NLP Pipeline

This repository contains an text machine learning pipeline and some of the projects i've worked on, either for fun/education or competition on Kaggle.
Each project have it's own readme containing information about the specific problematics of each.

I train the models locally on my pc using a Nvidia 1080 GPU.

Data

The data is not in the repository directly if you want to launch a model on one the projects in here you must download the data and change the config file in the task folder to be adequate.
Links to the datasets are in the tasks README's.

Projects

---

The projects folder contains the specific code about each project :

• config.py file containing most of the parameters necessary to train a model.
• feature_eng.py that contain the specific pre-processing functions you want to perform on the text for training, validation and testing

What if i want to add a new project ?

To add a new project you'll to create a few things :

• a new folder in the projects/ folder containing a __init__.py file, a config.py file and a feature_eng.py file
• copy and paste the content of another config.py file and change the information to be adequate with your task
• Add the functions you require to the feature_eng.py file

Training

To start training a model on any project use this command in terminal :

python -m train --project=tweet_disaster

You can replace the tweet_disaster with any folder in projects/. Default parameters train using a DISTILBERT model. You can change these parameters as such :

python -m train --project=commonlit --model_name=BERT --run_note=test_bert_1

The parameters can take different values :

• project : The project you want to train a model on, atm you can train a model on the aerial_cactus task, melanoma, blindness_detection & leaf_disease projects.
• model_name : You can choose any model that is in the models/ folder, name must be typed in MAJ like in the example above.
• run_note : A note or title to your run, it will be added to saved models name

Inference

To start prediction on new data for a project you can use this :

python -m predict --project=tweet_disaster --model_name=DISTILBERT --run_note=test

To do

---

• Add code the allow for Q&A projects
• Add scheduler
• Add more models
• Add more loss functions available
• Add metrics
• Add logger
• Add notebooks (for model evaluation - EDA - hyperparameter optimization etc...)