I implemented a custom PyTorch Dataset for my project's Dataloader to use. However, it is running slower than expected so profiling was chosen to troubleshoot the bottleneck.
Looked into vprof but I am unsure of the function to be profiled if I am interested in only profiling the Dataset implementation. Using PyTorch 1.5 with Python 3.7 on Ubuntu 18.04.
How can we perform performance profiling (CPU and memory) on just the custom Dataset implementation?
Thanks!
