BEVFormer icon indicating copy to clipboard operation
BEVFormer copied to clipboard

Published 20 hours ago • verified publisher icon fundamentalvision

out of memory

Open Another285 opened this issue 1 year ago • 1 comments

I got 2 gpus on my server,I run the train as ./tools/dist_train.sh ./projects/configs/bevformer/bevformer_base.py 2. It says that CUDA out of memory. Tried to allocate 252.00 MiB (GPU 0; 23.69 GiB total capacity; 267.51 MiB already allocated; 195.12 MiB free; 282.00 MiB reserved in total by PyTorch) It seems that it didn't use 2gpus. Iturned "2" into "3",it goes: RuntimeError: CUDA error: out of memory CUDA kernel errors might be asynchronously reported at some other API call,so the stacktrace below might be incorrect. Has everyone ever met this?

Another285 avatar Dec 07 '23 13:12 Another285

Download cudnn and give it a try

wuhen777 avatar Dec 22 '23 09:12 wuhen777