search

NVlabs / ODISE

Official PyTorch implementation of ODISE: Open-Vocabulary Panoptic Segmentation with Text-to-Image Diffusion Models [CVPR 2023 Highlight]
https://arxiv.org/abs/2303.04803
Other
845 stars 45 forks source link

out of memory #34

Open elhamAm opened 11 months ago

elhamAm commented 11 months ago

CUDA out of memory. Tried to allocate 32.00 MiB (GPU 0; 23.69 GiB total capacity; 21.31 GiB already allocated; 12.06 MiB free; 21.46 GiB reserved in total by PyTorch) If reserved memory is >> allocated memory try setting

I am working on a gpu with 24 gb and I run out of space and I am using crops of 512x512 with batch size 1

Is this a memory leak problem? I am also using garbage collector but that only delays running into the 'out of memory' error. Help is appreciated :)

yeyimilk commented 7 months ago

in your case, it looks like your GPU was occupied by other tasks.