v0.28.1.7

Intel® Optimization for Horovod* v0.28.1.7 Release Notes

Intel® Optimization for Horovod* is the Intel optimized distributed training framework to extend the official Horovod (based on v0.28.1) which aims at providing distributed ability to run TensorFlow workloads on Intel GPU clusters. This release contains the following major features:

• Supports Intel® Deep Learning Essentials (version 2025.0.1).
• Supports TensorFlow 2.15.1 and Intel® Extension for TensorFlow* v2.15.0.3.
• Turns TensorFlow NextPluggableDevice mode by default for Intel Device.
• Supports both scale-up and scale-out on the Intel® Data Center Max GPU cluster
• Fixes potential overflow of displacement arrays for large number of ranks and msg sizes

What's Changed

• Adds support for PyTorch build with pre-C++11 ABI
• Adds CMake build support to compile Horovod's SYCL kernels for _GLIBCXX_USE_CXX11_ABI=0
• Adds checks on return status of MPI_Init_thread
• Fixes build issue for sycl compiler 2025.1 with update to sycl include path

Documentation

• Distributed Training (example using Intel® Extension for TensorFlow*) with Intel® Optimization for Horovod* on Intel® GPU
• Distributed Training (example using Intel® Extension for PyTorch*) with Intel® Optimization for Horovod* on Intel® GPU