search

hikvision-research / SSOD

An official implementation of CVPR 2022 paper "Label Matching Semi-Supervised Object Detection".
Apache License 2.0
138 stars 14 forks source link

CUDA out of memory #6

Open SouthMaiya opened 2 years ago

SouthMaiya commented 2 years ago

batch_size=4 , 3090 Ti, the GPU's video memory is 24GB. Why does the code report the error"CUDA out of memory" after several iterations ？

AceCoooool commented 2 years ago
  1. We use V100 (GPU=32G) in our setting. You can use checkpoint in the backbone (need to combine label and unlabeled data in training) or use fp16 to save memory.
  2. "CUDA out of memory" after several iterations are caused by Resize pipeline.