iurada / fairness_crossdomain

Official repository of our work "Fairness meets Cross-Domain Learning: a Benchmark of Models and Metrics" published on IEEE Access
https://fairvisionlearning.github.io/
5 stars 0 forks source link

list index out of range error #1

Open AWbosman opened 1 month ago

AWbosman commented 1 month ago

Hi!

I am trying to reproduce some part of your interesting work and just running your scripts from this repo. I have followed the installation guide etc but I get the following error when running your scripts:

Traceback (most recent call last):
  File "main.py", line 75, in <module>
    main()
  File "main.py", line 20, in main
    args = parse_arguments()
  File "x/fairness_crossdomain/parse_args.py", line 30, in parse_arguments
    exp_type = os.path.normpath(args.experiment).split(os.sep)[1]
IndexError: list index out of range

Can you help me with solving this? Or point me to where this error might originate?

iurada commented 4 weeks ago

Hello!

Thank you so much for the kind words! This error is a consequence of a (very bad) choice of input validation I did a while ago. The idea was to check that the user would run a task (eg. classification) only on the datasets that support it. I will work on a fix for this.

In the meantime, can you please share some more details about how you launched the experiments and your configuration (eg. os, python version...)

AWbosman commented 4 weeks ago

Hi! Thank you for the fast reply. O the choices that haunt us! Its nice that you are willing to work on this. In the meantime, would it be at all possible to get a file with weights from one of the architectures you used in the paper? Preferably from the celeb dataset, as I am still waiting for access to the other datasets. This would really help me, as I can check whether the architectures work at all for the application I am working on?

Some details. I used the same python version as you (3.7.6) to create a conda environment. I am running the experiments on a cluster. Is using conda a problem for any of the requirements?

Kindly

iurada commented 3 weeks ago

Hi again,

unfortunately, I have no longer access to the checkpoints. However, I pushed some really minor fixes and tried running the experiments again. If everything has been set up correctly, then you should be able to run the code by simply editing the launch scripts in the scripts/ folder (eg. to reproduce the baseline on CelebA, with "EyeBags" as protected attribute, you can launch the experiment as ./scripts/celeba_eyebags_baseline.sh. Just remember to change the file paths to suit your configuration).

Regarding the details, conda should definitely not be a problem.

Hope this can help!