city-super / Octree-GS

Octree-GS: Towards Consistent Real-time Rendering with LOD-Structured 3D Gaussians
Other
561 stars 40 forks source link

CUDA out of memory #58

Open WS517 opened 2 days ago

WS517 commented 2 days ago

Thanks for your great work, I tried to render on a large scene, more than 2000 images, but my gpu is only 24G and the following problem occurs: RuntimeError: CUDA out of memory. Tried to allocate 18.00 MiB (GPU 0; 23.70 GiB total capacity; 20.71 GiB already allocated; 1.38 MiB free; 22.63 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 My configuration is as follows: scene="1015_Z’ exp_name="baseline’ gpu=-1 ratio=1 resolution=-1 appearance_dim=0

fork=2 base_layer=4 #original is 12 visible_threshold=0.9 dist2level="round’ update_ratio=0.2

progressive="True’ dist_ratio=0.999 #0.99 dist_ratio=0.999 #0.99 init_level=-1 extra_ratio=0.25 extra_up=0.01 Can you tell me how I should change the parameters to make the rendering successful? Looking forward to your reply, thank you very much!

fshlol commented 8 hours ago

I have the same CUDA out of memory error, looking for solution too