search

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
358 stars 48 forks source link

Encoder ends with two LayerNorm layers #11

Closed joakimjohnander closed 2 years ago

joakimjohnander commented 2 years ago

Hey,

First and foremost, thank you for your great work. I have one question. The encoder (using default settings) seems to terminate with two nn.LayerNorm . One is within TransformerEncoderLayer and the other within TransformerEncoder. Is this intentional?

Best regards, Joakim

DeppMeng commented 2 years ago

Hi,

It is not intentional. We just use the same code as DETR on this part. We did not study how it affect the performance.