Mayank Daga, Director of Deep Learning software, AMD
Source | TensorFlow public number
We’re excited to announce the launch of TensorFlow V1.8 for ROCm GpuS, which includes the Radeon Instinct MI25. This is a significant milestone for AMD’s ongoing deep learning acceleration efforts. ROCm, or Radeon Open Ecosystem, is our open-source software foundation for GPU computing on Linux. Our TensorFlow implementation uses MIOpen, a highly optimized GPU routine library for deep learning.
AMD provides a pre-built WHL package with a simple installation process similar to installing Linux general TensorFlow. We have released the installation instructions and pre-built Docker images.
In addition to supporting TensorFlow V1.8, we are working on adding all ROCm enhancements to the TensorFlow main storage. Some of these patches have been consolidated upstream and several others are under active review. As we strive to fully incorporate enhancements, we will also release and maintain future rocM-supported versions of TensorFlow, such as V1.10.
We believe that the future of deep learning optimization, portability, and scalability is rooted in domain-specific compilers. Inspired by the early results of XLA, we are also working on enabling and optimizing XLA for AMD Gpus.
For more information on AMD’s work in this area, please visit www.amd.com/deeplearnin…