Interactive Jupyter notebooks demonstrating the LLM framework capabilities.
| Notebook | Description |
|---|---|
| quick_start.ipynb | Complete tutorial: model building, training, inference, advanced features |
# Install Jupyter if needed
uv pip install jupyter
# Launch Jupyter
cd notebooks
jupyter notebook- Building a Model - Create a Decoder-only Transformer
- Training - Train on synthetic data with loss visualization
- Inference - Text generation (streaming and non-streaming)
- Advanced Features:
- Gradient Checkpointing (memory-efficient training)
- Grouped Query Attention (GQA)
- SwiGLU Activation
- Mixture of Experts (MoE)
- E2E Pipeline - Full train → evaluate → inference workflow