How to prune and convert a customed .pt to .engine?

Jetson NX 16G
not enough space for deepstream, 2G emmc remaining
yolov5_6.1 customed via ghostConv, CABlock


For pruning, you can check if there is any pruning tool in the PyTorch.
Or you can check our TAO toolkit but it needs to use the supported model and retraining.

For TensorRT, please generate an ONNX format first.
This can be done with this sample on a desktop environment.

Then copy the model to the Jetson and run the following commands:

$ /usr/src/tensorrt/bin/trtexec --onnx=[your/model]


Thank you much.

This topic was automatically closed 14 days after the last reply. New replies are no longer allowed.