I have a custom pretrained pytorch model and I need to convert it into a .wts file to run a library on it but I am lost as to how to accomplish this.
Details: I used YOLOv5 to train a model and save it as a .pt file. I need to run this model on Jetson Nano and the library I am using takes in a .wts file extension. I found only one website referring to this conversion here: https://github.com/wang-xinyu/tensorrtx/blob/master/tutorials/getting_started.md however, the README does not explain how to actually accomplish this.
If not this, my main objective is to efficiently run inference using a pretrained pytorch model on a 2GB Jetson Nano!
Things I have tried:
- I used YOLOv5 to convert the .pt model into a ONNX model.I am not sure how to convert the ONNX model into a .wts file.
- I tried using trtexec to convert the file into something TensorRT can use but that converts it into a .trt file.
- I tried using pytorch2trt but again that is accomplishing something else
