Installation

Install one ONNX Runtime backend appropriate for the target platform.

Platform	Command
CPU	pip install "redimnet2-onnx[cpu]"
NVIDIA CUDA	pip install "redimnet2-onnx[cuda]"
Intel OpenVINO	pip install "redimnet2-onnx[openvino]"
Windows DirectML	pip install "redimnet2-onnx[directml]"
Qualcomm QNN	pip install "redimnet2-onnx[qnn]"

Add the waveform extra when raw-waveform preprocessing is needed:

pip install "redimnet2-onnx[cuda,waveform]"

The same combination supports the optimized TensorRT runtime when compatible CUDA and TensorRT libraries are available on the host.

TensorRT version

Set TENSORRT_VERSION when the host does not expose TensorRT package metadata.

Runtime dependencies

Installing redimnet2-onnx without an extra installs neither ONNX Runtime nor PyTorch.