What can we expect regarding Torch Vs Torch Lightning Vs TensorRT Inference Speed.
I have converted the PTL (segmentation_models.pytorch) to onnx.
How do I convert it to TensorRT for accelerated inference and what speed up can I expect e.g. 1.5x or 5x faster with TRT.
Is Lightning already faster than vanilla torch?
We have 1M images to run the inference pipeline on.