deepinsight / deepinsight/insightface

How should DALI be used and why is it not speeding up?

Open
#2,514 0 comments 0 reactions 0 assignees View on GitHub
Dominant language
Python
Stars
29.7k
Forks
6.1k
PR merge metrics
No merged PRs in 30d

Description

Hi, all:
When I trained face recognition, I just generated mxrecord and set config.dali=True in the config file. I found that there was no speed-up effect with R200.
My question is:
1. Is this configuration correct? Is there anything else that needs to be pay attention to?
2. I usd the same configuration as the training log of WF42M-PFC-0.2-r100(256*16) on multi-host GPUs, training on V100-SXM2-32G x16 by reading images directly. I found that the training speed without DALI is even faster than your training speed with DALI in log, is this normal?
3. Does everyone use DALI to speed up? By how much?

Can anyone provide some suggestions? Thank you!

Contributor guide

No contributing guide indexed for this repository

Assessment

This issue has not been assessed yet.

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.