TensorFlow-DirectML is an accessible machine learning (ML) tool that has a wide range of applications in hardware design (especially GPU). Microsoft is working with AMD to combine AMD’s RDNA 2 architecture technology on the Windows platform. On the test platform using AMD Radeon RX 6900 XT and RX 6600 XT graphics cards, the inference performance was increased to 3.7 times, and the overall AI Benchmark Alpha score was also increased by 4.4 times.
Microsoft stated on the official blog that it is very happy to launch the first universal TensorFlow-DirectML toolkit, which can be used by both students and professional developers. In order to solve various problems in the world, powerful and easy-to-access machine learning tools are required, and suitable for all types of hardware. From scientists, researchers, professional designers to students who are interested in learning this knowledge, they all hope to have the right platform.
According to Microsoft, by collecting customer feedback, Microsoft found that the users generally want a better Windows PC platform to run machine learning tool solutions. In order to make it easier for Windows users to use machine learning tools, Microsoft launched a preview version that supports GPU-accelerated training work last year and can use DirectML’s machine learning framework in Windows and Windows Linux Subsystem (WSL).
Thanks to all the performance improvements implemented by Microsoft and AMD throughout the TensorFlow-DirectML development process, users can now take the machine learning workflow to a new level when using AMD graphics cards.
This GPU accelerated training is applicable to any GPU that supports DirectX 12 and fully supports AMD’s Radeon/Radeon Pro graphics product line. Users who use the latest AMD Radeon Software Adrenalin driver can immediately use TensorFlow-DirectML.