Late last year, AMD announced ROCM 5The Radeon RX 7900 XTX, Radeon RX 7900 XT, and Radeon Pro W7900 graphics cards are already supported by 7 and PyTorch, covering the top of the range with RDNA 3 architecture, allowing developers to choose the solution that best suits their needs.
Recently, AMD announced the launch of ROCM 60, further expanding the scope of support to support Radeon Pro W7800 and RX 7900 GRE graphics cards, meaning that all desktop and workstation graphics cards with N**i 31 GPUs are fully supported. By supporting a broad portfolio of products, AMD hopes to help the artificial intelligence (AI) community acquire more desktop graphics cards at different price points and performance levels for better development efforts. AMD also complements its solution stack with support for the Onnx Runtime. Onnx, short for Open Neural Network Exchange, is an intermediate machine learning framework that serves as an open standard for machine learning algorithms and software tools for converting AI models between different ML frameworks.
PyTorch is a deep learning library developed by Facebook AI Research as an open-source tool that has been welcomed by the industry and is widely supported by CPU and GPU manufacturers. AMD was initially slow to integrate PyTorch support, especially for consumer Radeon and Radeon Pro graphics cards, but this has shifted significantly over the past year. In addition, the priority to support flagship graphics cards is also due to the need for large capacity video memory for AI applications such as large language models (LLMs).