Commits

Branch selector

 master

User selector

All users

Datepicker

All time

Commit History

Commits on Oct 25, 2024

• Explicitly tell the GpuCompiler which stream to use from PJRT during the build step.

  Show description for 58dc89e
  

  klucke

  authored and

  tensorflower-gardener

  committed
  58dc89e

  Copy full SHA for 58dc89e

• [Triton] Allow reorderValues to handle downcast with dot_op layout on 16-bit -> 8-bit in the same way it handles 8-bit -> 16-bit. We already needed to do something similar for 16/32 bits previously.

  Show description for 3fc4b0f
  

  Moerafaat

  authored and

  tensorflower-gardener

  committed
  3fc4b0f

  Copy full SHA for 3fc4b0f

• [XLA:GPU] Fix includes in AllGatherDynamicSliceSimplifier

  Show description for d65152d
  

  golechwierowicz

  authored and

  tensorflower-gardener

  committed
  d65152d

  Copy full SHA for d65152d

• [XLA:GPU] Update comment for AllGatherOptimizer.

  Show description for 8331c81
  

  golechwierowicz

  authored and

  tensorflower-gardener

  committed
  8331c81

  Copy full SHA for 8331c81

• Integrate LLVM at llvm/llvm-project@33363521ca24

  Show description for af5a4d3
  

  krasimirgg

  authored and

  tensorflower-gardener

  committed
  af5a4d3

  Copy full SHA for af5a4d3

• Fix a use-after-free bug in third_party/tensorflow/compiler/mlir/quantization/common/attrs_and_constraints_test.cc

  Show description for cd53e4f
  tensorflower-gardener

  committed
  cd53e4f

  Copy full SHA for cd53e4f

• compat: Update forward compatibility horizon to 2024-10-25

  Show description for 417c495
  tensorflower-gardener

  committed
  417c495

  Copy full SHA for 417c495

• [XLA:GPU] Add a pass to ensure that dots have contracting dimensions.

  Show description for fc34e27
  

  akuegel

  authored and

  tensorflower-gardener

  committed
  fc34e27

  Copy full SHA for fc34e27

• Automated Code Change

  Show description for 76bd181
  tensorflower-gardener

  committed
  76bd181

  Copy full SHA for 76bd181

• [XLA:GPU] Remove untested `int4` path in the Triton fusion emitter.

  Show description for d1712ef
  

  bchetioui

  authored and

  tensorflower-gardener

  committed
  d1712ef

  Copy full SHA for d1712ef

• Update GraphDef version to 2026.

  Show description for 1395e62
  tensorflower-gardener

  committed
  1395e62

  Copy full SHA for 1395e62

• [PJRT:Python] Pass key value store to XLA compilation options.

  Show description for a2d47b0
  tensorflower-gardener

  committed
  a2d47b0

  Copy full SHA for a2d47b0

• PR #17259: Adding Strictness level to PGLE accuracy checker.

  Show description for 7e5e11c
  

  shraiysh

  authored and

  tensorflower-gardener

  committed
  7e5e11c

  Copy full SHA for 7e5e11c

• Automated Code Change

  Show description for c2a246f
  tensorflower-gardener

  committed
  c2a246f

  Copy full SHA for c2a246f

• [XLA:GPU] Deprecate xla_gpu_enable_priority_fusion flag.

  Show description for 7577c10
  

  olegshyshkov

  authored and

  tensorflower-gardener

  committed
  7577c10

  Copy full SHA for 7577c10

• Automated Code Change

  Show description for 3061518
  tensorflower-gardener

  committed
  3061518

  Copy full SHA for 3061518

• Automated Code Change

  Show description for 619e8fa
  tensorflower-gardener

  committed
  619e8fa

  Copy full SHA for 619e8fa

• [SPMD] Always colocate the `XlaSharding` metadata op with the `ReadVariableOp`.

  Show description for d91061b
  

  mrry

  authored and

  tensorflower-gardener

  committed
  d91061b

  Copy full SHA for d91061b

• Automated Code Change

  Show description for 4d82442
  tensorflower-gardener

  committed
  4d82442

  Copy full SHA for 4d82442

• Reverts 91bb94cdfa6c22dcb6a4ce460717b9450307e561

  Show description for 18c0dc8
  tensorflower-gardener

  committed
  18c0dc8

  Copy full SHA for 18c0dc8

• Check op_kernel is not null

  Show description for 9c3de1b
  

  deqiangc

  authored and

  tensorflower-gardener

  committed
  9c3de1b

  Copy full SHA for 9c3de1b

• Fix Reshape on BMM output pattern.

  Show description for eba9747
  

  vamsimanchala

  authored and

  tensorflower-gardener

  committed
  eba9747

  Copy full SHA for eba9747

• Add slice Op legalization to QC compiler plugin.

  Show description for 85f28a2
  tensorflower-gardener

  committed
  85f28a2

  Copy full SHA for 85f28a2

Commits on Oct 24, 2024

• Updates the `BasicStringArray` class to use `absl::Cord` as the element type.

  Show description for b40a341
  tensorflower-gardener

  committed
  b40a341

  Copy full SHA for b40a341

• PR #102: Update tflite/README.md

  Show description for 1a69fc0
  

  junjiang-lab

  authored and

  tensorflower-gardener

  committed
  1a69fc0

  Copy full SHA for 1a69fc0

• Pass also non-string values as Dispatch API and Dispatch Delegate options

  Show description for 3888f99
  tensorflower-gardener

  committed
  3888f99

  Copy full SHA for 3888f99

• Add a TODO to return a valid platform version for `TfrtCpuClient`.

  Show description for 7765727
  

  yangustc07

  authored and

  tensorflower-gardener

  committed
  7765727

  Copy full SHA for 7765727

• Temporarily remove stablehlo_current_version from PJRT_GetPluginCAttributes.

  Show description for 45d15fb
  

  GleasonK

  authored and

  tensorflower-gardener

  committed
  45d15fb

  Copy full SHA for 45d15fb

• [XLA] Small speedups to layout assignment.

  Show description for 86fb750
  

  hawkinsp

  authored and

  tensorflower-gardener

  committed
  86fb750

  Copy full SHA for 86fb750

• Fix nightly TF builds.

  Show description for 0a2637b
  tensorflower-gardener

  committed
  0a2637b

  Copy full SHA for 0a2637b

• [IFRT] Add IFRT IR passes for compiling atom programs.

  Show description for 1f8791e
  

  ICGog

  authored and

  tensorflower-gardener

  committed
  1f8791e

  Copy full SHA for 1f8791e

• Migrate mlir_graph_optimization_pass to use ConvertGraphToTfExecutor instead of ConvertGraphToMlir.

  Show description for 59b9cd8
  

  rocketas

  authored and

  tensorflower-gardener

  committed
  59b9cd8

  Copy full SHA for 59b9cd8

• [tflite] Apply the default delegates when using the signature runner.

  Show description for e8ab4a4
  

  qukhan

  authored and

  tensorflower-gardener

  committed
  e8ab4a4

  Copy full SHA for e8ab4a4

• 1. Fix asan, msan race by invoking Stop() within PassBarrier() as opposed to repeatedly within done callbacks.

  Show description for 6665b35
  tensorflower-gardener

  committed
  6665b35

  Copy full SHA for 6665b35

• Migrate import_model to use ConvertGraphToTfExecutor instead of ConvertGraphToMlir.

  Show description for 82572cd
  

  rocketas

  authored and

  tensorflower-gardener

  committed
  82572cd

  Copy full SHA for 82572cd

Pagination

Previous

Next