Hi. I have a C++ process with multiple models, for which I use multiple threads. I want each model to run in a separate core and set affinity to reflect that. In ONNX, I know there’s a way to set affinity for the particular session. However, switching to ONNX would involve a lot of overhead for me. I was hoping to be able to select affinity in torch for each model. Is there a way to do this? So far I see thread and affinity in torch is global and that seems like a problem