Operations
Use this page for operational folder hygiene, validation checks, and release/security workflows. Prerequisite: project dependencies installed in your active environment.
Artifact Directories
data/: prepared datasets, includingdata/meta.jsonforbinruns.checkpoints/: training outputs (ckpt_last.pt,train_log.json).outputs/: exported artifacts (hf_export/,gguf/, fine-tuning outputs).runs/: optional experiment logs or external tracker exports (not auto-created by core scripts).
Note
The core training scripts write to checkpoints/ and outputs/. Keep those directories versioned in your run notes, but do not commit large artifacts.
Command(s)
Local quality checks:
python -m pytest -q
ruff check src scripts tests train.py generate.py demo_gradio.py --select E9,F63,F7,F82
Docs build checks:
Output Files / Artifacts Produced
- Test and lint logs in terminal output
- Built docs in
site/aftermkdocs build - CI pipelines:
.github/workflows/ci.ymland.github/workflows/docs.yml
Security and Release
Security reporting (see SECURITY.md):
- Use GitHub Security Advisories for vulnerabilities
- Include impact, affected components, and reproduction steps
Release flow (see RELEASE.md):
- Confirm CI is green.
- Run local validation commands.
- Update version and changelog.
- Tag and publish the release.
Common Errors
- Missing dependencies: see Torch not installed.
- Metadata and path confusion: see Meta path mismatch.
- CUDA expected but unavailable: see CUDA not detected.