Overview
This project was developed for the Introduction to Deep Learning and GenAI course at IIT Madras. The task was multi-label emotion classification — predicting multiple emotions simultaneously from input text.
Approach
- Framed the problem as a multi-label classification task with binary cross-entropy loss
- Used PyTorch Lightning to structure the training loop cleanly
- Tracked experiments, metrics, and model checkpoints with Weights & Biases (wandb)
Tech Stack
- PyTorch — model definition and training
- PyTorch Lightning — training framework
- Weights & Biases (wandb) — experiment tracking and visualization
- Python — core implementation