fairlib

https://libr-ai.github.io/fairlib/
Admin
fairlib is a unified Python framework for assessing and improving classification fairness with...
Tokens:49,248
Snippets:392
Trust Score:9.7
Update:5 months ago
Show doc for...
Tokens:
Raw
### Example of pip installation output

Source: https://libr-ai.github.io/fairlib/tutorial/Structured_Inputs.html

This output shows the process of installing fairlib and its dependencies, including PyYAML, torch, transformers, and others. It indicates which packages are already satisfied and which are being downloaded and collected.

```text
Looking in indexes: https://pypi.org/simple, https://us-python.pkg.dev/colab-wheels/public/simple/
Collecting fairlib
  Downloading fairlib-0.0.9-py3-none-any.whl (85 kB)
     |████████████████████████████████| 85 kB 2.6 MB/s
Requirement already satisfied: PyYAML in /usr/local/lib/python3.7/dist-packages (from fairlib) (3.13)
Requirement already satisfied: torch in /usr/local/lib/python3.7/dist-packages (from fairlib) (1.12.0+cu113)
Collecting transformers
  Downloading transformers-4.20.1-py3-none-any.whl (4.4 MB)
     |████████████████████████████████| 4.4 MB 22.6 MB/s
Requirement already satisfied: seaborn in /usr/local/lib/python3.7/dist-packages (from fairlib) (0.11.2)
Requirement already satisfied: scikit-learn in /usr/local/lib/python3.7/dist-packages (from fairlib) (1.0.2)
Requirement already satisfied: tqdm in /usr/local/lib/python3.7/dist-packages (from fairlib) (4.64.0)
Requirement already satisfied: docopt in /usr/local/lib/python3.7/dist-packages (from fairlib) (0.6.2)
Collecting sacremoses
  Downloading sacremoses-0.0.53.tar.gz (880 kB)
     |████████████████████████████████| 880 kB 62.2 MB/s
Requirement already satisfied: numpy in /usr/local/lib/python3.7/dist-packages (from fairlib) (1.21.6)
Requirement already satisfied: pandas in /usr/local/lib/python3.7/dist-packages (from fairlib) (1.3.5)
Requirement already satisfied: matplotlib in /usr/local/lib/python3.7/dist-packages (from fairlib) (3.2.2)
Collecting pickle5
  Downloading pickle5-0.0.12-cp37-cp37m-manylinux_2_5_x86_64.manylinux1_x86_64.whl (256 kB)
     |████████████████████████████████| 256 kB 47.9 MB/s
Requirement already satisfied: pyparsing!=2.0.4,!=2.1.2,!=2.1.6,>=2.0.1 in /usr/local/lib/python3.7/dist-packages (from matplotlib->fairlib) (3.0.9)
Requirement already satisfied: kiwisolver>=1.0.1 in /usr/local/lib/python3.7/dist-packages (from matplotlib->fairlib) (1.4.4)
Requirement already satisfied: python-dateutil>=2.1 in /usr/local/lib/python3.7/dist-packages (from matplotlib->fairlib) (2.8.2)
Requirement already satisfied: cycler>=0.10 in /usr/local/lib/python3.7/dist-packages (from matplotlib->fairlib) (0.11.0)
Requirement already satisfied: typing-extensions in /usr/local/lib/python3.7/dist-packages (from kiwisolver>=1.0.1->matplotlib->fairlib) (4.1.1)
Requirement already satisfied: six>=1.5 in /usr/local/lib/python3.7/dist-packages (from python-dateutil>=2.1->matplotlib->fairlib) (1.15.0)
Requirement already satisfied: pytz>=2017.3 in /usr/local/lib/python3.7/dist-packages (from pandas->fairlib) (2022.1)
Requirement already satisfied: regex in /usr/local/lib/python3.7/dist-packages (from sacremoses->fairlib) (2022.6.2)
Requirement already satisfied: click in /usr/local/lib/python3.7/dist-packages (from sacremoses->fairlib) (7.1.2)
Requirement already satisfied: joblib in /usr/local/lib/python3.7/dist-packages (from sacremoses->fairlib) (1.1.0)
Requirement already satisfied: scipy>=1.1.0 in /usr/local/lib/python3.7/dist-packages (from scikit-learn->fairlib) (1.7.3)
Requirement already satisfied: threadpoolctl>=2.0.0 in /usr/local/lib/python3.7/dist-packages (from scikit-learn->fairlib) (3.1.0)
Collecting huggingface-hub<1.0,>=0.1.0
  Downloading huggingface_hub-0.8.1-py3-none-any.whl (101 kB)
     |████████████████████████████████| 101 kB 11.5 MB/s
Requirement already satisfied: packaging>=20.0 in /usr/local/lib/python3.7/dist-packages (from transformers->fairlib) (21.3)
Requirement already satisfied: requests in /usr/local/lib/python3.7/dist-packages (from transformers->fairlib) (2.23.0)
Requirement already satisfied: importlib-metadata in /usr/local/lib/python3.7/dist-packages (from transformers->fairlib) (4.12.0)
Requirement already satisfied: filelock in /usr/local/lib/python3.7/dist-packages (from transformers->fairlib) (3.7.1)
Collecting PyYAML
  Downloading PyYAML-6.0-cp37-cp37m-manylinux_2_5_x86_64.manylinux1_x86_64.manylinux_2_12_x86_64.manylinux2010_x86_64.whl (596 kB)
     |████████████████████████████████| 596 kB 11.8 MB/s
Collecting tokenizers!=0.11.3,<0.13,>=0.11.1
  Downloading tokenizers-0.12.1-cp37-cp37m-manylinux_2_12_x86_64.manylinux2010_x86_64.whl (6.6 MB)
     |████████████████████████████████| 6.6 MB 33.9 MB/s
Requirement already satisfied: zipp>=0.5 in /usr/local/lib/python3.7/dist-packages (from importlib-metadata->transformers->fairlib) (3.8.1)
Requirement already satisfied: idna<3,>=2.5 in /usr/local/lib/python3.7/dist-packages (from requests->transformers->fairlib) (2.10)
Requirement already satisfied: chardet<4,>=3.0.2 in /usr/local/lib/python3.7/dist-packages (from requests->transformers->fairlib) (3.0.4)
Requirement already satisfied: urllib3!=1.25.0,!=1.25.1,<1.26,>=1.21.1 in /usr/local/lib/python3.7/dist-packages (from requests->transformers->fairlib) (1.24.3)

```

--------------------------------

### Install _fairlib_ with pip

Source: https://libr-ai.github.io/fairlib/README.html

Installs the _fairlib_ package using pip. It is recommended to start within a new virtual environment.

```bash
conda create -n fairlib python=3.7
conda activate fairlib

pip install fairlib
```

--------------------------------

### Install _fairlib_ for Development

Source: https://libr-ai.github.io/fairlib/README.html

Sets up a development environment by cloning the repository and installing _fairlib_ in development mode.

```bash
git clone https://github.com/HanXudong/fairlib.git ~/fairlib
cd ~/fairlib
python setup.py develop
```

--------------------------------

### fairlib Installation Details

Source: https://libr-ai.github.io/fairlib/tutorial_customized_evaluation.html

This output shows the successful installation of the fairlib package and its dependencies, including versions of numpy, scikit-learn, and transformers.

```text
Collecting fairlib
  Downloading fairlib-0.0.3-py3-none-any.whl (63 kB)
[K     |████████████████████████████████| 63 kB 2.0 MB/s 
[?25hRequirement already satisfied: numpy in /usr/local/lib/python3.7/dist-packages (from fairlib) (1.21.6)
Requirement already satisfied: scikit-learn in /usr/local/lib/python3.7/dist-packages (from fairlib) (1.0.2)
Requirement already satisfied: seaborn in /usr/local/lib/python3.7/dist-packages (from fairlib) (0.11.2)
Requirement already satisfied: tqdm in /usr/local/lib/python3.7/dist-packages (from fairlib) (4.64.0)
Requirement already satisfied: docopt in /usr/local/lib/python3.7/dist-packages (from fairlib) (0.6.2)
Requirement already satisfied: pandas in /usr/local/lib/python3.7/dist-packages (from fairlib) (1.3.5)
Requirement already satisfied: PyYAML in /usr/local/lib/python3.7/dist-packages (from fairlib) (3.13)
Collecting pickle5
  Downloading pickle5-0.0.12-cp37-cp37m-manylinux_2_5_x86_64.manylinux1_x86_64.whl (256 kB)
[K     |████████████████████████████████| 256 kB 29.1 MB/s 
[?25hRequirement already satisfied: matplotlib in /usr/local/lib/python3.7/dist-packages (from fairlib) (3.2.2)
Collecting transformers
  Downloading transformers-4.18.0-py3-none-any.whl (4.0 MB)
[K     |████████████████████████████████| 4.0 MB 55.2 MB/s 
[?25hRequirement already satisfied: torch in /usr/local/lib/python3.7/dist-packages (from fairlib) (1.11.0+cu113)
Requirement already satisfied: python-dateutil>=2.1 in /usr/local/lib/python3.7/dist-packages (from matplotlib->fairlib) (2.8.2)
Requirement already satisfied: kiwisolver>=1.0.1 in /usr/local/lib/python3.7/dist-packages (from matplotlib->fairlib) (1.4.2)
Requirement already satisfied: cycler>=0.10 in /usr/local/lib/python3.7/dist-packages (from matplotlib->fairlib) (0.11.0)
Requirement already satisfied: pyparsing!=2.0.4,!=2.1.2,!=2.1.6,>=2.0.1 in /usr/local/lib/python3.7/dist-packages (from matplotlib->fairlib) (3.0.8)
Requirement already satisfied: typing-extensions in /usr/local/lib/python3.7/dist-packages (from kiwisolver>=1.0.1->matplotlib->fairlib) (4.2.0)
Requirement already satisfied: six>=1.5 in /usr/local/lib/python3.7/dist-packages (from python-dateutil>=2.1->matplotlib->fairlib) (1.15.0)
Requirement already satisfied: pytz>=2017.3 in /usr/local/lib/python3.7/dist-packages (from pandas->fairlib) (2022.1)
Requirement already satisfied: scipy>=1.1.0 in /usr/local/lib/python3.7/dist-packages (from scikit-learn->fairlib) (1.4.1)
Requirement already satisfied: threadpoolctl>=2.0.0 in /usr/local/lib/python3.7/dist-packages (from scikit-learn->fairlib) (3.1.0)
Requirement already satisfied: joblib>=0.11 in /usr/local/lib/python3.7/dist-packages (from scikit-learn->fairlib) (1.1.0)
Collecting sacremoses
  Downloading sacremoses-0.0.53.tar.gz (880 kB)
[K     |████████████████████████████████| 880 kB 51.9 MB/s 
[?25hRequirement already satisfied: requests in /usr/local/lib/python3.7/dist-packages (from transformers->fairlib) (2.23.0)
Requirement already satisfied: filelock in /usr/local/lib/python3.7/dist-packages (from transformers->fairlib) (3.6.0)
Requirement already satisfied: importlib-metadata in /usr/local/lib/python3.7/dist-packages (from transformers->fairlib) (4.11.3)
Collecting huggingface-hub<1.0,>=0.1.0
  Downloading huggingface_hub-0.5.1-py3-none-any.whl (77 kB)
[K     |████████████████████████████████| 77 kB 6.9 MB/s 
[?25hCollecting PyYAML
  Downloading PyYAML-6.0-cp37-cp37m-manylinux_2_5_x86_64.manylinux1_x86_64.manylinux_2_12_x86_64.manylinux2010_x86_64.whl (596 kB)
[K     |████████████████████████████████| 596 kB 60.7 MB/s 
[?25hRequirement already satisfied: regex!=2019.12.17 in /usr/local/lib/python3.7/dist-packages (from transformers->fairlib) (2019.12.20)
Collecting tokenizers!=0.11.3,<0.13,>=0.11.1
  Downloading tokenizers-0.12.1-cp37-cp37m-manylinux_2_12_x86_64.manylinux2010_x86_64.whl (6.6 MB)
[K     |████████████████████████████████| 6.6 MB 54.8 MB/s 
[?25hRequirement already satisfied: packaging>=20.0 in /usr/local/lib/python3.7/dist-packages (from transformers->fairlib) (21.3)
Requirement already satisfied: zipp>=0.5 in /usr/local/lib/python3.7/dist-packages (from importlib-metadata->transformers->fairlib) (3.8.0)
Requirement already satisfied: urllib3!=1.25.0,!=1.25.1,<1.26,>=1.21.1 in /usr/local/lib/python3.7/dist-packages (from requests->transformers->fairlib) (1.24.3)
Requirement already satisfied: certifi>=2017.4.17 in /usr/local/lib/python3.7/dist-packages (from requests->transformers->fairlib) (2021.10.8)
Requirement already satisfied: idna<3,>=2.5 in /usr/local/lib/python3.7/dist-packages (from requests->transformers->fairlib) (2.10)
Requirement already satisfied: chardet<4,>=3.0.2 in /usr/local/lib/python3.7/dist-packages (from requests->transformers->fairlib) (3.0.4)
Requirement already satisfied: click in /usr/local/lib/python3.7/dist-packages (from sacremoses->transformers->fairlib) (7.1.2)

```

--------------------------------

### Fairlib Installation Dependencies

Source: https://libr-ai.github.io/fairlib/tutorial/evaluation.html

This output shows the successful installation of Fairlib and lists all its dependencies, including numpy, scikit-learn, seaborn, and others, along with their installed versions. It also indicates which packages were already satisfied.

```text
Collecting fairlib
  Downloading fairlib-0.0.3-py3-none-any.whl (63 kB)
     |████████████████████████████████| 63 kB 2.0 MB/s
Requirement already satisfied: numpy in /usr/local/lib/python3.7/dist-packages (from fairlib) (1.21.6)
Requirement already satisfied: scikit-learn in /usr/local/lib/python3.7/dist-packages (from fairlib) (1.0.2)
Requirement already satisfied: seaborn in /usr/local/lib/python3.7/dist-packages (from fairlib) (0.11.2)
Requirement already satisfied: tqdm in /usr/local/lib/python3.7/dist-packages (from fairlib) (4.64.0)
Requirement already satisfied: docopt in /usr/local/lib/python3.7/dist-packages (from fairlib) (0.6.2)
Requirement already satisfied: pandas in /usr/local/lib/python3.7/dist-packages (from fairlib) (1.3.5)
Requirement already satisfied: PyYAML in /usr/local/lib/python3.7/dist-packages (from fairlib) (3.13)
Collecting pickle5
  Downloading pickle5-0.0.12-cp37-cp37m-manylinux_2_5_x86_64.manylinux1_x86_64.whl (256 kB)
     |████████████████████████████████| 256 kB 29.1 MB/s
Requirement already satisfied: matplotlib in /usr/local/lib/python3.7/dist-packages (from fairlib) (3.2.2)
Collecting transformers
  Downloading transformers-4.18.0-py3-none-any.whl (4.0 MB)
     |████████████████████████████████| 4.0 MB 55.2 MB/s
Requirement already satisfied: torch in /usr/local/lib/python3.7/dist-packages (from fairlib) (1.11.0+cu113)
Requirement already satisfied: python-dateutil>=2.1 in /usr/local/lib/python3.7/dist-packages (from matplotlib->fairlib) (2.8.2)
Requirement already satisfied: kiwisolver>=1.0.1 in /usr/local/lib/python3.7/dist-packages (from matplotlib->fairlib) (1.4.2)
Requirement already satisfied: cycler>=0.10 in /usr/local/lib/python3.7/dist-packages (from matplotlib->fairlib) (0.11.0)
Requirement already satisfied: pyparsing!=2.0.4,!=2.1.2,!=2.1.6,>=2.0.1 in /usr/local/lib/python3.7/dist-packages (from matplotlib->fairlib) (3.0.8)
Requirement already satisfied: typing-extensions in /usr/local/lib/python3.7/dist-packages (from kiwisolver>=1.0.1->matplotlib->fairlib) (4.2.0)
Requirement already satisfied: six>=1.5 in /usr/local/lib/python3.7/dist-packages (from python-dateutil>=2.1->matplotlib->fairlib) (1.15.0)
Requirement already satisfied: pytz>=2017.3 in /usr/local/lib/python3.7/dist-packages (from pandas->fairlib) (2022.1)
Requirement already satisfied: scipy>=1.1.0 in /usr/local/lib/python3.7/dist-packages (from scikit-learn->fairlib) (1.4.1)
Requirement already satisfied: threadpoolctl>=2.0.0 in /usr/local/lib/python3.7/dist-packages (from scikit-learn->fairlib) (3.1.0)
Requirement already satisfied: joblib>=0.11 in /usr/local/lib/python3.7/dist-packages (from scikit-learn->fairlib) (1.1.0)
Collecting sacremoses
  Downloading sacremoses-0.0.53.tar.gz (880 kB)
     |████████████████████████████████| 880 kB 51.9 MB/s
Requirement already satisfied: requests in /usr/local/lib/python3.7/dist-packages (from transformers->fairlib) (2.23.0)
Requirement already satisfied: filelock in /usr/local/lib/python3.7/dist-packages (from transformers->fairlib) (3.6.0)
Requirement already satisfied: importlib-metadata in /usr/local/lib/python3.7/dist-packages (from transformers->fairlib) (4.11.3)
Collecting huggingface-hub<1.0,>=0.1.0
  Downloading huggingface_hub-0.5.1-py3-none-any.whl (77 kB)
     |████████████████████████████████| 77 kB 6.9 MB/s
Collecting PyYAML
  Downloading PyYAML-6.0-cp37-cp37m-manylinux_2_5_x86_64.manylinux1_x86_64.manylinux_2_12_x86_64.manylinux2010_x86_64.whl (596 kB)
     |████████████████████████████████| 596 kB 60.7 MB/s
Requirement already satisfied: regex!=2019.12.17 in /usr/local/lib/python3.7/dist-packages (from transformers->fairlib) (2019.12.20)
Collecting tokenizers!=0.11.3,<0.13,>=0.11.1
  Downloading tokenizers-0.12.1-cp37-cp37m-manylinux_2_12_x86_64.manylinux2010_x86_64.whl (6.6 MB)
     |████████████████████████████████| 6.6 MB 54.8 MB/s
Requirement already satisfied: packaging>=20.0 in /usr/local/lib/python3.7/dist-packages (from transformers->fairlib) (21.3)
Requirement already satisfied: zipp>=0.5 in /usr/local/lib/python3.7/dist-packages (from importlib-metadata->transformers->fairlib) (3.8.0)
Requirement already satisfied: urllib3!=1.25.0,!=1.25.1,<1.26,>=1.21.1 in /usr/local/lib/python3.7/dist-packages (from requests->transformers->fairlib) (1.24.3)
Requirement already satisfied: certifi>=2017.4.17 in /usr/local/lib/python3.7/dist-packages (from requests->transformers->fairlib) (2021.10.8)
Requirement already satisfied: idna<3,>=2.5 in /usr/local/lib/python3.7/dist-packages (from requests->transformers->fairlib) (2.10)
Requirement already satisfied: chardet<4,>=3.0.2 in /usr/local/lib/python3.7/dist-packages (from requests->transformers->fairlib) (3.0.4)
Requirement already satisfied: click in /usr/local/lib/python3.7/dist-packages (from sacremoses->transformers->fairlib) (7.1.2)
Building wheels for collected packages: sacremoses

```

--------------------------------

### Setup Plotting Environment

Source: https://libr-ai.github.io/fairlib/tutorial/interactive_plots.html

Set up the plotting environment using Matplotlib and Altair. The `%matplotlib inline` magic command is used for displaying plots directly in the notebook. Ensure necessary libraries like altair and vega_datasets are installed.

```python
# Notice: make sure the enable widget when using interactive plots
# %matplotlib widget

%matplotlib inline

import altair as alt
from vega_datasets import data
```

--------------------------------

### DAdv Example Command

Source: https://libr-ai.github.io/fairlib/reference_component_hyperparameter_tuning.html

Example command to run DAdv with specific hyperparameter settings. Use --adv_num_subDiscriminator to set the number of subdiscriminators and --adv_diverse_lambda to control the strength of the difference loss.

```bash
python main.py --adv_debiasing --adv_num_subDiscriminator 3 --adv_diverse_lambda 10
```

--------------------------------

### Import and Use _fairlib_ Package

Source: https://libr-ai.github.io/fairlib/README.html

Demonstrates how to import and use _fairlib_ as a Python package. This involves loading options from a configuration file, initializing a model, and starting the training process with debiasing.

```python
from fairlib.base_options import options
from src import networks

config_file = 'opt.yaml'
# Get options
state = options.get_state(conf_file=config_file)

# Init the model
model = networks.get_main_model(state)

# Training with debiasing
model.train_self()
```

--------------------------------

### Import Fairlib Library

Source: https://libr-ai.github.io/fairlib/tutorial/interactive_plots.html

Import the Fairlib library to access its analysis and plotting functionalities. Ensure Fairlib is installed in your environment.

```python
import fairlib
```

--------------------------------

### Example LATEX Table Output

Source: https://libr-ai.github.io/fairlib/tutorial/fairlib_demo.html

This is an example of the LATEX table output generated by the `to_latex()` method. It includes various performance and fairness metrics for different models.

```latex
\begin{tabular}{lrrrrrrrrrlll}
\toprule
        Models &  test_performance mean &  test_performance std &  test_fairness mean &  test_fairness std &  dev_performance mean &  dev_performance std &  dev_fairness mean &  dev_fairness std &      DTO &         epoch list &                                       opt_dir list &  is_pareto \\
\midrule
          GDEO &               0.752763 &              0.004999 &            0.892255 &           0.007860 &              0.749350 &             0.003494 &           0.912672 &          0.002766 & 0.269694 &  [2, 12, 5, 9, 10] & [G:\\Experimental_results\\GroupDifference\\Moji\\G... &       True \\
   BTFairBatch &               0.746837 &              0.003407 &            0.899351 &           0.004936 &              0.743975 &             0.004236 &           0.919254 &          0.004731 & 0.272437 &    [8, 6, 9, 5, 7] & [G:\\Experimental_results\\FairBatch\\Moji\\BTInitF... &      False \\
       Vanilla &               0.722981 &              0.004576 &            0.611870 &           0.014356 &              0.726650 &             0.003673 &           0.632302 &          0.013370 & 0.476849 &   [2, 11, 2, 5, 2] & [G:\\Experimental_results\\vanilla\\Moji\\0\\opt.yam... &       True \\
          BTEO &               0.753927 &              0.001433 &            0.877469 &           0.003756 &              0.746325 &             0.000998 &           0.896874 &          0.005401 & 0.274892 &    [8, 6, 9, 4, 5] & [G:\\Experimental_results\\GatedBT\\Moji\\GatedBT_R... &       True \\
     GatedDAdv &               0.750163 &              0.006945 &            0.908679 &           0.021678 &              0.745600 &             0.004828 &           0.928670 &          0.022488 & 0.266004 & [24, 13, 19, 4, 3] & [G:\\Experimental_results\\hypertune3\\Moji\\hypert... &      False \\
     FairBatch &               0.751488 &              0.005772 &            0.904373 &           0.008213 &              0.746050 &             0.003896 &           0.914526 &          0.006020 & 0.266276 &    [9, 9, 6, 6, 5] & [G:\\Experimental_results\\FairBatch\\Moji\\FairBat... &       True \\
      GatedAdv &               0.753113 &              0.005196 &            0.890065 &           0.013302 &              0.748975 &             0.003805 &           0.910838 &          0.010314 & 0.270257 & [11, 4, 4, 13, 12] & [G:\\Experimental_results\\hypertune2\\Moji\\hypert... &      False \\
DelayedCLS_Adv &               0.761015 &              0.003081 &            0.882425 &           0.015918 &              0.751675 &             0.003481 &           0.899346 &          0.011417 & 0.266341 &  [13, 1, 3, 13, 1] & [/data/cephfs/punim1421/Fair_NLP_Classification... &       True \\
        GDMean &               0.752163 &              0.002130 &            0.901389 &           0.003916 &              0.749050 &             0.001368 &           0.922430 &          0.005829 & 0.266735 &  [11, 12, 7, 2, 2] & [G:\\Experimental_results\\GroupDifference\\Moji\\G... &       True \\
     GatedBTEO &               0.762106 &              0.002592 &            0.900764 &           0.014701 &              0.759775 &             0.003798 &           0.909445 &          0.006631 & 0.257762 &   [3, 1, 3, 1, 11] & [G:\\Experimental_results\\GatedBT\\Moji\\GatedBT_R... &       True \\
    BTGatedAdv &               0.735459 &              0.028830 &            0.866150 &           0.028232 &              0.730150 &             0.024594 &           0.886862 &          0.030537 & 0.296476 &  [5, 19, 19, 6, 1] & [G:\\Experimental_results\\hypertune2\\Moji\\hypert... &       True \\
           Adv &               0.756414 &              0.007271 &            0.893286 &           0.005623 &              0.747425 &             0.004549 &           0.912125 &          0.008507 & 0.265936 & [9, 14, 18, 5, 16] & [G:\\Experimental_results\\hypertune\\Moji\\hypertu... &       True \\
  OldFairBatch &               0.750638 &              0.006012 &            0.905537 &           0.005046 &              0.744525 &             0.004995 &           0.917734 &          0.004761 & 0.266655 &   [5, 7, 8, 13, 7] & [G:\\Experimental_results\\Original_FairBatch\\Moj... &       True \\

```

--------------------------------

### Main Function for Model Training and Post-processing

Source: https://libr-ai.github.io/fairlib/tutorial_customized_methods.html

The main function initializes options, gets the model state, trains the model, and then conditionally runs post-processing methods like INLP if specified in the state.

```python
def main():
    options = BaseOptions()
    state = options.get_state()


    # Init the model
    model = networks.get_main_model(state)
    # state.opt.main_model = model
    logging.info('Model Initialized!')


    model.train_self()
    logging.info('Model Trained!')


    if state.INLP:
        logging.info('Run INLP')
        from src.networks.INLP import get_INLP_trade_offs
        get_INLP_trade_offs(model, state)


    logging.info('Finished!')
```

--------------------------------

### Custom Dataset for Image Loading

Source: https://libr-ai.github.io/fairlib/tutorial/Image_Inputs.html

Create a custom dataset by inheriting from BaseDataset. This example defines a transform for image normalization and loads data from a specified file.

```python
class CustomizedDataset(dataloaders.utils.BaseDataset):

    transform=transforms.Compose([
        transforms.ToTensor(),
        transforms.Normalize((0.1307, 0.1307, 0.), (0.3081, 0.3081, 0.3081))])

    def load_data(self):

        self.data_dir = os.path.join(self.args.data_dir, "colored_MNIST_{}.pt".format(self.split))

        data = torch.load(self.data_dir)

        self.X = [self.transform(_img) for _img in data[0]]
        self.y = data[1]
        self.protected_label = data[2]
```

--------------------------------

### Model Training Output Log

Source: https://libr-ai.github.io/fairlib/tutorial/fairlib_demo.html

Example log output from the `train_self` method, showing epoch progress, loss, data time, training time, and evaluation metrics like accuracy and F-score.

```log
2022-07-21 06:53:10 [INFO ]  Epoch:    0 [      0/  99998 ( 0%)]        Loss: 0.6907     Data Time: 0.06s       Train Time: 0.18s
2022-07-21 06:53:14 [INFO ]  Epoch:    0 [  51200/  99998 (51%)]        Loss: 0.4075     Data Time: 0.44s       Train Time: 3.34s
2022-07-21 06:53:18 [INFO ]  Evaluation at Epoch 0
2022-07-21 06:53:18 [INFO ]  Validation accuracy: 71.44 macro_fscore: 71.35     micro_fscore: 71.44     TPR_GAP: 39.59  FPR_GAP: 39.59  PPR_GAP: 38.67
2022-07-21 06:53:18 [INFO ]  Test accuracy: 71.21       macro_fscore: 71.12     micro_fscore: 71.21     TPR_GAP: 39.42  FPR_GAP: 39.42  PPR_GAP: 38.49
2022-07-21 06:53:18 [INFO ]  Epoch:    1 [      0/  99998 ( 0%)]        Loss: 0.4195     Data Time: 0.01s       Train Time: 0.07s
2022-07-21 06:53:22 [INFO ]  Epoch:    1 [  51200/  99998 (51%)]        Loss: 0.3780     Data Time: 0.42s       Train Time: 3.29s
2022-07-21 06:53:26 [INFO ]  Evaluation at Epoch 1
2022-07-21 06:53:26 [INFO ]  Validation accuracy: 72.10 macro_fscore: 72.06     micro_fscore: 72.10     TPR_GAP: 37.69  FPR_GAP: 37.69  PPR_GAP: 37.00
2022-07-21 06:53:26 [INFO ]  Test accuracy: 71.72       macro_fscore: 71.68     micro_fscore: 71.72     TPR_GAP: 37.11  FPR_GAP: 37.11  PPR_GAP: 36.41
2022-07-21 06:53:27 [INFO ]  Epoch:    2 [      0/  99998 ( 0%)]        Loss: 0.3608     Data Time: 0.01s       Train Time: 0.07s
2022-07-21 06:53:30 [INFO ]  Epoch:    2 [  51200/  99998 (51%)]        Loss: 0.3567     Data Time: 0.42s       Train Time: 3.29s
2022-07-21 06:53:34 [INFO ]  Epochs since last improvement: 1
2022-07-21 06:53:34 [INFO ]  Evaluation at Epoch 2
2022-07-21 06:53:34 [INFO ]  Validation accuracy: 71.53 macro_fscore: 71.44     micro_fscore: 71.53     TPR_GAP: 39.40  FPR_GAP: 39.40  PPR_GAP: 38.40
2022-07-21 06:53:34 [INFO ]  Test accuracy: 71.38       macro_fscore: 71.28     micro_fscore: 71.38     TPR_GAP: 39.07  FPR_GAP: 39.07  PPR_GAP: 37.99
2022-07-21 06:53:35 [INFO ]  Epoch:    3 [      0/  99998 ( 0%)]        Loss: 0.3584     Data Time: 0.01s       Train Time: 0.07s
2022-07-21 06:53:38 [INFO ]  Epoch:    3 [  51200/  99998 (51%)]        Loss: 0.3430     Data Time: 0.42s       Train Time: 3.28s
2022-07-21 06:53:42 [INFO ]  Epochs since last improvement: 2
2022-07-21 06:53:42 [INFO ]  Evaluation at Epoch 3
2022-07-21 06:53:43 [INFO ]  Validation accuracy: 71.08 macro_fscore: 70.73     micro_fscore: 71.08     TPR_GAP: 41.08  FPR_GAP: 41.08  PPR_GAP: 38.75
2022-07-21 06:53:43 [INFO ]  Test accuracy: 70.87       macro_fscore: 70.49     micro_fscore: 70.87     TPR_GAP: 40.67  FPR_GAP: 40.67  PPR_GAP: 38.42
2022-07-21 06:53:43 [INFO ]  Epoch:    4 [      0/  99998 ( 0%)]        Loss: 0.3578     Data Time: 0.02s       Train Time: 0.07s
2022-07-21 06:53:46 [INFO ]  Epoch:    4 [  51200/  99998 (51%)]        Loss: 0.3881     Data Time: 0.44s       Train Time: 3.36s
2022-07-21 06:53:50 [INFO ]  Epochs since last improvement: 3
2022-07-21 06:53:50 [INFO ]  Evaluation at Epoch 4
2022-07-21 06:53:51 [INFO ]  Validation accuracy: 72.06 macro_fscore: 71.96     micro_fscore: 72.06     TPR_GAP: 38.04  FPR_GAP: 38.04  PPR_GAP: 36.87
2022-07-21 06:53:51 [INFO ]  Test accuracy: 71.72       macro_fscore: 71.61     micro_fscore: 71.72     TPR_GAP: 38.09  FPR_GAP: 38.09  PPR_GAP: 36.96
2022-07-21 06:53:51 [INFO ]  Epoch:    5 [      0/  99998 ( 0%)]        Loss: 0.4008     Data Time: 0.02s       Train Time: 0.07s
2022-07-21 06:53:54 [INFO ]  Epoch:    5 [  51200/  99998 (51%)]        Loss: 0.3393     Data Time: 0.43s       Train Time: 3.29s
2022-07-21 06:53:58 [INFO ]  Epochs since last improvement: 4
2022-07-21 06:53:58 [INFO ]  Evaluation at Epoch 5
2022-07-21 06:53:59 [INFO ]  Validation accuracy: 71.44 macro_fscore: 71.41     micro_fscore: 71.44     TPR_GAP: 40.50  FPR_GAP: 40.50  PPR_GAP: 39.97
2022-07-21 06:53:59 [INFO ]  Test accuracy: 71.13       macro_fscore: 71.11     micro_fscore: 71.13     TPR_GAP: 39.69  FPR_GAP: 39.69  PPR_GAP: 39.19
2022-07-21 06:53:59 [INFO ]  Epoch:    6 [      0/  99998 ( 0%)]        Loss: 0.3890     Data Time: 0.01s       Train Time: 0.07s
2022-07-21 06:54:03 [INFO ]  Epoch:    6 [  51200/  99998 (51%)]        Loss: 0.3668     Data Time: 0.44s       Train Time: 3.32s
2022-07-21 06:54:07 [INFO ]  Epochs since last improvement: 5
2022-07-21 06:54:07 [INFO ]  Evaluation at Epoch 6
2022-07-21 06:54:07 [INFO ]  Validation accuracy: 71.96 macro_fscore: 71.92     micro_fscore: 71.96     TPR_GAP: 37.52  FPR_GAP: 37.52  PPR_GAP: 36.82
2022-07-21 06:54:07 [INFO ]  Test accuracy: 72.04       macro_fscore: 72.00     micro_fscore: 72.04     TPR_GAP: 36.75  FPR_GAP: 36.75  PPR_GAP: 36.01
```

--------------------------------

### Configure Debiasing Arguments in Fairlib

Source: https://libr-ai.github.io/fairlib/tutorial/fairlib_demo.html

Specify arguments for debiasing methods, including experiment ID, adversarial debiasing flag, and Balanced Training (BT) parameters like 'Downsampling' and 'EO' objective. This setup is essential for enabling advanced bias mitigation techniques.

```python
debiasing_args = args.copy()

# Update the experiment name
debiasi_args["exp_id"] = "BT_Adv"

# Perform adversarial training if True
debiasi_args["adv_debiasing"] = True

# Specify the hyperparameters for Balanced Training
debiasi_args["BT"] = "Downsampling"
debiasi_args["BTObj"] = "EO"

debias_options = fairlib.BaseOptions()
debias_state = debias_options.get_state(args=debiasing_args, silence=True)

fairlib.utils.seed_everything(2022)

debias_model = fairlib.networks.get_main_model(debias_state)
```

--------------------------------

### Initialize Fairlib Options and State

Source: https://libr-ai.github.io/fairlib/tutorial/evaluation.html

Set up experiment-specific arguments by merging shared options and initializing the Fairlib BaseOptions. This prepares the environment for training.

```python
args = {
    "dataset":Shared_options["dataset"],
    "data_dir":Shared_options["data_dir"],
    "device_id":Shared_options["device_id"],

    # Give a name to the exp, which will be used in the path
    "exp_id":"vanilla",
}

# Init the argument
options = fairlib.BaseOptions()
state = options.get_state(args=args, silence=True)

fairlib.utils.seed_everything(2022)

```

--------------------------------

### Configure Experiment Options

Source: https://libr-ai.github.io/fairlib/tutorial/fairlib_demo.html

Initialize experiment arguments and state, including dataset, data directory, device ID, and experiment ID. This sets up the configuration for training.

```python
args = {
    # The name of the dataset, corresponding dataloader will be used,
    "dataset":  "Moji",

    # Specifiy the path to the input data
    "data_dir": "data/deepmoji",

    # Device for computing, -1 is the cpu; non-negative numbers indicate GPU id.
    "device_id":    -1,

    # Give a name to the exp, which will be used in the path
    "exp_id":"vanilla",
}

# Init the argument
options = fairlib.BaseOptions()
state = options.get_state(args=args, silence=True)
```

--------------------------------

### Reproduce Experiments with Configuration File

Source: https://libr-ai.github.io/fairlib/tutorial_usage.html

Use a configuration file to reproduce experiments by specifying the path to the opt.yaml file.

```python
python fairlib --conf_file PATH_TO_EXP/opt.yaml

```

--------------------------------

### Initialize Fairlib Options and Seed

Source: https://libr-ai.github.io/fairlib/tutorial_customized_evaluation.html

Set up experiment-specific arguments by inheriting from shared options, then initialize the fairlib BaseOptions and seed the random number generator for reproducibility.

```python
args = {
    "dataset":Shared_options["dataset"], 
    "data_dir":Shared_options["data_dir"],
    "device_id":Shared_options["device_id"],

    # Give a name to the exp, which will be used in the path
    "exp_id":"vanilla",
}

# Init the argument
options = fairlib.BaseOptions()
state = options.get_state(args=args, silence=True)

fairlib.utils.seed_everything(2022)

```

--------------------------------

### Get data distribution after sampling

Source: https://libr-ai.github.io/fairlib/tutorial/manipulate_data_distribution.html

Calculate the data distribution of the resampled data to verify the effect of generalized sampling.

```python
balanced_g_distribution = get_data_distribution(
    y_data = y[balanced_g_indices],
    g_data = g[balanced_g_indices])
```

--------------------------------

### Enable Inline Plotting

Source: https://libr-ai.github.io/fairlib/tutorial/fairlib_demo.html

This command enables matplotlib plots to be displayed directly within the notebook or script. It's a standard setup for data visualization.

```python
%matplotlib inline
```

--------------------------------

### Run INLP with main.py

Source: https://libr-ai.github.io/fairlib/reference_component_hyperparameter_tuning.html

This command initiates the INLP process by running the main script with the --INLP flag. It serves as the entry point for applying the INLP method.

```python
python main.py --INLP

```
Tokens:
Raw