search

longcw / faster_rcnn_pytorch

Faster RCNN with PyTorch
MIT License
1.71k stars 466 forks source link

out of memory if don`t fix VGG16 param #87

Open aRookieMan opened 5 years ago

aRookieMan commented 5 years ago

If dont fix pretrained VGG16s parameters , the model will increasing untill memory booms. But if fix these parameters , it works. Why?

code in train.py

for param in net.rpn.features.parameters():
    param.requires_grad = False
SmallSmallQiu commented 5 years ago

谢谢