RuntimeError: CUDA out of memory. Tried to allocate 64.00 MiB (GPU 0; 10.76 GiB total capacity; 9.87 GiB already allocated; 34.44 MiB free; 9.89 GiB reserved in total by PyTorch)

princeton-vl / pytorch_stacked_hourglass

Pytorch implementation of the ECCV 2016 paper "Stacked Hourglass Networks for Human Pose Estimation"

BSD 3-Clause "New" or "Revised" License

469 stars 94 forks source link

Closed Scheng-xiang closed 4 years ago

Scheng-xiang commented 4 years ago

why it happend?

Scheng-xiang commented 4 years ago

My gpu is 2080ti and have 11g memory, does it need more than 11g memory to trian or something was wrong happend?

crockwell commented 4 years ago

Depends on your batch size. For the full 8HG network & batch size of 16 or 32, you'll probably need multiple GPUs.