zentf Release v5.2

ZenDNN Plugin for TensorFlow (zentf) — Release Notes v5.2

---

Overview

zentf 5.2 is a major release that continues our focus on optimizing inference for Recommender Systems and Large Language Models on AMD EPYC™ CPUs.

---

What's New in zentf 5.2

TensorFlow Version Support

Component	Details
TensorFlow 2.20.0	Primary supported version with optimal performance. Distributed as a Python wheel via PyPI and as a C++ package.
TensorFlow-Java main (75402bef)	Java User Interface — Fully supported (available via source build only).

---

Improvements

1. TF 2.20 Integration

• zentf 5.2 is built for and validated against TensorFlow v2.20.0.
• Bazel 7.4.1 — Upgraded from the Bazel 5.3–6.5 range to a single supported version (7.4.1).
• Because TensorFlow-Java is not released for TensorFlow 2.20.0, zentf supports TensorFlow-Java main (75402bef) via source build only.

2. Migrate from Legacy ZenDNN Library to ZenDNNL

• CMake-based ZenDNNL integration using rules_foreign_cc.
• All operator kernels (MatMul, Conv2D, BatchMatMul, Softmax, Pooling) have been rewritten to use the ZenDNNL Low Overhead API (LOA), replacing the legacy ZenDNN primitives.
• Old third-party dependencies on zen_dnn and amd_blis (BLIS) have been removed, replaced by ZenDNNL with integrated AOCL-DLP.

3. Removed Legacy Components

• Mempool optimization has been completely removed; equivalent performance is achieved using jemalloc as the memory allocator instead.
• INT8 support has been removed.
• Non-performant ops removed — ZenTranspose, ZenReshape, Binary ops.

4. Performance Optimizations

• Enhanced Operations with LOA: Low Overhead API optimizations for improved performance.

---

Breaking Changes

Caution

• Dropped TensorFlow Backward Compatibility: Backward compatibility with previous TensorFlow versions has been discontinued due to major changes in TensorFlow 2.20.0.
• Removed Mempool Support: Dropped support for mempool optimization.
• Dropped INT8 Support: Previously available only for the ResNet50 model; now fully removed.
• Removed Ops: Cleaned up non-performant ops and obsolete fusions — ZenTranspose, ZenReshape, Binary ops, BatchNorm fusions.

zentf Release v5.1

zentf 5.1 is the TensorFlow plugin which comes with the ZenDNN 5.1 release. ZenDNN 5.1 continues to deliver inference performance for deep learning models on AMD EPYC™ CPUs, with a focus on optimizing large scale Recommender Systems. We've introduced several key optimizations to boost the performance of recommender models, such as DIEN.

The zentf 5.1 plugin is optimized to be compatible with TensorFlow 2.19. We've also enabled support for PluggableDevice in TensorFlow-Java. This feature has been officially contributed and upstreamed to the TensorFlow-Java repository, strengthening its core capabilities.

zentf Release v5.0.2

zentf 5.0.2 is the TensorFlow plugin which comes with ZenDNN 5.0.2, which is a minor release building upon the major ZenDNN 5.0 release. This upgrade continues the focus on optimizing inference with Recommender Systems and Large Language Models on AMD EPYC™ CPUs.

zentf 5.0.2 includes enhancements for bfloat16 performance, primarily by leveraging microkernels and operators from the ZenDNN 5.0.2 library. These operators are designed to better leverage the EPYC microarchitecture and cache hierarchy.
The zentf 5.0.2 plugin works seamlessly with TensorFlow versions from the latest 2.18 to 2.16, offering a high-performance experience for deep learning on AMD EPYC™ platforms.

This release adds support for Java interfaces through TensorFlow Java v1.0.0

zentf Release v5.0.1

zentf 5.0.1 is the TensorFlow plugin which comes with ZenDNN 5.0.1, which is a minor release building upon the major ZenDNN 5.0 release. This upgrade continues the focus on optimizing inference with Recommender Systems and Large Language Models on AMD EPYC™ CPUs.

zentf 5.0.1 includes enhancements for bfloat16 performance, primarily by leveraging microkernels and operators from the ZenDNN 5.0.1 library. These operators are designed to better leverage the EPYC microarchitecture and cache hierarchy.

The zentf 5.0.1 plugin works seamlessly with TensorFlow versions from the latest 2.18 to 2.16, offering a high-performance experience for deep learning on AMD EPYC™ platforms.

zentf Release v5.0

This release of AMD's CPU solution for TensorFlow provides a binary built with the Pluggable Device approach.
This zentf release:

• Supports TensorFlow v2.17.
• Integrates with ZenDNN v5.0 as the core inference library and is compiled with GCC v12.2.
• Merged BF16 and FP32 compute flows and added broadcasting support for BatchMatMul kernel.
• INT8 support for the ResNet50 model.
• Softmax kernel now supports up to 5D.
• Deprecated blocked format support for convolution ops and restriction of rewrite for the fused ops based on the post ops.
• Provides experimental support of C++ APIs.

zentf Release v4.2

This zentf release:

• Is compatible with TensorFlow v2.16 and later.
• Extends TensorFlow to provide a performant AI inference solution for AMD EPYC^TM servers leveraging the ZenDNN v4.2 library.
• Includes graph optimizations and fusions tailored for AMD EPYC^TM architectures.
• Supports BF16 execution through auto-mixed precision to provide performance improvements with minimal changes in accuracy.

TensorFlow-ZenDNN Plugin Release v0.2

This is the first open-sourced release for TensorFlow-ZenDNN Plugin.

• This release supports the latest TensorFlow v2.14
• The backend library support in this release is upgraded from ZenDNNv3.3 to the latest released ZenDNN4.1
• This release continues to provide experimental support for leveraging the latest ZenDNN library via TensorFlow's pluggable device approach.