Commits

Branch selector

 master

User selector

cezheng

All time

Commit History

Commits on Jul 3, 2024

• [PJRT] Use AnyInvocable for WorkerThread.

  

  cezheng

  authored and

  tensorflower-gardener

  committedJul 3, 2024
  5743bda

Commits on May 22, 2024

• [XLA] Generalize iota tile assignment transpose handling.

  

  cezheng

  authored and

  tensorflower-gardener

  committedMay 22, 2024
  4bc3f5a

Commits on May 20, 2024

• [XLA] Support transposing iota tile assignment with tile dimensions combining multiple reshape dimensions.

  

  cezheng

  authored and

  tensorflower-gardener

  committedMay 20, 2024
  8f3b195

Commits on May 18, 2024

• [XLA] Support more cases in IotaTileAssignment::Transpose.

  

  cezheng

  authored and

  tensorflower-gardener

  committedMay 18, 2024
  3c5c0ed

Commits on Apr 24, 2024

• [XLA] Avoid crash in LatencyHidingScheduler when it fails to schedule a node and improve error message.

  

  cezheng

  authored and

  tensorflower-gardener

  committedApr 24, 2024
  75ef9c3

Commits on Apr 20, 2024

• [PJRT] Change CreateErrorBuffer to take PjRtMemorySpace rather than PjRtDevice.

  

  cezheng

  authored and

  tensorflower-gardener

  committedApr 20, 2024
  a3eb026

Commits on Apr 10, 2024

• [ifrt] Make xla::ifrt::DeviceList default constructible.

  

  cezheng

  authored and

  tensorflower-gardener

  committedApr 10, 2024
  99d1501

• [ifrt] Change Client::GetTopologyForDevices to take a DeviceList instead of a Span of device pointers. This would enable client implementations to make use of the cached hash of the DeviceList.

  

  cezheng

  authored and

  tensorflower-gardener

  committedApr 10, 2024
  9c0eafe

Commits on Jan 23, 2024

• [XLA] Avoid unnecessary deep copy of device groups in SPMDCollectiveOpsCreator's std::function captures.

  

  cezheng

  authored and

  tensorflower-gardener

  committedJan 23, 2024
  a8383eb

Commits on Jan 16, 2024

• [XLA] Remove dead field HloInstruction::operation_queue_id_ after openxla/xla#8452.

  

  cezheng

  authored and

  tensorflower-gardener

  committedJan 16, 2024
  e2a335e

Commits on Jan 4, 2024

• [PJRT] Use absl::AnyInvocable<void() &&> rather than std::function<void()> for BufferFromHostBuffer's on_done_with_host_buffer callback, so that it supports move-only captures and is ensured to be …

  

  cezheng

  authored and

  tensorflower-gardener

  committedJan 4, 2024
  51affe8

Commits on Dec 22, 2023

• [PJRT] Remove a deprecated overload of PjRtBuffer::ToLiteral.

  

  cezheng

  authored and

  tensorflower-gardener

  committedDec 22, 2023
  ea45ca4

Commits on Dec 20, 2023

• [PJRT] Remove PjRtBuffer::OnReady.

  

  cezheng

  authored and

  tensorflower-gardener

  committedDec 20, 2023
  3149c15

Commits on Nov 1, 2023

• [XLA] Speedup IsSubTilingOrEqualSharding a bit.

  

  cezheng

  authored and

  tensorflower-gardener

  committedNov 1, 2023
  7253c28

Commits on Oct 12, 2023

• [XLA:GPU] Fix segfault when accessing HloModule::config(). Don't use the const ref before running passes since some passes may destory the old config object due to the recent copy-on-write change.

  

  cezheng

  authored and

  tensorflower-gardener

  committedOct 12, 2023
  b46b4bc

Commits on Oct 11, 2023

• [XLA] Make HloModule::config_ copy-on-write.

  

  cezheng

  authored and

  tensorflower-gardener

  committedOct 11, 2023
  10a9d17

Commits on Oct 6, 2023

• [PJRT:CPU] Add CreateErrorBuffer implementation.

  

  cezheng

  authored and

  tensorflower-gardener

  committedOct 6, 2023
  de71498

Commits on Sep 18, 2023

• [XLA] Disable async call syntax sugar when the async computation is nontrvial (i.e. does not expand to a single instruction) as that is not lossless. This is to ensure the two formats can roundtrip.

  

  cezheng

  authored and

  tensorflower-gardener

  committedSep 18, 2023
  59a577e

Commits on Sep 15, 2023

• [XLA] Fallback to V1 sharding when it's not possible to preserve V2 in PartialTile.

  

  cezheng

  authored and

  tensorflower-gardener

  committedSep 15, 2023
  57c7009

Commits on Aug 24, 2023

• [XLA] Delete _xla_host_transfer_original_type and _xla_host_transfer_is_lower_bits.

  

  cezheng

  authored and

  tensorflower-gardener

  committedAug 24, 2023
  3f1ca93

Commits on Aug 8, 2023

• Cleanup test modules to use V2 sharding if it's shorter.

  

  cezheng

  authored and

  tensorflower-gardener

  committedAug 8, 2023
  50da242

Commits on Aug 7, 2023

• [XLA:Client] Make HloSharding::iota_tile actually produce V2 shardings.

  

  cezheng

  authored and

  tensorflower-gardener

  committedAug 7, 2023
  2045c07

• [XLA] Increase coverage for V2 sharding.

  

  cezheng

  authored and

  tensorflower-gardener

  committedAug 7, 2023
  9ab9d7a

Commits on Jul 27, 2023

• [PJRT] Add PjRtDevice::PoisonExecution.

  

  cezheng

  authored and

  tensorflower-gardener

  committedJul 27, 2023
  54eeb2a

Commits on Jun 22, 2023

• [XLA] Introduce HloShardingV2 (3/N).

  

  cezheng

  authored and

  tensorflower-gardener

  committedJun 22, 2023
  20c8946

• [XLA] Introduce HloShardingV2 (2/N).

  

  cezheng

  authored and

  tensorflower-gardener

  committedJun 22, 2023
  385bbb1

Commits on Jun 21, 2023

• [XLA] Introduce HloShardingV2 (1/N).

  

  cezheng

  authored and

  tensorflower-gardener

  committedJun 21, 2023
  0a0a4b2

Commits on Jun 13, 2023

• [XLA] Delete one unused overload of HloSharding::PartialTile.

  

  cezheng

  authored and

  tensorflower-gardener

  committedJun 13, 2023
  e5f0117

Commits on Jun 9, 2023

• [XLA] Cleanup build deps for some xla subdirectories.

  

  cezheng

  authored and

  tensorflower-gardener

  committedJun 9, 2023
  d6dfaf8

Commits on May 25, 2023

• [XLA:Client] Change replicate_last_dim to subgroup_types in HloSharding.iota_tile to cover arbitrary subgroups, adding necessary accessors.

  

  cezheng

  authored and

  tensorflower-gardener

  committedMay 25, 2023
  28ecd88

Commits on May 23, 2023

• [XLA:Client] Expose HloSharding pybind factories for iota tile/partial tile, replicated and manual sharding,

  

  cezheng

  authored and

  tensorflower-gardener

  committedMay 23, 2023
  be11761

Commits on Mar 24, 2023

• [TSL:Metric] Reduce contention in Sampller::GetCell.

  

  cezheng

  authored and

  tensorflower-gardener

  committedMar 24, 2023
  bf76570

Commits on Mar 22, 2023

• [XLA] Code cleanups that also reduce code size.

  

  cezheng

  authored and

  tensorflower-gardener

  committedMar 22, 2023
  307adc1

Commits on Mar 16, 2023

• [XLA] Fix debug build failure when CopySliceFrom is provided a empty slice.

  

  cezheng

  authored and

  tensorflower-gardener

  committedMar 16, 2023
  329261a

• Increase shard_count for prng_test.

  

  cezheng

  authored and

  tensorflower-gardener

  committedMar 16, 2023
  8b04f45

Pagination

PreviousNext