Skip to content

CONTRIBUTING.md

Latest commit

 

History

History
54 lines (37 loc) · 1.7 KB

CONTRIBUTING.md

File metadata and controls

54 lines (37 loc) · 1.7 KB

Contributing

This repository is a research artifact first. Keep changes small, attributable, and easy to validate.

Local Setup

export CC=gcc
export CXX=g++
export PYO3_USE_ABI3_FORWARD_COMPATIBILITY=1
uv sync --extra dev

Default Validation

Run from the repository root:

uv run ruff check .
uv run pytest
uv run python -m telemetry_adaptation.cli --help

Smoke Benchmark Reproduction

Use the smoke config for the fastest end-to-end validation:

SMOKE_RUN_DIR="$(uv run python -m telemetry_adaptation.cli run-benchmark --config configs/benchmark/smoke.yaml --print-run-dir)"
uv run python -m telemetry_adaptation.cli summarize-results --input "$SMOKE_RUN_DIR" --output docs/results-smoke.md

This writes a timestamped run directory under artifacts/results/smoke/. Large raw benchmark outputs remain intentionally untracked, while compact evidence snapshots under docs/results-evidence/ and docs/results-smoke-evidence/ are kept in version control.

If you want to reproduce the accepted full benchmark that includes SMD, prepare the dataset first:

uv run python -m telemetry_adaptation.cli prepare-data --dataset smd --subset machine-1-1

Research Changes

If you change benchmark logic, policy behavior, or thesis-facing documentation:

  1. Read the current docs/results.md.
  2. Keep a local experiment log or design note outside tracked repository files.
  3. Preserve evidence for any new public claim before updating committed summaries.

Pull Requests

  • Prefer one bounded idea per pull request.
  • Include the exact validation commands you ran.
  • Do not commit large artifacts/results/ directories.
  • Do not rewrite unrelated user changes in a dirty worktree.