search

Vegeta2020 / SE-SSD

SE-SSD: Self-Ensembling Single-Stage Object Detector From Point Cloud, CVPR 2021.
Apache License 2.0
810 stars 128 forks source link

why sessd need so much memory? #47

Closed AndyYuan96 closed 2 years ago

AndyYuan96 commented 3 years ago

Hi, I just wonder why sessd need so much memory, the model looks not so large, for v100 32g, I can only set batch size =1, when batch size = 2, error happen: runtime error: cudnn error: CUDNN_STSTUS_NOT_SUPPORTED, this error may appear if you passed in a non-contiguous input. I just follow other change model_ema = build_detector() to solve the multiprocessing problem.

AndyYuan96 commented 3 years ago

but the actually memory occupied when set batch size = 1 is 12GB.

Vegeta2020 commented 2 years ago

@AndyYuan96 The student SSD should be initialized with a slightly or well converged pre-trained model, otherwise, it may produce some unexpected values in the ODIoU calculation, which may consume very large memory. Actually, the normal training of our SE-SSD is complelely affordable with a common GPU.