pytorch-loss icon indicating copy to clipboard operation
pytorch-loss copied to clipboard

Published 20 hours ago • verified publisher icon CoinCheung

The probability caculation of negative class in 'Label smooth.py'

Open kaijieshi7 opened this issue 3 years ago • 0 comments

'lb_neg' should be "lb_smooth/(numclass -1)" instead of " lb_smooth/numclass" .

kaijieshi7 avatar Feb 24 '21 05:02 kaijieshi7