yenchenlin / nerf-pytorch

A PyTorch implementation of NeRF (Neural Radiance Fields) that reproduces the results.
MIT License
5.38k stars 1.05k forks source link

torch.autograd.set_detect_anomaly(True) seems slow down the training #42

Closed zfkuang closed 2 years ago

zfkuang commented 3 years ago

Hi,

Thanks for the work on this repo! I just found that calling this function (line 2 of run_nerf_helpers.py) can somehow slow down the training process by a large amount (almost 50% in my computer). Since this is a debugging option I'd like to suggest to turn it off as default. How do you think?

Thanks!

yenchenlin commented 2 years ago

Sounds good, done. Huge thanks for the suggestion!