Backends

The AMD Inference Server supports a number of backends. Each backend has its own sets of features and capabilities, which are summarized at a high-level in the table below. You can see more information about each backend on its own page.

Backend	Hardware	Model Formats	Model Support
CPlusPlus	CPU | GPU | FPGA	.so	✔
MIGraphX	GPU	.mxr | .onnx	✔
PT+ZenDNN	CPU	.pt	⚠
TF+ZenDNN	CPU	.tf	⚠
Vitis AI	FPGA	.xmodel	✔