intel/intel-extension-for-tensorflow v2.15.0.2

Intel® Extension for TensorFlow* 2.15.0.2

on GitHub

Features and Improvements

Intel® Extension for TensorFlow* extends the official TensorFlow capabilities, allowing TensorFlow workloads to run on Intel® Data Center GPU Max Series. This release includes the following features and improvements:

• Toolkit Support: Supports Intel® oneAPI Base Toolkit (version 2025.0.1)

• Updated Support: The Intel® Extension for TensorFlow* has been upgraded to support oneDNN v3.6.2

Bug Fixes

• Fixes page fault issues in TruncatedMod and UniqueOp discovered by stricter checks for memory access.
• Fixes missing env_check.py under the Intel® Extension for TensorFlow* library path when installing previously released xpu wheels.
• Fixes build issues with GCC14.
• Fixes build issues related to oneDNN upgrade.
• Fixes issues related to oneAPI DPC++ Compiler upgrade
• Fixes build and linking issues originating from crosstool_wrapper_driver.tpl

What's Changed

• XeTLA patch is updated to support the latest Intel® oneAPI DPC++ Compiler
• SYCL Shuffle API has been updated
• traceme_encode.h is updated to avoid potential compile warnings
• Code is refactored to avoid potential nested queue submit issues.

Documentation

• Welcome to Intel® Extension for TensorFlow* documentation
• TensorFlow Serving Installation Guide
• Distributed support: Intel® Optimization for Horovod*
• Intel® Extension for TensorFlow* Installation guide
• Frequently Asked Questions