Release Notes

Version Mapping

Product Information

Product Name	Kunpeng BoostKit
Product Version	26.1.RC1
Software Name	Kunpeng TensorFlow Inference Optimization
Software Version	v1.1.0

OS, Compiler, and CPU

OS	CPU	Compiler
openEuler 22.03 LTS SP3	New Kunpeng 920 processor model	GCC 12.3.1
openEuler 24.03 LTS SP3	Kunpeng 950 processor	GCC 12.3.1

Virus Scan Results

Software package release and the following virus scanning are not involved.

Important Notes

Precautions

For details, see Installation Guide.

v1.1.0

Change Description

• Added the TensorFlow ANNC static graph fusion feature, adapted to Kunpeng 950 processor and supporting static graph fusion of operators such as KPFusedEmbeddingActionIdGather.
• Added the TensorFlow ANNC constant folding optimization, adapted to Kunpeng 950 processor.

Modified Features

None

Removed Features

None

Resolved Issues

None

Known Issues

None

v1.0.0

Change Description

Added the TensorFlow KDNN thread passthrough feature, supporting operators such as batchmatmul, concat, and softmax.

Modified Features

None

Removed Features

None

Resolved Issues

None

Known Issues

None

25.2.RC1

Change Description

New Features

Algorithm	Update Description
TensorFlow graph fusion	Added the graph fusion and graph rewriting functions at the TensorFlow model level.
XLA operator fusion	Added the ANNC XLA graph fusion feature.
Operator optimization	Added the ANNC operator optimization feature.

Modified Features

None

Removed Features

None

Resolved Issues

None

Known Issues

None

25.1.RC1

Change Description

New Features

Algorithm Name	Update Description
Batch operator scheduling	Added the operator scheduling optimization and XLA thread pool management optimization features.
Thread affinity isolation	Added the thread affinity isolation feature.

Modified Features

None

Removed Features

None

Resolved Issues

None

Known Issues

None

Related Documentation

v1.1.0 Related Documentation

Document Name	Overview	Delivery Method
Release Notes	Provides the version release information about Kunpeng TensorFlow.	Open-source repository
Feature Introduction	Describes how to use Kunpeng TensorFlow optimization features.	Open-source repository
Installation Guide	Describes how to install and use Kunpeng TensorFlow.	Open-source repository
Quick Start	Provides a quick start guide to the Kunpeng TensorFlow inference optimization features.	Open-source repository
API Reference	Describes how to use the APIs of the Kunpeng TensorFlow inference optimization features.	Open-source repository

Obtaining Documentation

Visit the open-source repository to view or download related documents.