Open Ericgone opened 4 months ago
Thank you for your interest and for bringing up this issue. Based on my evaluations, here are the peak CUDA memory consumptions for the nine scenes in Mip-NeRF 360 dataset: Bicycle: 19599.0 MB Bonsai: 5303.0 MB Counter: 8949.0 MB Flowers: 15765.0 MB Garden: 17695.0 MB Kitchen: 7513.0 MB Room: 7333.0 MB Stump: 14475.0 MB Treehill: 16177.0 MB To help reduce memory consumption, you might find these open-source projects useful:
hi,
thanks for your great work which i'm pretty interested in. but when I tried your pipeline, cuda OOM appeared. It seems that at some time point, the instant cuda mem required is extreamly high, which may be a misoperation by me. So I wonder how much the largest vram for this project is. thanks
best,