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

fine-tuning the model in the third step of training, #4

Open
gwx541 opened this issue Nov 1, 2024 · 4 comments
Open

fine-tuning the model in the third step of training, #4

gwx541 opened this issue Nov 1, 2024 · 4 comments

Comments

@gwx541
Copy link

gwx541 commented Nov 1, 2024

When fine-tuning the model in the third step of training, should the config of the main.py file be changed to the final_checkpoint.pth file in Pre-training and the train_eps_10.0_s.yaml file in PRIVIMAGE+D? After I changed it, opacus reported an error.

@SunnierLee
Copy link
Owner

Hi! You also need to change the "data.path" and "data.fid_stats" into yours in train_eps_10.0_s.yaml. Could you show me the error?

@gwx541
Copy link
Author

gwx541 commented Nov 2, 2024

d7943f5376c5b6437c1994d5f63d428 I had changed data.pth、data.fid_satas and model.ckpt The error is following 9b98afcbe68c5f00164eb6cac2005b1 9ef0957ff1288c218385694a78c9251

@gwx541
Copy link
Author

gwx541 commented Nov 2, 2024

Hi! You also need to change the "data.path" and "data.fid_stats" into yours in train_eps_10.0_s.yaml. Could you show me the error?
I had show the error.

@SunnierLee
Copy link
Owner

SunnierLee commented Nov 3, 2024

  1. Please run the following command to reinstall opacus:

cd xxx/DP-ImaGen
pip uninstall opacus
pip install -e src/opacus

  1. Please replace your DP-ImaGen/src/PRIVIMAGE+D/runners/train_dpdm_base.py and DP-ImaGen/src/PRIVIMAGE+D/model/layers.py with the latest files in the repo.

Besides, we fixed a bug in query_semantics.py yesterday. If you ran query_semantics.py before, you may need to update query_semantics.py, re-run it ,and then pre-train your models! You can check if the synthetic images generated by the pre-trained models are normal.

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

2 participants