Home » Oreps » TensorFlow » All Releases » v2.12.1 Release
tensorflow/tensorflow
An Open Source Machine Learning Framework for Everyone
Watch
6
Commit Activity
Directory Browser
Velocity History
Releasesv2.12.1
Basic Details Issues Addressed Top Contributors Directory Browser Release Notes
TensorFlow v2.12.1

TensorFlow: v2.12.1 Release

Release date:
August 28, 2023
Previous version:
v2.12.0 (released March 20, 2023)
Magnitude:
87,565 Diff Delta
Contributors:
278 total committers
Data confidence:
Commits:
2,904 impacting commits

166 Features Released with v2.12.1

[NVIDIA XLA]Replace broadcast of trivial matrix bias by optional custom-call for CUDA >= 12
#60409
Added Arm NN Settings for Arm NN delegate
#60352
[NVIDIA XLA] Support Conv-Bias-Relu6/LeakyRelu fusion in XLA using cuDNN runtime fusion
#59614
[tosa] fix strided_slice legalization
#59880
Add //tensorflow/tsl/protobuf:protos_all_cc_impl to tf_portable_proto_lib in tensorflow/tsl/platform/default/build_config.bzl
#60304
Update release notes for TensorFlow 2.13.0
#60463
Update the RBE images to the latest container versions
#60260
[ROCm] gpu solvers support
#57102
[Linaro:ARM_CI] Add job to run c++ tests nightly
#60481
Use fp32 compute result as the reference for tf32 test in ROCm as MIOโ€ฆ
#60696
Add handling of TensorRT's new enumeration value `nvinfer1::DataType::kFP8`
#60046
Adding large tensor mlir kernels, batch 1
#60303
r2.13 cherry-pick: 6142f87fb7f "Fix TPUExecute for TPU embedding."
#60865
Lite: Fix CMake Build Issue for Metal Delegate
#60077
skip comparison if input is already a boolean tensor for count_nonzero
#60782
[oneDNN] Enable oneDNN reduced precision floating point arithmetic mode with new TF_SEโ€ฆ
#59305
Update the RBE images to the latest container versions
#60600
[oneDNN v3.x]: Enabled weight caching in convolution
#60475
[oneDNN v3.x]: Support for convolution (fwd + bwd) and transpose
#60307
[NVIDIA TF] Disable TF32 in remapper_test on A100+ GPUs.
#60550
[Tosa] Add legalization of BroadcastTo
#60692
[NextPluggableDevice] Disable alignment check for AsyncValueTensor
#60520
[NextPluggableDevice]fix NextPluggableDevice allocator(use pjrt allocator) and Sync()
#59898
r2.13 cherry-pick: 517d4ac6925 "Expose `sync()` in checkpoint manager."
#60699
Update training information in get_started_low_level.md
#60372
[AMD-ZENDNN] Blocked format support with new env variable
#60818
[NextPluggableDevice] expose necessary c headers
#60496
[INTEL oneDNN][Bug Fix] Fix incorrect use of int for dim size related to oneDNN gemm and matmul op
#60568
[NextPluggableDevice]add pjrt compile on demand control flag
#60422
Minor fixes for RELEASE.md
#60553

Browse Other Releases

Latest Pending Unreleased ๐Ÿ˜Ž
v2.13.1 Released September 12, 2023
12,192 ฮ”
v2.13.0 Released June 28, 2023
8 ฮ”
v2.13.0-rc2 Released June 16, 2023
1,122 ฮ”
v2.13.0-rc1 Released May 26, 2023
74 ฮ”
v2.13.0-rc0 Released June 27, 2023
626 ฮ”
v2.12.1 Released August 28, 2023
87,565 ฮ”
v2.12.0 Released March 20, 2023
100 ฮ”
v2.12.0-rc1 Released May 30, 2023
29,815 ฮ”
v2.12.0-rc0 Released March 16, 2023
1,626 ฮ”
v2.11.1 Released March 16, 2023
58,221 ฮ”

Top Contributors in v2.12.1

tensorflower-gardener
BrianWieder
jam14j
faizan-m
fionalang
cezheng
rjpower
rainwoodman
qlzh727
jblespiau

Directory Browser for v2.12.1

We haven't yet finished calculating and confirming the files and directories changed in this release. Please check back soon.

Release Notes Published

Release 2.12.1

Bug Fixes and Other Changes

  • The use of the ambe config to build and test aarch64 is not needed. The ambe config will be removed in the future. Making cpu_arm64_pip.sh and cpu_arm64_nonpip.sh more similar for easier future maintenance.
GitClear uses cookies to ensure you get the best experience on our website. Learn more