google / ml-fairness-gym Goto Github PK

Links in Environments readme are broken, for example, link to College Admission - https://github.com/google/ml-fairness-gym/blob/master/examples/docs/college_admission_example.md (no such file in the repository)

This is what I see trying to run the commands in the quick start, after git status:

	__pycache__/
	agents/__pycache__/
	bin/
	environments/__pycache__/
	examples/__pycache__/
	include/
	lib/
	metrics/__pycache__/
	spaces/__pycache__/

Should this project have a .gitignore file?

Hello world, I'm testing new notification settings.

I installed ml-fairness-gym using the instructions and Python version 3.7, but when I ran tests.sh a number of the tests failed and eventually I had to cancel the script because it froze my computer.

When I try running lending_demo.py I get the following error. I'm not sure what the issue is and haven't been able to debug it thus far.

Do you have a github version of this ml-fairness-code with tensorflow 1.8 or later? Currently requirements.txt uses tensorflow==1.15 which is not available for pip installation anymore. I have tensorflow 1.8 but that (obviously) creates a lot of dependancies issues. I am struggling to reproduce the results.

Please let me know.

Many thanks,
Ali

Hi,

Thanks for this really nice environment. I have been testing it out. However, I was looking to read the paper: "Fairness is not Static: Deeper understanding of long term fairness via agents and environments. KDD workshop on Explainable AI (XAI) 2019 " that I assume this work is based on.

Can you provide link to the provide if it is available somewhere?

Thanks!

Hello. Thanks for ML for fairness!

(venv) ➜ experiments git:(master) ✗ python ./lending_demo.py --num_steps=500
Traceback (most recent call last):
File "./lending_demo.py", line 26, in
from agents import threshold_policies

I already use the PATH code
export PYTHONPATH="${PYTHONPATH}:/path/to/ml-fairness-gym"
After ./tests.sh
File "/Library/Developer/CommandLineTools/Library/Frameworks/Python3.framework/Versions/3.7/lib/python3.7/traceback.py", line 285, in line
self._line = linecache.getline(self.filename, self.lineno).strip()
File "/Library/Developer/CommandLineTools/Library/Frameworks/Python3.framework/Versions/3.7/lib/python3.7/linecache.py", line 16, in getline
lines = getlines(filename, module_globals)
File "/Library/Developer/CommandLineTools/Library/Frameworks/Python3.framework/Versions/3.7/lib/python3.7/linecache.py", line 48, in getlines
for mod in sys.modules.values():
RuntimeError: dictionary changed size during iteration

And there I can not proceed anymore from here.

hello, thanks for sharing! 👍

The quick start doesn't work for me:

Traceback (most recent call last):
  File "examples/lending_demo.py", line 26, in <module>
    from agents import threshold_policies
ModuleNotFoundError: No module named 'agents'

I think this is trying to import a file, but agents/threshold_policies is in the root and not the examples folder.

This works though:

$ cp examples/lending_demo.py .
$ python ./lending_demo.py --num_steps=500

...and opens a bunch of windows with charts! :)

So maybe these paths or imports need to be updated, or the quick start instructions? I'm using conda so if this is just because of that vs. virtualenv then sorry if I don't know enough to tell the difference, just trying to be helpful :)

The examples link is broken:

https://github.com/google/ml-fairness-gym/blob/master/examples

I get the following error when trying to run the example:
AssertionError: Observation {'bank_cash': array(10001.), 'applicant_features': array([0, 1, 0, 0, 0, 0, 0], dtype=int32), 'group': array([1., 0.])} is not contained in self.observation_space
In call to configurable 'run_simulation' (<function run_simulation at 0x7fa07162a440>)

all the requirements should be installed, however I get a few failed tests when running ./tests.sh

• some undefined metric warnings: UndefinedMetricWarning: No positive samples in y_true, true positive value should be meaningless
• AssertionErrors as in the example.
  

Any help appreciated

While the other two experiments run, i keep getting this error concerned with Multi processing pool.
multiprocessing.pool.MaybeEncodingError: Error sending result: '<multiprocessing.pool.ExceptionWithTraceback object at 0x114b00f10>'. Reason: 'PicklingError("Can't pickle <class 'TypeError'>: it's not the same object as builtins.TypeError")'