AI Glossary

Model compression

What does “Model compression” mean?

Model compression: The family of techniques for making models smaller, faster or cheaper to run while keeping as much quality as possible. Quantization, pruning and distillation are the main approaches, and they are often combined.

Model compression visual guide: The family of techniques for making models smaller, faster or cheaper to run while keeping as much quality as possible. Quantization, pruning and distillation are the main approaches, and they are often combined.
Visual guide Open full size (new tab)

What is an example of “Model compression”?

Compression is what lets features first built on data-center hardware eventually move onto phones.

Keep learning

See also

QuantizationTermPruningTermKnowledge distillationTermOn-device AITerm

Same topic

GPU memoryTermModel servingTermInterconnectTermNPUTermInference serverTerm