search

Amshaker / unetr_plus_plus

[IEEE TMI-2024] UNETR++: Delving into Efficient and Accurate 3D Medical Image Segmentation
Apache License 2.0
340 stars 32 forks source link

utilization of GPU is lower #67

Open QianLingjun opened 8 months ago

QianLingjun commented 8 months ago

When I increased the training data to 1000 CT cases, the training of 4 12G 1080ti was normal, but the utilization of GPU would intermittently all be 0. The end result is more time consuming.

How can I adjust the code to increase the utilization of the GPU? thank you.

Amshaker commented 7 months ago

Hi @QianLingjun ,

Check if your data loading and preprocessing steps are optimized. Inefficient data handling can lead to bottleneck issues. Also, experiment with different batch sizes and check the GPU utilization. Sometimes, using a smaller or larger batch size can impact GPU utilization.