JunyuanDeng / NeRF-LOAM

[ICCV2023] NeRF-LOAM: Neural Implicit Representation for Large-Scale Incremental LiDAR Odometry and Mapping
MIT License
491 stars 29 forks source link

cuda out of memory #11

Closed whwh747 closed 6 months ago

whwh747 commented 6 months ago

Hello, thank you for your kind words. To run the demo on a TiTan XP graphics card without encountering a CUDA OOM (Out of Memory) issue at 20 frames, you would need a graphics card with at least how much memory?

JunyuanDeng commented 6 months ago

I tested my code with RTX 3090 and NVIDIA TITAN RTX 24GB, both of which have 24GB memory.

In my personal experience, a GPU with 16GB memory is sufficient.

您好,大概需要16GB的显存,我之前测试都是用24GB显存的卡跑,没什么特别问题

whwh747 commented 6 months ago

thank you