mmdetection to nvidia triton inference server, best way to do it?

[Grushus]

Hello! Just wondering what the best way to get the model after training to nvidia triton inference server?

The model is in the .pth format which isnt supported by triton, I've tried converting to .pt with a simple torch.load(.pth) torch.save(.pt)

Ive researched regarding mmdetection pytorch2onnx but mmdetection doesnt support it anymore and i havent gotten mmdeploy to work. There are no docs as far as I can tell regarding mmdeploy pytorch2onnx. Kinda stuck atm and would appreciate any help or pointers in the right direction, feels like ive explored most options I can think of myself.