How to accelerate regseg on tensorRT about super-gradients HOT 3 CLOSED

Hi,
So we are using the Deci platform in order to compile and optimize the runtime performance. You choose there the target HW that you will use, and if it's Nvidia's HW it will compile it using trt.
You can try using the platform and load there your onnx model - https://console.deci.ai/
we can also assist you with deployment issues via Deci community slack channel - https://join.slack.com/t/decicommunity/shared_invite/zt-1b0pfclld-SEyaCUQxafr~7LWuPM9H~Q

from super-gradients.

Since RegSeg's original repository saved model as zip file, how did you convert it to onnx?

from super-gradients.

We are converting torch files into onnx - you can do it on the pre-trained checkpoints with torch.onnx.export() fonction
we wrote a blog about it:
https://deci.ai/blog/how-to-convert-a-pytorch-model-to-onnx/

from super-gradients.