2d-gaussian-splatting icon indicating copy to clipboard operation
2d-gaussian-splatting copied to clipboard

Published 20 hours ago • verified publisher icon hbb1

CUDA out of memory

Open karthik101200 opened this issue 8 months ago • 6 comments

num_rendered, color, depth, radii, geomBuffer, binningBuffer, imgBuffer = _C.rasterize_gaussians(*args) torch.cuda.OutOfMemoryError: CUDA out of memory. Tried to allocate 35.79 GiB (GPU 0; 23.49 GiB total capacity; 990.78 MiB already allocated; 21.66 GiB free; 1.14 GiB reserved in total by PyTorch) If reserved memory is >> allocated memory try setting max_split_size_mb to avoid fragmentation. See documentation for Memory Management and PYTORCH_CUDA_ALLOC_CONF

35.79 gb seems like a lot?

When I run on more powerful mobile GPU (Ada A2000) but with less VRAM (8GB) it starts training but goes out of memory after 4.5k iterations. Any way to solve either of the issues?

karthik101200 avatar Jun 26 '24 14:06 karthik101200