search

PRBonn / PIN_SLAM

📍PIN-SLAM: LiDAR SLAM Using a Point-Based Implicit Neural Representation for Achieving Global Map Consistency [TRO' 24]
MIT License
321 stars 30 forks source link

torch.cuda.OutOfMemoryError: CUDA out of memory #18

Closed zfkdsg closed 4 months ago

zfkdsg commented 4 months ago

Hello, when I try to run your code run_demo.yaml，the following error has occurred: torch.cuda.OutOfMemoryError: CUDA out of memory. Tried to allocate 396.00 MiB. GPU 0 has a total capacity of 3.81 GiB of which 348.00 MiB is free. My gpu is RTX3050,How can I modify this program to make it run properly. Thank you!

zfkdsg commented 4 months ago

By the way，my GPU memory is 4GB

kevintsq commented 4 months ago

Reducing the batch size in the config file may help.

zfkdsg commented 4 months ago

thanks I will try it out。

YuePanEdward commented 4 months ago

Additionally, you may

  1. decrease the size of the voxel hashing buffer: https://github.com/PRBonn/PIN_SLAM/blob/1c2fc2fcc910520e0aa9d7e8da2d889f456f0c4b/utils/config.py#L94 These would also take up self.buffer_size*8 Bytes, which equals to 400 MB here.

  2. decrease the training data pool capacity (default value 1e7): This can be directly set in the config file, add:

    continual:
pool_capacity: 1e6