search

zouchuhang / LayoutNetv2

PyTorch implementation of our IJCV paper: "Manhattan Room Layout Reconstruction from a Single 360 image: A Comparative Study of State-of-the-art Methods"
https://arxiv.org/pdf/1910.04099.pdf
MIT License
221 stars 34 forks source link

the number of epochs #10

Closed godhj93 closed 1 year ago

godhj93 commented 1 year ago

Hi, i am training the network as you described in this repository.

I found that the number of epochs set 99999.

is there any reason to set it 99999?

I am wondering if I can set the number of epochs 100 since it takes too much time to train.

thanks :)

zouchuhang commented 1 year ago

The training script will save the model only if the validation loss achieves the smallest. So you can stop the training job at any time when you feel that the validation loss will never go down.