Radeon Machine Learning
Radeon Machine Learning (Radeon ML or RML) is an AMD SDK for high-performance deep learning inference on GPUs. This library is designed to support any desktop OS and any vendor’s GPU with a single API to simplify the usage of ML inference.
RML is built on DirectML (DirectX12), MiOpen (OpenCL) and MPS (Metal). It uses the native low-level API to achieve the best possible performance as on every OS.
RML provides for interoperability with your application via DirectX/OpenCL/Metal.
Backends
Radeon ML can use one of the following underlying backends for inference:
Microsoft DirectML
AMD MIOpen
Apple MPS
The choice of the backend depends on your operating system:
Backend |
Windows |
Linux |
OSX |
|---|---|---|---|
Microsoft DirectML |
Supported |
Not supported |
Not Supported |
AMD MIOpen |
Supported |
Supported |
Not supported |
Apple MPS |
Not Supported |
Not Supported |
Supported |
System Requirements
[For DirectML] Windows 10 SDK version 1903 or later
[For MIOpen] Pre-built MIOpen libraries/binaries
[For Apple MPS] OSX 10.13.4 or later