search

potterhsu / easy-faster-rcnn.pytorch

An easy implementation of Faster R-CNN (https://arxiv.org/pdf/1506.01497.pdf) in PyTorch.
MIT License
165 stars 57 forks source link

About the learning rate #26

Open czs1886 opened 4 years ago

czs1886 commented 4 years ago

For the WarmUpMultiStepLR, if we maintain the default factor and warm up steps, the learning rate will tend to be 0 after several steps since the lr is mulplicated by factor for number of steps times.

Rajat-Mehta commented 4 years ago

check this #7