Atten4Vis / ConditionalDETR

This repository is an official implementation of the ICCV 2021 paper "Conditional DETR for Fast Training Convergence". (https://arxiv.org/abs/2108.06152)
Apache License 2.0
368 stars 50 forks source link

some question about lr #27

Closed LYMDLUT closed 2 years ago

LYMDLUT commented 2 years ago

Non-DC5 models were trained with batch size 2, and DC5 with 1,should the learning rate changed?

charlesCXK commented 2 years ago

Hi, we didn't change learning rate.