Microsoft open sources TensorFlow-DirectML – SD Occasions


Microsoft has introduced it’s open sourcing its extension of TensorFlow for Home windows. The supply code for TensorFlow-DirectML is now accessible on GitHub. 

“TensorFlow-DirectML broadens the attain of TensorFlow past its conventional Graphics Processing Unit (GPU) help, by enabling high-performance coaching and inferencing of machine studying fashions on any Home windows gadgets with a DirectX 12-capable GPU by way of DirectML, a hardware accelerated deep studying API on Home windows,” wrote Microsoft’s senior program supervisor Clarke Rahrig, senior software program engineer Justin Stoecker, and principal software program engineer lead Chai Chaoweeraprasit. 

TensorFlow-DirectML helps native Win32 and Home windows Subsystem for Linux, and is appropriate for college kids, freshmen and fans seeking to speed up mannequin coaching and prediction on DirectX 12 GPU machines.

Based on the staff, TensorFlow is extensively used for machine studying fashions, however takes an incredible quantity of computation. Whereas TensorFlow leverages each Central Processing Items and GPUs, GPU acceleration is proscribed to vendor-specific platforms. “Bringing the complete machine studying coaching functionality to Home windows, on any GPU, has been a preferred request from the Home windows developer group,” the staff wrote. 

DirectML is designed to increase the platform with high-performance implementations of mathematical operations. “We’re bringing high-performance coaching and inferencing on the breadth of Home windows hardware by leveraging DirectML within the TensorFlow framework. Not solely does this prolong TensorFlow’s GPU attain on Home windows, but it surely additionally applies to the Home windows Subsystem for Linux (WSL),” the staff wrote. 

The primary developer preview of the answer was launched in June. By shifting the challenge to GitHub, the staff hopes to have interaction with the group and deal with points customers need essentially the most.