NLP-GENIUS 🎵 ✨

NLP-GENIUS is a cutting-edge natural language processing project that combines the power of transformers and machine learning to create an innovative music technology suite. It features:

🎹 Title to Lyrics Generator (Transformer)

Transform any title into complete song lyrics using a fine-tuned GPT-2 model trained on over 1M songs. The model captures musical patterns and lyrical structures to generate coherent and creative lyrics that match the input title's theme.

🎯 Genre Classifier

An advanced machine learning classifier that can predict a song's genre based on its lyrics. Using a combination of logistic regression and naive Bayes algorithms, it achieves impressive accuracy in categorizing songs into different musical styles.

🔬 Key Technical Features

• Fine-tuned GPT-2 transformer model for lyric generation
• Sophisticated text preprocessing pipeline
• Multi-model ensemble classification
• Docker containerization for easy deployment
• Comprehensive evaluation metrics and benchmarking

Title to Lyrics Transformer

Getting the weights

Use the pre-computed weights

• A long training has already been made over 1M lyrics (RTX 4090 for 18 hours) :
• Download the weights here.
• Extract the archive in the transformer directory. Your file tree should look like this :

/NLP-GENIUS$ tree -L 2
.
├── README.md
├── dataset
│   └── song_lyrics.csv
 ...

└── transformers
    ├── Dockerfile
    ├── __pycache__
    ├── classes.py
    ├── preprocessing.py
    ├── requirements.txt
    ├── title_to_lyrics_transformer.py
    ├── trained_model # <- Extracted archive
    └── training.py

.. Or train from scratch

Configure a .env file in the transformer directory. The following config will train the model for 10 epochs over 1000 rows :

# /NLP_GENIUS/transformers/.env
TSF_TRAINING=True
TSF_EPOCHS=10
TSF_ROWS=1000

Then install the dependencies and run the python script :

/NLP-GENIUS/transformers$ pip install -r requirements.txt
/NLP-GENIUS/transformers$ python title_to_lyrics_transformer.py

Generating lyrics

Build the Docker image

Make sure to be at the root of the project. First image build will take some time (copying the whole dataset) ~ 12mn.

 pwd # /NLP-GENIUS

 docker build -t nlp_transformer -f transformers/Dockerfile .

Run the Docker container

docker run -e TSF_TITLE="Hello darkness" -it nlp_transformer