Building production LLM and NLP systems: retrieval, semantic search, embeddings, and agents that run at scale, not in a notebook.
Notes on LLM systems, semantic search, embeddings, retrieval, attention, PyTorch, and NLP engineering.
Concise explanations for specific technical problems, with examples and implementation details.
Understanding why we use negative infinity instead of zero for masking in transformer attention mechanisms. Learn about causal masking, softmax behavior, and practical implementation in PyTorch.
Learn about Unicode normalization, non-breaking spaces (\xa0), and how to handle text encoding issues in Python. Understand NFC, NFD, NFKD, and NFKC normalization forms for better text processing....
Reposts and adaptations of the notes above, syndicated to Medium and Towards AI. Start with Implementation Notes for the original technical depth.