Commits

Branch selector

 master

User selector

All users

All time

Commit History

Commits on Jan 13, 2024

• Increase maxZoom factor for hlo graph explorer

  Show description for 93e492a
  

  zzzaries

  authored and

  tensorflower-gardener

  committed
  93e492a

  Copy full SHA for 93e492a

• Ensure the allocation type is kTfLiteCustom when doing shallow copies in DeepOrShallowCopyTensorsShapeTypeData.

  Show description for 798b2eb
  

  timonvo

  authored and

  tensorflower-gardener

  committed
  798b2eb

  Copy full SHA for 798b2eb

• Introduce some macros to define XLA errors.

  Show description for 8feffce
  

  klucke

  authored and

  tensorflower-gardener

  committed
  8feffce

  Copy full SHA for 8feffce

• [XLA] Remove slice to dynamic custom call with pad to static operand

  Show description for a2bbced
  

  blakehechtman

  authored and

  tensorflower-gardener

  committed
  a2bbced

  Copy full SHA for a2bbced

• Adds options for Pixel devices.

  Show description for 1360a12
  tensorflower-gardener

  committed
  1360a12

  Copy full SHA for 1360a12

• Run combined bridge first

  Show description for fddc3db
  tensorflower-gardener

  committed
  fddc3db

  Copy full SHA for fddc3db

Commits on Jan 12, 2024

• #tf-data Return error if symbolic checkpointing fails in parallel_interleave

  Show description for c0cc1dd
  tensorflower-gardener

  committed
  c0cc1dd

  Copy full SHA for c0cc1dd

• [xla:gpu] Add GpuExecutable post-initialization rendezvous to avoid deadlocks

  Show description for a43ec75
  

  ezhulenev

  authored and

  tensorflower-gardener

  committed
  a43ec75

  Copy full SHA for a43ec75

• [xla:gpu] Add support for scheduling conditional into a command buffer

  Show description for c43070c
  

  tyb0807

  authored and

  tensorflower-gardener

  committed
  c43070c

  Copy full SHA for c43070c

• Replace `std::vector<int64_t>` with `DimensionVector` in `hlo_sharding_util` if the length of the vector is similar to the rank of a tensor.

  Show description for b679eb2
  

  ZixuanJiang

  authored and

  tensorflower-gardener

  committed
  b679eb2

  Copy full SHA for b679eb2

• Migrate wasm32 constraint to emscripten

  Show description for e7ca442
  tensorflower-gardener

  committed
  e7ca442

  Copy full SHA for e7ca442

• - Handles new TFCI format for any and bisect

  Show description for d0a969c
  

  angerson

  authored and

  tensorflower-gardener

  committed
  d0a969c

  Copy full SHA for d0a969c

• [xla:gpu] Support emitting ConditionalThunk for command buffers

  Show description for 7f82464
  

  tyb0807

  authored and

  tensorflower-gardener

  committed
  7f82464

  Copy full SHA for 7f82464

• [xla:gpu] Remove WhileThunkNotSupported guard

  Show description for a3364f1
  

  tyb0807

  authored and

  tensorflower-gardener

  committed
  a3364f1

  Copy full SHA for a3364f1

• Catch error and print understandable message from loading SavedModel in SavedModel CLI, and also skip loading checkpoints for better efficiency.

  Show description for 6a818ff
  

  k-w-w

  authored and

  tensorflower-gardener

  committed
  6a818ff

  Copy full SHA for 6a818ff

• [xla:gpu] Add support for emitting SelectAndScatter from HLO

  Show description for 190e9b5
  

  tyb0807

  authored and

  tensorflower-gardener

  committed
  190e9b5

  Copy full SHA for 190e9b5

• [xla:gpu] Support adding cu_threefry2x32 custom kernel to command buffer

  Show description for 4004d64
  

  anlunx

  authored and

  tensorflower-gardener

  committed
  4004d64

  Copy full SHA for 4004d64

• Remove multi_platform_manager_listener module initializer sequencing. Listener support was previously removed (and had no initialization anyway).

  Show description for 7a673cd
  

  klucke

  authored and

  tensorflower-gardener

  committed
  7a673cd

  Copy full SHA for 7a673cd

• Add a counter to track replicated graphs without a device type

  Show description for 98b7442
  

  jcai19

  authored and

  tensorflower-gardener

  committed
  98b7442

  Copy full SHA for 98b7442

• [xla:gpu] Clean up NCCL utils library and extract a separate nccl_clique lib

  Show description for cae6994
  

  ezhulenev

  authored and

  tensorflower-gardener

  committed
  cae6994

  Copy full SHA for cae6994

• [XLA] Purge AsyncOpCanonicalizer

  Show description for e387177
  

  vsytch

  authored and

  tensorflower-gardener

  committed
  e387177

  Copy full SHA for e387177

• Roll back previous change "Cleans up wrong callers of FindNonTrivialHero."

  Show description for d44eb61
  

  haoyuz

  authored and

  tensorflower-gardener

  committed
  d44eb61

  Copy full SHA for d44eb61

• Accelerate and optimize MergeShardingIfCompatible.

  Show description for ecf1cd3
  

  ZixuanJiang

  authored and

  tensorflower-gardener

  committed
  ecf1cd3

  Copy full SHA for ecf1cd3

• Update `xspace_to_tools_data_from_byte_string` to accept options parameter to help support tensorboard.gcp profiler.

  Show description for bdbd288
  tensorflower-gardener

  committed
  bdbd288

  Copy full SHA for bdbd288

• #tf-data improve rebatch throughput by ~2.75x (when the data are movable) via avoiding copying

  Show description for 1b7a667
  

  jimlinntu

  authored and

  tensorflower-gardener

  committed
  1b7a667

  Copy full SHA for 1b7a667

• [xla:gpu] Add support for HLO-only BuildInitializerThunk

  Show description for eba17b4
  

  tyb0807

  authored and

  tensorflower-gardener

  committed
  eba17b4

  Copy full SHA for eba17b4

• Early exit shape inference for `XlaCallModule` if possible

  Show description for 98525a1
  

  junwhanahn

  authored and

  tensorflower-gardener

  committed
  98525a1

  Copy full SHA for 98525a1

• [XLA] Require that allocations use the same offset for their aliased required assignment offset at use

  Show description for c1fba99
  

  berkinilbeyi

  authored and

  tensorflower-gardener

  committed
  c1fba99

  Copy full SHA for c1fba99

• Consider `scratch_bytes` for top autotuning results whose times are within an error margin.

  Show description for 264d3ed
  

  dimitar-asenov

  authored and

  tensorflower-gardener

  committed
  264d3ed

  Copy full SHA for 264d3ed

• PR #7843: [GPU] NCCL User Buffer registration (1/3)

  Show description for a43114f
  

  trevor-m

  authored and

  tensorflower-gardener

  committed
  a43114f

  Copy full SHA for a43114f

• Import openai/triton from GitHub.

  Show description for 0edd03b
  

  Moerafaat

  authored and

  tensorflower-gardener

  committed
  0edd03b

  Copy full SHA for 0edd03b

• [xla:gpu] Handle HLO-only case in BuildConstantInitializerThunk

  Show description for cd79fb1
  

  tyb0807

  authored and

  tensorflower-gardener

  committed
  cd79fb1

  Copy full SHA for cd79fb1

• [xla:gpu][NFC] Remove unnecessary static helper for BuildConstantInitializerThunk

  Show description for fb15a8c
  

  tyb0807

  authored and

  tensorflower-gardener

  committed
  fb15a8c

  Copy full SHA for fb15a8c

• Fix broken build.

  Show description for ed542c2
  

  dimitar-asenov

  authored and

  tensorflower-gardener

  committed
  ed542c2

  Copy full SHA for ed542c2

• Fix Bazel test directory path in macos_x86 CI config

  Show description for c72fc7b
  

  nitins17

  authored and

  tensorflower-gardener

  committed
  c72fc7b

  Copy full SHA for c72fc7b

Pagination

Previous

Next