How to avoid CUDA out of memory

[Goru1890]

I'm trying to test my own dataset and I'm getting back a CUDA out of memory message:

RuntimeError: CUDA out of memory. Tried to allocate 1.19 GiB (GPU 0; 10.76 GiB total capacity; 6.91 GiB already allocated; 851.81 MiB free; 1.51 GiB cached)

Could it be because of my pytorch version? pytorch version ==1.3.1

[user-wu]

I'm trying to test my own dataset and I'm getting back a CUDA out of memory message:

RuntimeError: CUDA out of memory. Tried to allocate 1.19 GiB (GPU 0; 10.76 GiB total capacity; 6.91 GiB already allocated; 851.81 MiB free; 1.51 GiB cached)

Could it be because of my pytorch version? pytorch version ==1.3.1

maybe you need a RTX 3090