DeepStylometry

Embedding subtle stylometric features.

---

A deep learning architecture designed to investigate the best way of embedding stylistic features from text using contrastive learning and interpretability.

Requirements

• 3.9 <= Python <= 3.12
• A HuggingFace account in order to run the preprocessing script or get access to some pre-trained models.

Logging into your HuggingFace account

huggingface-cli login

Installation

pip install -r requirements.txt

Usage

It is recommended to use the shell scripts provided in the scripts folder to run the code. Make sure to modify the parameters in the scripts's but also in the configs's file you want to use to fit your need. The scripts are designed to be run from any directory.

If you still want to run the Python scripts directly:

Hyperparameter tuning

Set the hyperparameters you want in tune.yml.

usage: tune.py [-h] --config_path CONFIG_PATH --ray_storage_path
               RAY_STORAGE_PATH --logs_dir LOGS_DIR [--num_proc NUM_PROC]
               [--cache_dir CACHE_DIR]

Arguments used for hyper-parameter tuning.

options:
  -h, --help            show this help message and exit
  --config_path CONFIG_PATH
                        Path to the config file.
  --ray_storage_path RAY_STORAGE_PATH
                        Directory where Ray will save the logs and experiments
                        results.
  --logs_dir LOGS_DIR   Directory where the logs will be saved.
  --num_proc NUM_PROC   Number of processes to use. Default is the number of
                        CPUs minus one.
  --cache_dir CACHE_DIR
                        Path to the cache directory for HuggingFace.

Training/Fine-tuning

Set the hyperparameters you want in train.yml.

usage: train.py [-h] --config_path CONFIG_PATH --logs_dir LOGS_DIR
                [--num_proc NUM_PROC] [--checkpoint_dir CHECKPOINT_DIR]
                [--checkpoint_path CHECKPOINT_PATH] [--cache_dir CACHE_DIR]

Arguments used to train/fine-tune a model.

options:
  -h, --help            show this help message and exit
  --config_path CONFIG_PATH
                        Path to the config file.
  --logs_dir LOGS_DIR   Directory where the logs are stored.
  --num_proc NUM_PROC   Number of processes to use. Default is the number of
                        CPUs.
  --checkpoint_dir CHECKPOINT_DIR
                        Directory where the model's checkpoints are stored.
  --checkpoint_path CHECKPOINT_PATH
                        Path to a checkpoint file if it exists. This argument
                        is only used when testing an existing model.
  --cache_dir CACHE_DIR
                        Path to the cache directory for HuggingFace.

Citation

To cite DeepStylometry:

TBD

License

This project is licensed under the Apache License 2.0.