Open insinfo opened 1 year ago
This repo (mdx-net) contains only the code for training the models, this issue might fit better in the repo that contains the models.
If using the repo that contains the models, it defaults to using CPU.
In some versions of onnxruntime it might throw an error to specify the provider to use in predict_blend.py
ort.InferenceSession()
. In this case you are using only the CPU to separate the tracks, so it should be edited to
_ort = ort.InferenceSession(f'{onnx_name}/{model.target_name}.onnx', providers=['CPUExecutionProvider'])
As with OpenVINO, I don't think the code have used that toolkit.
It is possible to separate a music track with this model on a computer with an Intel CPU without an Nvidia GPU, with or without OpenVINO