchedulin |
|
— |
2026-08-15 |
| pr |
0.00 |
[XLA:CPU] Handle complex memrefs with non-identity layouts d |
|
— |
2026-08-20 |
| pr |
0.00 |
Relax TpuShapeVerifier for host calls |
|
— |
2026-08-20 |
| pr |
0.00 |
Support dynamic shape on TFRT with PjRt-IFRT |
|
— |
2026-08-20 |
| pr |
0.00 |
Add a new IFRT API for copying multiple multi-device arrays' |
|
— |
2026-08-19 |
| pr |
0.00 |
[XLA:CPU] Vectorize copysign op. |
|
— |
2026-08-20 |
| pr |
0.00 |
Add "manual" tag to GPU_2DEVS_BACKEND in DTensor OSS build d |
|
— |
2026-08-20 |
| pr |
0.00 |
Integrate LLVM at llvm/llvm-project@6df891b91674 |
|
— |
2026-08-20 |
| pr |
0.00 |
Automated Code Change |
|
— |
2026-07-18 |
| pr |
0.00 |
Add mosaic annotations for LLO |
|
— |
2026-08-19 |
| pr |
0.00 |
Introduce SparseCore TF ops for positional weighted BWD pass |
|
— |
2026-08-20 |
| pr |
0.00 |
Add out-of-bounds check for reverse_index_map in SparseFillE |
|
— |
2026-08-20 |
| pr |
0.00 |
PR #46618: [XLA:CPU] Fix crash and wrong results for scan-li |
|
— |
2026-08-03 |
| pr |
0.00 |
PR #47012: [xla:gpu] Fine grained PJRT definition events for |
|
— |
2026-08-14 |
| pr |
0.00 |
Use two-pass variance calculation in BatchNormExpander to pr |
|
— |
2026-08-19 |
| pr |
0.00 |
Add support for emulated dtypes in the cost model. |
|
— |
2026-08-19 |
| pr |
0.00 |
Deprecate ComputeBuffersAt and enhance diagnostics in GetUni |
|
— |
2026-08-17 |
| pr |
0.00 |
PR #45997: [GPU][Triton][RaggedDot] Emitter: infrastructure |
|
— |
2026-08-20 |
| pr |
0.00 |
Integrate StableHLO at openxla/stablehlo@3c9301c1 |
|
— |
2026-08-20 |
| pr |
0.00 |
Fix out-of-bounds read in TFLite BIDIRECTIONAL_SEQUENCE_LSTM |
|
— |
2026-08-18 |
| pr |
0.00 |
Update the GitHub Actions workflow files to use the latest v |
|
— |
2026-08-20 |
| pr |
0.00 |
Set OpMetadata for derived instructions only when the origin |
|
— |
2026-08-20 |
| pr |
0.00 |
[XLA] Add thread filtering to HloLiveRange |
|
— |
2026-08-04 |
| pr |
0.00 |
[XLA:TPU] Fix iteration index propagation for index 0 in HLO |
|
— |
2026-08-19 |
| pr |
0.00 |
Remove linearization related CommonPjRtClient subclass funct |
|
— |
2026-08-20 |
| pr |
0.00 |
Use rules_ml_toolchain for Linux x86_64 -> macOS aarch64 cro |
|
— |
2026-06-18 |
| pr |
0.00 |
Update ops library from eager to lazy importing. |
|
— |
2026-04-09 |
| pr |
0.00 |
Copybara autoresearch 2 |
|
— |
2026-08-20 |
| pr |
0.00 |
Fix SDY round-trip import for decomposed func result shardin |
|
— |
2026-08-18 |
| pr |
0.00 |
Automated Code Change |
|
— |
2026-07-18 |
| pr |
0.00 |
[NFC]Rename the autotuner_pass to config_assigner pass to co |
|
— |
2026-08-19 |
| pr |
0.00 |
Skip batching down in `PriorityTaskQueue` if batch down retu |
|
— |
2026-08-20 |
| pr |
0.00 |
PR #46104: [NVIDIA GPU] Add a new hlo op CollectiveReduce |
|
— |
2026-08-17 |
| pr |
0.00 |
[Mosaic] Allow ND vectors in tpu.vector_store_idx |
|
— |
2026-08-11 |
| pr |
0.00 |
[XLA] Set -Werror=mismatched-tags on clang. |
|
— |
2026-08-20 |
| pr |
0.00 |
Add prefer_estimated_configs to config assigner to pick conf |
|
— |
2026-08-13 |
| pr |
0.00 |
Automated Code Change |
|
— |
2026-07-18 |
| pr |
0.00 |
Automated Code Change |
|
— |
2026-07-18 |
| pr |
0.00 |
Migrate tf2xla and tf_jit to use PjRtCompiler and PjRtClient |
|
— |
2026-08-18 |
| pr |
0.00 |
Automated Code Change |
|
— |
2026-07-18 |
| pr |
0.00 |
C++-side support for batched_copy_to_host_async. |
|
— |
2026-08-20 |
| pr |
0.00 |
Internal changes only |
|
— |
2026-08-18 |
| pr |
0.00 |
Clean up JNI target labels. |
|
— |
2026-08-15 |
| pr |
0.00 |
PR #121412: Fix out-of-bounds write in `TensorListScatterInt |
|
— |
2026-08-20 |
| pr |
0.00 |
Fix forward declaration of amdgpu::Chipset to struct |
|
— |
2026-08-20 |
| pr |
0.00 |
[XLA:CPU] Fix mixed-precision integer dot accumulation (sile |
|
— |
2026-07-27 |
| pr |
0.00 |
[XLA:CPU] Add vectorized sin and cos intrinsic kernels for c |
|
— |
2026-07-22 |
| pr |
0.00 |
Allow stable_sort_expander to be subclassed for more policy |
|
— |
2026-08-07 |
| pr |
0.00 |
Automated Code Change |
|
— |
2026-07-18 |
| pr |
0.00 |
Automated Code Change |
|
— |
2026-07-18 |
| pr |
0.00 |
Refactor CUPTI Hardware Event System (HES) initialization. |
|
— |
2026-08-19 |
| pr |
0.00 |
[XLA:CPU] Use the correct complex division expansion. |
|
— |
2026-08-20 |
| pr |
0.00 |
[XLA]: Fix -1^x in XLA for negative exponents |
|
— |
2026-08-17 |
| pr |
0.00 |
Bump markdown from 3.8 to 3.8.1 in /ci/official/requirements |
|
— |
2026-03-06 |
| pr |
0.00 |
Automated Code Change |
|
— |
2026-07-18 |
| pr |
0.00 |
[XLA:CPU] Fix DynamicUpdateSlice in-place emission inside lo |
|
— |
2026-08-20 |
| pr |
0.00 |
Fix convertToSdySharding and ImportShardingsPass for 0-resul |
|
— |
2026-08-20 |
| pr |
0.00 |
PR #46893: [XLA] Initialize dynamic dimension sizes in Liter |
|
— |
2026-08-20 |
| pr |
0.00 |
Add usable limits section to E1001 error doc. |
|
— |
2026-08-20 |
| pr |
0.00 |
Add unit test to show constant folding or the lack thereof o |
|
— |
2026-08-20 |
| pr |
0.00 |
Align Blackwell small-tile Tensor Core FLOPS derate threshol |
|
— |
2026-08-19 |
| pr |
0.00 |
[XLA:GPU] Extend xla_disable_hlo_passes (and xla_enable_pass |
|
— |
2026-08-20 |
| pr |
0.00 |
Enable MoE with ynnpack. |
|
— |
2026-08-20 |
| pr |
0.00 |
[CI - Windows] Add LLVM19 toolchain. |
|
— |
2026-08-19 |
| pr |
0.00 |
Create a heuristic as to whether we should fuse a concat int |
|
— |
2026-08-20 |
| pr |
0.00 |
[XLA:XTile] Move block_level_parameters to codegen/xtile |
|
— |
2026-08-18 |
| pr |
0.00 |
Enforce 32-byte host DMA alignment and padded transfer size |
|
— |
2026-08-14 |
| pr |
0.00 |
Automated Code Change |
|
— |
2026-08-20 |
