benchmarks
benchmarks copied to clipboard
how to use chrome://tracing/ to analyse performance bottleneck with --trace_file flas.
I am testing on AlexNet with 8 V100 PCIE GPUs using real imagenet dataset. The Result for one card is about 3700. two cards is 6400. three cards is 6000 and 8 cards is 3800. I want to use --trace_file to analyse the bottleneck, but don't know how to. I can see many pids and tids in chrome://tracing/. Any one has a tutorial?