search

LIAGM / LFattNet

Attention-based View Selection Networks for Light-field Disparity Estimation
MIT License
54 stars 15 forks source link

Questions about training #9

Closed toughaaa closed 3 years ago

toughaaa commented 3 years ago

Dear LIAGM, I have some questions about the training code：

  1. In the code, the epoch is set to 10000000. How long does each epoch takes, and how many epochs need to be trained in total?
  2. I fit the environment as your describe, but an OOM error occurred. My GPU memory is 11G. Finally, I adjusted the number of 3D convolution kernels from 150 to 100 to run successfully. Looking forward to your reply. Thanks.
LIAGM commented 3 years ago

Hi,

  1. Under the 1080TI GPU, each epoch may take about 4 hours and usually I will see whether the losses are converged or not, roughly it may take 30~40 epochs for training.
  2. Due to the network architecture, the memory consumption is huge and if your GPU also have to display for the screen, the OOM error may occur under 11G GPU.
toughaaa commented 3 years ago

@LIAGM Thanks again for your reply.