search

facebookresearch / nonrigid_nerf

Open source repository for the code accompanying the paper 'Non-Rigid Neural Radiance Fields Reconstruction and Novel View Synthesis of a Deforming Scene from Monocular Video'.
Other
362 stars 40 forks source link

Colab runtime crash due to out of system RAM #20

Open m0o0scar opened 1 year ago

m0o0scar commented 1 year ago

Description

Hi. Thanks for this amazing work!! I'm trying to follow the "Walkthrough With an Example Sequence" on Google Colab. After something like ~14 hours of training I finally reach the final rendering step. But Colab runtime crashed due to out of system RAM, even when I'm running on a high-RAM instance:

Snipaste_2023-03-03_09-49-29

Instructions To Reproduce the Issue:

Please visit this Colab notebook, I basically copy all the codes from "Walkthrough With an Example Sequence", the only modification is to connect to my Google Drive to save training result.

edgar-tr commented 1 year ago

Hi,that might well be the case. I think I only optimized the code to be a bit flexible with respect to VRAM and assumed sufficient RAM. I trained on machines with at least 100-200GB RAM. The issue might be that it probably stores all the intermediate information like the color per point sample on the ray of every ray/pixel in the image.

baiyuntao00 commented 1 year ago

Description

Hi. Thanks for this amazing work!! I'm trying to follow the "Walkthrough With an Example Sequence" on Google Colab. After something like ~14 hours of training I finally reach the final rendering step. But Colab runtime crashed due to out of system RAM, even when I'm running on a high-RAM instance:

Snipaste_2023-03-03_09-49-29

Instructions To Reproduce the Issue:

Please visit this Colab notebook, I basically copy all the codes from "Walkthrough With an Example Sequence", the only modification is to connect to my Google Drive to save training result.

Hi, I also have the same problem. Besides adding the RAM memory modules, do you have other solutions to the problem?