xingyizhou / ExtremeNet

Bottom-up Object Detection by Grouping Extreme and Center Points
BSD 3-Clause "New" or "Revised" License
1.03k stars 174 forks source link

How to train extremenet with multi GPUs, such as 4*1080Ti. #7

Closed Splendon closed 5 years ago

xingyizhou commented 5 years ago

Hi, You can change the batch size here https://github.com/xingyizhou/ExtremeNet/blob/master/config/ExtremeNet.json#L4 and images per batch here https://github.com/xingyizhou/ExtremeNet/blob/master/config/ExtremeNet.json#L22 (and make sure the sum of chunk_sizes equal to batch_size) to fit your GPUs.