Hi, thanks for the interesting work and for releasing the code.
I noticed that Tables 1–4 report parameters, MACs, and accuracy, but not throughput or inference latency.
Was throughput left out because it was outside the scope of the hypothesis, or because it was hard to measure the baselines fairly under a common setup?
Also, do you happen to have any internal throughput numbers (e.g. images/s on a single GPU at 224x224) that you could share? It would be very helpful as a reference.
Thanks!
Hi, thanks for the interesting work and for releasing the code.
I noticed that Tables 1–4 report parameters, MACs, and accuracy, but not throughput or inference latency.
Was throughput left out because it was outside the scope of the hypothesis, or because it was hard to measure the baselines fairly under a common setup?
Also, do you happen to have any internal throughput numbers (e.g. images/s on a single GPU at 224x224) that you could share? It would be very helpful as a reference.
Thanks!