tokaka22 / GMAA

38 stars 6 forks source link

Some question about training and debug #6

Open junyizeng opened 1 year ago

junyizeng commented 1 year ago

Thank you very much for open sourcing the code. I encountered some problems while replicating the attack effect and I hope to receive your response regarding the following: 1. In step 1, you used Celeba-HQ-Mask images for filtering in the process.py file, and in the last line, it says "assert ct=27255". However, after running the program, I only extracted 4081 images. Does this affect the subsequent model training? 2.When running tran.sh, I encountered the following error. How can I resolve it? 3. In your requirements.txt, you need to download mlflow and pre-commit without specifying the version. However, there is a conflict between the two libraries when it comes to the importlib-metadata library requirement, so I did not install pre-commit. Could you provide the specific versions for these two libraries? 4. I briefly looked at your code, and I would like to understand some details of the GMAA implementation through debugging. However, it seems inconvenient to use PyCharm for debugging using the current running method in the sh file. Therefore, I would like to ask how to debug the code. Looking forward to your reply.

preprocess 问题截图