CPU Operators Supported by VART ML Runtime#
When a model is compiled with CPU partition support, NPU-incompatible operators are routed to CPU implementations provided by the AMD Vitis™ AI compiler. The following operators are supported by VART ML Runtime for CPU execution. Each listed operator is supported on AArch64 (ARM).
If a model contains NPU-incompatible operators that are not in this list, it cannot be compiled with CPU partition support and must use standard compilation with ONNX Runtime instead.
Operator |
AArch64 (ARM) |
|---|---|
Add |
Yes |
ArgMax |
Yes |
Cast |
Yes |
Clip |
Yes |
Concat |
Yes |
Conv |
Yes |
DequantizeLinear |
Yes |
Div |
Yes |
Erf |
Yes |
Expand |
Yes |
Flatten |
Yes |
Gather |
Yes |
Gemm |
Yes |
Inverse |
Yes |
Log |
Yes |
MatMul |
Yes |
MaxPool |
Yes |
Mul |
Yes |
NonMaxSuppression |
Yes |
QuantizeLinear |
Yes |
ReduceMax |
Yes |
ReduceMean |
Yes |
Relu |
Yes |
Reshape |
Yes |
Resize |
Yes |
ScatterND |
Yes |
Sigmoid |
Yes |
Slice |
Yes |
Softmax |
Yes |
Split |
Yes |
Squeeze |
Yes |
Sub |
Yes |
Transpose |
Yes |
Unsqueeze |
Yes |
Inverse uses the ONNX contrib schema com.microsoft.Inverse.