search

sconlyshootery / FeatDepth

This is the offical codes for the methods described in the "Feature-metric Loss for Self-supervised Learning of Depth and Egomotion".
MIT License
247 stars 28 forks source link

Why use DistOptimizerHook? #79

Closed adizhol closed 2 years ago

adizhol commented 2 years ago

Hello, It seems that DistOptimizerHook is averaging the gradients across all processes\gpus, but shouldn't torch's DDP handle this?

Thanks. Adi

sconlyshootery commented 2 years ago

Since our work is based on an early version of MMCV, in which time, DDP is rarely used.