AMD UIF 1.2 Released With Initial Radeon GPU Support

AMD today published version 1.2 of their Unified Inference Frontend (UIF) that is for supporting deep learning inference across AMD CPUs, GPUs, as well as Versal adaptive SoCs and FPGAs. AMD UIF aims to accelerate AI inference across all AMD compute platforms with machine learning frameworks like PyTorch and TensorFlow. With UIF 1.2 adds initial support for Radeon GPUs.

The AMD Unified Inference Frontend has been supporting AMD Instinct accelerators with previous releases while UIF 1.2 marks the point of adding "support for AMD Radeon GPUs." This initially got me excited seeing them extending their unified inference solution to Radeon GPUs, but when reading further in the release notes it mentions: "Support for AMD Radeon™ PRO V620 and W6800 GPUs."

Thus this initial release of Radeon GPU support is rather sad... Just Radeon PRO and not the consumer parts, but even more is that it's for older RDNA2 hardware and not the newest RDNA3 parts like the Radeon PRO W7500/W7600. Hopefully upcoming AMD UIF updates will further iron out this Radeon GPU support so that it's more robust, on a similar note to AMD bringing ROCm to more RDNA3 GPUs later this year.

AMD UIF 1.2 also integrates rocAL support, the ROCm Augmentation Library. AMD rocAL is for setting up hybrid pipelines to help maximize throughput for machine learning applications. Leveraging rocAL can lead to more efficient image/video processing and more. The rocAL library works for both AMD's CPUs and GPUs.

The AMD UIF 1.2 release also has more models enabled for AMD CPUs and adding some new models for AMD GPUs as part of the AMD Model Zoo. There are also ROCm updates and improvements to the AMD Inference Server.

AMD UIF 1.2 can be downloaded via GitHub.