Sudden spike in validation loss with regularization

I’ve trained a image segmentation model using BCE loss and adam optimizer.
On this run, I just changed the weight_decay value from 0 to 1e-05 & every other things is same as first run. But I see in the validation curve there is a big spike in the middle


left plot corresponding first run and right for second run(having weight decay)

What could be the reason of this spike?