Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

about infer quert num #2

Open
shaoqb opened this issue Mar 2, 2023 · 3 comments
Open

about infer quert num #2

shaoqb opened this issue Mar 2, 2023 · 3 comments

Comments

@shaoqb
Copy link

shaoqb commented Mar 2, 2023

image
The paper says that the one-to-one branch is used for inference. Where is it reflected in the code?

@shaoqb
Copy link
Author

shaoqb commented Mar 2, 2023

Or you only use all_cls_scores and all_bbox_preds in your inference, but the number of queries is 2700?

@PkuRainBow
Copy link
Contributor

@LittlePey Please help to answer this question.

@LittlePey
Copy link
Contributor

@shaoqb Hi, to only use one-to-one branch in inference, you can set num_query=900, comment out Line494~Line498 and set self_attn_mask to None in Line499. To load ckpt trained with one-to-many, the num_query in Line328 should be 2700 and you need take the first 900 query in reference_points before using reference_points.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

3 participants