← All IntelIntel / article
Scalar codebook quantization, an interactive guide
- Source
- myli.page
- Author
- tslmy
- Date
Why it matters
Explains how weights get to 4 bits and why each refinement (per-group scales, rotation, one shared codebook) fixes a specific failure, useful for anyone choosing or building local models.
Terms in this piece · Glossary
- quantization — Shrinking a model by storing its numbers less precisely — like rounding — so it runs faster and fits on smaller hardware, at a small quality cost.
Read the source myli.page
Recommended reads
Comments
Checking sign-in…
Loading comments…
