Implementing core DL architectures and research papers from scratch — because theory without code is just reading.
Built under Eunice Labs.
| Concept | Notebook | Explanation |
|---|---|---|
| LSTM | basic_LSTM.ipynb | — |
| Seq2Seq | seq2seq_tutorial.ipynb | seq2seq.md |
| Bahdanau Attention | bahdanau.ipynb | bahdanau_explained.md |
| Luong Attention | luong.ipynb | luong_explained.md |
| Self Attention | self_attention.ipynb | self_attention_explained.md |
| Transformer | transformer.ipynb | transformer_explained.md |
Most DL courses stop at theory. This repo is the other half — every major architecture implemented in PyTorch, with explanations written to actually make sense.
The goal isn't to reproduce papers. It's to understand them well enough to break them.
- PyTorch
- Jupyter Notebooks
- Python 3.10+
- Multi-head attention
- Vision Transformer (ViT)
- RLHF / PPO from scratch
- Mixture of Experts
Gagan P — Founder, Eunice Labs · AI Researcher
If this helped you understand something better, star the repo.