Conversion of ONNX files to TRT for TensorRT?
#1
by
gaudi
- opened
Hello,
Does anyone know how someone can take the directory of ONNX files found for this model and create a .trt file that TensorRT can read? The bot from the source repo says the ONNX files can be consumed by backends like TensorRT; however, I'm unsure how to proceed when all the examples I've found have a single onnx file for the model (encoder and decoder are not split out into separate onnx files).
Has anyone else already done this conversion for different GPUs and have an available set of trt files that they can share? This would be a helpful file to have on HF as well.