Fast Technology reported on February 21 that AMD officially announced that the new ROCM 6The 0 development platform was officially launched, continuing to support new GPU graphics card products and AI development frameworks, and expanding client-based AI artificial intelligence and ML machine learning development from both software and hardware levels.
Previously released ROCM 57 Added support for the RX 7900 XT, RX 7900 XTX, Radeon Pro W7900 graphics cards, including the PyTorch framework.
Today's ROCM 60 has welcomed two new members of the RDNA 3 graphics card family: RX 7900 GRE, Radeon Pro W7800.
Engineers can use them for AI and ML development, which further expands the scope of the platform, lowers the barrier to entry, and provides developers with more choices in terms of performance and price.
In addition, ROCM 60 also supports the Onnx Runtime, an intermediate ML framework for converting AI models between different ML frameworks.
As a result, developers can perform inference on native AMD hardware with a wider range of source data, as well as process INT8 integer data with AMD's own graphical inference engine, MigraphX, which previously supported FP16 and FP32 floating-point data formats.
Of course, the Pytorch framework continues to be supported for manipulating FP16 FP32 mixed-precision data.
With ROCM 6The official release of 0 also indicates that AMD is accelerating and deepening the construction of the ROCM development platform, supporting more and more graphics card hardware and framework models, allowing developers to engage in AI and ML research more freely, and also providing the same powerful platform choices in addition to NVIDIA CUDA and Intel OpenAPI.