search

cherise215 / Cooperative_Training_and_Latent_Space_Data_Augmentation

[MICCAI 2021 Oral] Cooperative Training and Latent Space Data Augmentation for Robust Medical Image Segmentation
Other
36 stars 6 forks source link

GPU utilization problem #2

Closed zephyrize closed 1 year ago

zephyrize commented 1 year ago

In the past few days, I tried to run the program based on my own dataset, but found that the utilization rate of the GPU is surprisingly low...

zephyrize commented 1 year ago

I think it is because of the inefficiency caused by the garbage collection ( gc.collect() )and clearing cache statements (torch.empty_cache() ) in the code. When I delete all the two codes in the project, the GPU utilization has improved significantly.