Skip to content

README.md

Latest commit

 

History

History
30 lines (20 loc) · 1.1 KB

README.md

File metadata and controls

30 lines (20 loc) · 1.1 KB

TurboQuant Benchmarks

This directory contains deterministic, NumPy-only TurboQuant benchmark runners.

Current contents:

These runners are validation tooling only. They are not the package KV preview API and do not define TurboQuantKVPreviewCodec.compress, .decompress, or .memory_stats.

Non-goals:

  • real-model downloads
  • runtime adapters
  • hardware probing
  • machine-specific artifact storage

Typical usage from the package root:

PYTHONPATH=src python benchmarks/turboquant_paper_validation.py --output /tmp/turboquant-paper.json
PYTHONPATH=src python benchmarks/turboquant_synthetic_kv_benchmark.py --output /tmp/turboquant-kv.json

The benchmark report in this directory summarizes measured outputs from those runners and presents the current compression results in Markdown form.

These runners are intended for local validation and CI-adjacent smoke usage. They must remain deterministic by seed and must not require network access.