I am looking to do some reproducible benchmarks on number of hardware to measure framework performance. Are there any benchmarking scripts that the community uses / recommends to measure peformance of Pytorch models from the model hub on varied hardware, similar to the TensorFlow benchmarks[1]