Commits

Branch selector

 master

User selector

All users

Datepicker

All time

Commit History

Commits on Dec 15, 2022

• Fix half/bfloat16 convert bug.

  Show description for d4ea582
  

  cantonios

  authored and

  tensorflower-gardener

  committed
  d4ea582

  Copy full SHA for d4ea582

• This CL ensures a minimum number of reductions by only scaling the work group size

  Show description for 325de9c
  tensorflower-gardener

  committed
  325de9c

  Copy full SHA for 325de9c

• [XLA:GPU] Add Dnn VersionInfo to GpuTargetConfig

  Show description for e5d2ae4
  

  anlunx

  authored and

  tensorflower-gardener

  committed
  e5d2ae4

  Copy full SHA for e5d2ae4

• Make xla_compile_main.cc work when cuda is not enabled.

  Show description for 4b2df3a
  

  ddunl

  authored and

  tensorflower-gardener

  committed
  4b2df3a

  Copy full SHA for 4b2df3a

• Improve side effect modeling of `CollectiveGatherV2`

  Show description for 04c0fa8
  

  mgester

  authored and

  tensorflower-gardener

  committed
  04c0fa8

  Copy full SHA for 04c0fa8

• Remove _map_resources() from all Trackable-derived classes

  Show description for 1461c25
  

  BlaziusMaximus

  authored and

  tensorflower-gardener

  committed
  1461c25

  Copy full SHA for 1461c25

• [XLA:GPU] Make GpuBfloat16Support compatible with AOT compilation

  Show description for d6f5d78
  

  anlunx

  authored and

  tensorflower-gardener

  committed
  d6f5d78

  Copy full SHA for d6f5d78

• [mhlo] Add type inference for stablehlo.scatter

  Show description for 5133fce
  

  Xin Zhou

  authored and

  tensorflower-gardener

  committed
  5133fce

  Copy full SHA for 5133fce

• Provide PreemptionWatcher API for preemption detection and recovery for Borg TPU with TPUClusterResolver, and export it as `tf.distribute.experimental.PreemptionWatcher`.

  Show description for c9622b2
  

  Nan Wei

  authored and

  tensorflower-gardener

  committed
  c9622b2

  Copy full SHA for c9622b2

• Add support for FP8 in various helper functions and classes.

  Show description for 5d27ce5
  

  reedwm

  authored and

  tensorflower-gardener

  committed
  5d27ce5

  Copy full SHA for 5d27ce5

• (rollforward) Create `Callable` as an allowed return-type for `Trackable._serialize_to_tensors`.

  Show description for ecf5a42
  

  k-w-w

  authored and

  tensorflower-gardener

  committed
  ecf5a42

  Copy full SHA for ecf5a42

• Simplify the evaluator logic for hlo::convolution op

  Show description for 7080a68
  

  sdasgup3

  authored and

  tensorflower-gardener

  committed
  7080a68

  Copy full SHA for 7080a68

• [tfrt] Peel linalg.map and linalg.fill

  Show description for 5fc4109
  tensorflower-gardener

  committed
  5fc4109

  Copy full SHA for 5fc4109

• Fix bounds checks for gather and scatter.

  Show description for 4236743
  

  jreiffers

  authored and

  tensorflower-gardener

  committed
  4236743

  Copy full SHA for 4236743

• Update TFRT dependency to use revision

  Show description for 383d68d
  tensorflower-gardener

  committed
  383d68d

  Copy full SHA for 383d68d

• add tsl/protobuf:dnn_proto_cc_impl to xla_cc_binary

  Show description for fd94993
  

  ddunl

  authored and

  tensorflower-gardener

  committed
  fd94993

  Copy full SHA for fd94993

• Fix a couple typos.

  Show description for 1782c21
  tensorflower-gardener

  committed
  1782c21

  Copy full SHA for 1782c21

• [mhlo] Add type inference for stablehlo.select_and_scatter

  Show description for e3ddbed
  

  Xin Zhou

  authored and

  tensorflower-gardener

  committed
  e3ddbed

  Copy full SHA for e3ddbed

• [NFC] Use absl::flat_hash_map instead of std::map for compiler factories

  Show description for bd003ed
  

  jurahul

  authored and

  tensorflower-gardener

  committed
  bd003ed

  Copy full SHA for bd003ed

• Use std::nullopt instead of llvm::None

  Show description for e8b0009
  tensorflower-gardener

  committed
  e8b0009

  Copy full SHA for e8b0009

• Allow numpy casting between custom types.

  Show description for 11945b5
  

  cantonios

  authored and

  tensorflower-gardener

  committed
  11945b5

  Copy full SHA for 11945b5

• [tfrt] Tile linalg.fill

  Show description for fc9f157
  tensorflower-gardener

  committed
  fc9f157

  Copy full SHA for fc9f157

• Fixing a typo in the image embedder instructions.

  Show description for f7085cb
  tensorflower-gardener

  committed
  f7085cb

  Copy full SHA for f7085cb

• [tfrt] Add vectorization of memref.copy to JitRt pipeline

  Show description for 89a740b
  tensorflower-gardener

  committed
  89a740b

  Copy full SHA for 89a740b

• Merge pull request #58793 from trevor-m:tmorris-bf16-inplace

  Show description for 6c84269
  tensorflower-gardener

  committed
  6c84269

  Copy full SHA for 6c84269

• Merge pull request #57634 from elfringham:aarch64_docker

  Show description for a801b27
  tensorflower-gardener

  committed
  a801b27

  Copy full SHA for a801b27

• Merge pull request #58792 from trevor-m:tmorris-bf16-einsum

  Show description for e479574
  tensorflower-gardener

  committed
  e479574

  Copy full SHA for e479574

• Assorted fixes for StableHLO <=> MHLO roundtrip

  Show description for 9fb197d
  

  Eugene Burmako

  authored and

  tensorflower-gardener

  committed
  9fb197d

  Copy full SHA for 9fb197d

• [XLA:Python] [ifrt] Fix mismatched client and device buffer.

  Show description for d1e8a81
  

  hawkinsp

  authored and

  tensorflower-gardener

  committed
  d1e8a81

  Copy full SHA for d1e8a81

• Merge pull request #58882 from 103yiran:103yiran-lrn

  Show description for 959db87
  tensorflower-gardener

  committed
  959db87

  Copy full SHA for 959db87

• Don't require 'NVIDIA' in PR titles for TF-TRT PRs coming from NVIDIA

  Show description for a6d5ce5
  

  poulsbo

  authored and

  tensorflower-gardener

  committed
  a6d5ce5

  Copy full SHA for a6d5ce5

• Update include dirs for when mlir_hlo is used as a dependency in a different project.

  Show description for 158f85d
  

  ddunl

  authored and

  tensorflower-gardener

  committed
  158f85d

  Copy full SHA for 158f85d

• Avoid a direct dependency from `tensorflow/lite/tools/serialization/option_writer_generator.cc` to `TfLiteNode` so that it can be used with code that works with `TfLiteOpaqueNode`.

  Show description for 1aa2277
  

  MatthiasKreileder

  authored and

  tensorflower-gardener

  committed
  1aa2277

  Copy full SHA for 1aa2277

• [XLA] Minor fixes to the latency-hiding-scheduler.

  Show description for f406d68
  tensorflower-gardener

  committed
  f406d68

  Copy full SHA for f406d68

• [GmlSt] Fuse inputs into linalg.matmul for triton.

  Show description for 5e87799
  

  olegshyshkov

  authored and

  tensorflower-gardener

  committed
  5e87799

  Copy full SHA for 5e87799

Pagination

Previous

Next