Optimization and training dynamics · MLD-0349
Effective batch size
The total amount of data contributing to a parameter update across accumulation and workers.
Original reference page · detailed expansion pending. Original definitions, examples, notes and citations are preserved below. New examples are illustrative, not benchmark results.
Original example
Combine per-device batches across eight GPUs.
Original caveat
Variable-length token batches need a clear counting convention.
Context and aliases
Level: Intermediate. Stages: Training · Theory.
Tags: ML · DL · SLM · LLM.
Aliases: None recorded.
Related concepts
Concepts linking here
Original references
- Deep Learning: online textbook ↗ · Consulted
- PyTorch optimization algorithms ↗ · Reference link
- PyTorch documentation ↗ · Reference link
- Original reference URL ↗