Caché multiquery
One KV head sharply reduces logical cache size.
Cálculo / simulación localExperimento educativo local, sin API externa. Los datos sintéticos no son mediciones de producción.
Study guide & lesson outline
LEARNING CONTRACTUnderstand → implement → investigate
Goal: explain the mechanism in this lesson, reproduce one example and test a failure case. Reading alone does not mark mastery.
- Understand: state assumptions, shapes/units and a counterexample.
- Implement: reproduce a numerical example and inspect boundaries.
- Investigate: compare one alternative under a matched budget.
02 · Predice y configura
Las notas técnicas, los parámetros y las respuestas están actualmente en inglés.
03 · Analiza la evidencia
Métodos y referencias
32 layers, head dimension 128. Dense cache only; excludes model weights, activations, block tables and sharing. Per-sequence allocation rounds up to the next page.
Text / 原始公式
KV bytes=2×layers×KV heads×head dimension×sequence×batch×bytes/value
An estimated cache footprint is not a measured GPU throughput or total memory requirement.
GQA (2023); PagedAttention (2023) ↗Preguntas de práctica
kv-3 / 01What does “Multi-query cache” test?
One KV head sharply reduces logical cache size.
Original teaching answer / local derivation. Method reference: GQA (2023); PagedAttention (2023) ↗
Abrir laboratorio →kv-3 / 02Which computation produces the result in “Multi-query cache”?
KV
Original teaching answer / local derivation. Method reference: GQA (2023); PagedAttention (2023) ↗
Abrir laboratorio →kv-3 / 03What is the baseline configuration for “Multi-query cache”?
{"sequence":4096,"batch":1,"heads":1,"bytes":2,"block":16}
Original teaching answer / local derivation. Method reference: GQA (2023); PagedAttention (2023) ↗
Abrir laboratorio →kv-3 / 04What numerical result should you check for “Multi-query cache”?
{"KV cache (GiB)":0.0625,"Allocated KV (GiB)":0.0625,"Padding tokens":0}
Original teaching answer / local derivation. Method reference: GQA (2023); PagedAttention (2023) ↗
Abrir laboratorio →kv-3 / 05How would you run a controlled comparison for “Multi-query cache”?
Keep the supplied fixture and other settings fixed; change sequence. Compare logical KV GiB, allocated GiB and padding tokens.
Original teaching answer / local derivation. Method reference: GQA (2023); PagedAttention (2023) ↗
Abrir laboratorio →kv-3 / 06What failure could invalidate a conclusion from “Multi-query cache”?
An estimated cache footprint is not a measured GPU throughput or total memory requirement.
Original teaching answer / local derivation. Method reference: GQA (2023); PagedAttention (2023) ↗
Abrir laboratorio →kv-3 / 07Is “Multi-query cache” a benchmark of a deployed language model?
No. 32 layers, head dimension 128. Dense cache only; excludes model weights, activations, block tables and sharing. Per-sequence allocation rounds up to the next page.
Original teaching answer / local derivation. Method reference: GQA (2023); PagedAttention (2023) ↗
Abrir laboratorio →kv-3 / 08Which evidence should you report for “Multi-query cache”?
Compare logical KV GiB, allocated GiB and padding tokens. Export inputs, engine, result and method; distinguish synthetic fixtures from observed production data.
Original teaching answer / local derivation. Method reference: GQA (2023); PagedAttention (2023) ↗
Abrir laboratorio →kv-3 / 09How can another researcher reproduce “Multi-query cache”?
Open kv-3, reset the preset, retain the exact inputs and run. The calculation is deterministic for identical inputs.
Original teaching answer / local derivation. Method reference: GQA (2023); PagedAttention (2023) ↗
Abrir laboratorio →kv-3 / 10What is the next validation step after “Multi-query cache”?
An estimated cache footprint is not a measured GPU throughput or total memory requirement. Test the hypothesis on a representative held-out workload and document a baseline, success metric and stopping rule before claiming deployment readiness.
Original teaching answer / local derivation. Method reference: GQA (2023); PagedAttention (2023) ↗
Abrir laboratorio →Assumptions, implementation & acceptance
Count both K and V, KV heads rather than query heads, and sequence×batch exactly once. The byte formula excludes activations, weights, allocator fragmentation and attention workspace.
Engine formulas, scenario presets and numerical safeguards reviewed. Seeded synthetic inputs are not a production benchmark.
Before running: predict a boundary case
One KV head sharply reduces logical cache size.
Hint: isolate the assumption
Keep the baseline input fixed, change one parameter or boundary, and compare both the numerical result and the claimed units. A passing example verifies this fixture only.
Minimal executable example
Save the example and linked modules in one folder. Use Node.js 22 or newer, then run node kv-3.mjs. No third-party runtime dependencies.
import {calculate,network} from './academy-core.mjs';
const params={
"sequence": 4096,
"batch": 1,
"heads": 1,
"bytes": 2,
"block": 16
};
const result=calculate('kv',params);
console.log(result.metrics);Test and transfer
One KV head sharply reduces logical cache size.
Record input, code version, expected result and observed result in the notebook below. Then construct a counterexample that violates the assumption above and explain why the original conclusion no longer follows.
Download complete kit (.tar.gz) · Complete reference kit and regression checksPredict, test and keep evidence
Before revealing an answer, predict one result. After the exercise, record the discrepancy and a transfer test. Saved locally across language editions.
Open notebook / export / clear →The mathematical review scope is recorded per lesson. It does not establish empirical paper reproduction; report suspected errors in your notes. Review scope