AMD Expands ML Development Capabilities with ROCm 6.0 and Extended GPU Support 

Feb. 15, 2024 -- AMD has announced the expansion of its machine learning (ML) development offerings with the introduction of AMD ROCm 6.0, building upon the previously announced support for the AMD Radeon RX 7900 XT, XTX, Radeon PRO W7900 GPUs, and PyTorch integration.

Credit: Shutterstock

Key advancements include:

• Enhanced GPU Support: The AI research and ML engineering community can now leverage the Radeon PRO W7800 and Radeon RX 7900 GRE GPUs for development. This expansion broadens the availability of desktop graphics cards across various price points and performance levels, reaffirming AMD's commitment to facilitating AI research and development.
• ONNX Runtime Integration: AMD's solution stack now includes support for ONNX (Open Neural Network Exchange) Runtime. ONNX serves as an intermediary ML framework, enabling the conversion of AI models between different ML frameworks. This integration allows users to perform inference on a wider range of source data using local AMD hardware. Additionally, AMD introduces support for INT8 data type through MIGraphX, AMD's graph inference engine, complementing the existing data types like FP32 and FP16.
• Continued PyTorch Framework Support: With ROCm 6.0, AMD continues to support the PyTorch framework by introducing mixed precision capabilities (FP32/FP16) in ML training workflows.

AMD's ongoing enhancements in both hardware and software domains are aimed at empowering AI researchers and ML engineers with more versatile and powerful tools for developing cutting-edge ML applications.

For more information from AMD, click here.

About AMD

For more than 50 years AMD has driven innovation in high-performance computing, graphics and visualization technologies. Billions of people, leading Fortune 500 businesses and cutting-edge scientific research institutions around the world rely on AMD technology daily to improve how they live, work and play. AMD employees are focused on building leadership high-performance and adaptive products that push the boundaries of what is possible.

Source: David Diederichs, AMD