Microsoft has announced it is open sourcing its extension of TensorFlow for Windows. The source code for TensorFlow-DirectML is now available on GitHub.
“TensorFlow-DirectML broadens the reach of TensorFlow beyond its traditional Graphics Processing Unit (GPU) support, by enabling high-performance training and inferencing of machine learning models on any Windows devices with a DirectX 12-capable GPU through DirectML, a hardware accelerated deep learning API on Windows,” wrote Microsoft’s senior program manager Clarke Rahrig, senior software engineer Justin Stoecker, and principal software engineer lead Chai Chaoweeraprasit.
TensorFlow-DirectML supports native Win32 and Windows Subsystem for Linux, and is suitable for students, beginners and enthusiasts looking to accelerate model training and prediction on DirectX 12 GPU machines.
According to the team, TensorFlow is widely used for machine learning models, but takes a tremendous amount of computation. While TensorFlow leverages both Central Processing Units and GPUs, GPU acceleration is limited to vendor-specific platforms. “Bringing the full machine learning training capability to Windows, on any GPU, has been a popular request from the Windows developer community,” the team wrote.
DirectML is designed to extend the platform with high-performance implementations of mathematical operations. “We’re bringing high-performance training and inferencing on the breadth of Windows hardware by leveraging DirectML in the TensorFlow framework. Not only does this extend TensorFlow’s GPU reach on Windows, but it also applies to the Windows Subsystem for Linux (WSL),” the team wrote.
The first developer preview of the solution was released in June. By moving the project to GitHub, the team hopes to engage with the community and focus on issues users want the most.
