Largo Tutorials

Companion code for tutorials at largo.dev.

Practical, code-heavy implementations of ML concepts—embeddings, transformers, time series, edge deployment, and more.

Tutorials

Embeddings

• Sentence Embeddings from Scratch — Build a BiLSTM sentence encoder with PyTorch. Covers tokenization, embedding layers, and pooling strategies.

Predictive Maintenance

• Hard Drive Failure Prediction — Predict drive failures using Backblaze SMART data (24.8M records, Q4 2023). Compares XGBoost, LSTM, Transformer, and Mamba architectures.

  Model	Test AUC	F2	Recall	Notes
  XGBoost	0.920	0.095	54%	Feature engineering wins
  Transformer	0.916	0.006	80%	Conv1D preprocessing
  LSTM	0.907	0.006	78%	Bidirectional
  Mamba (SSM)	0.901	0.017	69%	Linear complexity O(n)

  Key techniques: balanced sampling, GPU acceleration with cuDF, Conv1D preprocessing, threshold optimization for extreme class imbalance (0.01% positive rate).

  New: model_mamba.py implements Mamba SSM using the official mamba-ssm library. All neural models achieve ~0.90 AUC (good ranking), but XGBoost with engineered features achieves 10x better F2.

Setup

# Clone the repo
git clone https://github.com/StoliRocks/largo-tutorials.git
cd largo-tutorials

# Install common dependencies
pip install -r requirements.txt

# Or install for a specific tutorial
pip install -r embeddings/sentence-embeddings-from-scratch/requirements.txt

Author

Steven W. White — largo.dev · LinkedIn

License

MIT