Flow-Guided-Feature-Aggregation icon indicating copy to clipboard operation
Flow-Guided-Feature-Aggregation copied to clipboard

Published 20 hours ago • verified publisher icon msracver

Why rescale_grad is set to 1.0 in train_end2end.py?

Open kindloaf opened this issue 6 years ago • 0 comments

It seems to me the rescale_grad should be set to 1/batch_size. In case of 4 GPUs training rescale_grad should be 0.25. Why is the value set to "1.0" in train_end2end.py?

kindloaf avatar Jan 03 '19 08:01 kindloaf