AI engineering glossary · Inference serving optimisation
What is per-channel quantisation?
Per-channel quantisation: Giving each output channel (row) of a weight matrix its own scale factor instead of one for the whole tensor. Because different channels have different magnitudes, this reduces rounding error noticeably and is a common default for weights.
Per-channel quantisation: an example
A layer with 4096 output channels stores 4096 scales alongside its INT8 weights.
Related terms
More Inference serving optimisation terms
- Per-tensor quantisation
- Percentile latency
- Post-training quantisation
- Preemption
- Prefill-decode disaggregation
- Prefix caching
- Prompt lookup decoding
- Pruning
Where you learn it
Per-channel quantisation is part of the AI Engineering Program, which teaches it with notes, a hands-on lab and an exam. The first lesson is free.