polarisZhao / PFLD-pytorch

PFLD pytorch Implementation
808 stars 198 forks source link

Training: how much GPU memory do you need/ #10

Closed foolhard closed 5 years ago

foolhard commented 5 years ago

I try to train on NV2080 (8G), but failed with out of memory. What's your training environment? What GPU do you use?

RuntimeError: CUDA out of memory. Tried to allocate 26.00 MiB (GPU 0; 7.76 GiB total capacity; 6.88 GiB already allocated; 7.56 MiB free; 34.09 MiB cached)

polarisZhao commented 5 years ago

cuda 10.0 cudnn 7.5.0 GPU: RTX 2080ti