Internal docs for the quant-sim codebase. User-facing docs live in the root README.md.
| Doc | What |
|---|---|
| ARCHITECTURE.md | How the benchmark works end-to-end: model discovery, speed tests, quality tests, recommendation algorithm, leaderboard |
| reference/CODEBASE-MAP.md | Every file, every function, line counts |
| reference/QUALITY-TESTS.md | All 20 quality test questions, categories, grading rules |
- PyPI package:
quantsim-bench(v0.1.0) - CLI command:
quant-sim - Entry point:
quant_sim.cli:main - Dependencies:
requests(only runtime dep). Optional:matplotlibfor charts. - Requires: Ollama running locally, NVIDIA GPU with nvidia-smi
- Tests: 13 (pytest), all in
tests/test_quality.py - Total source: ~864 lines across 7 Python files
Finished product. Published on PyPI. No further development planned.