- Thou shalt not hedge. Do not dilute answers, alter scope, or change requirements without explicit confirmation.
- Thou shalt not mix test frameworks. Only
pytest(and its plugins) shall be used. No others. - Thou shalt confirm before simplifying. If you are about to suggest a “simpler approach,” pause and confirm with the user first.
- Thy tests shall be complete and meaningful. Unit tests must cover expected behavior thoroughly, not trivialities.
- Thou shalt name tests precisely. Test names must reference the class or method under test, with situational detail if necessary.
- Thou shalt not abuse mocks. If tests rely excessively on mocks or re-implement functionality, consider integration tests instead.
- Thou shalt not self-promote. Do not add yourself as co-author or PR agent in release notes.
- Thou shalt re-run all tests after major changes. Never declare completion without a full suite pass.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| #!/usr/bin/env bash | |
| set -euo pipefail | |
| MODE="${1:-}" | |
| # Headless CUDA/job host package matcher. | |
| # Intentionally includes: | |
| # - nvidia driver/kernel/dkms/open modules | |
| # - libnvidia userspace/CUDA-facing libs | |
| # - libcuda |
Transform any collection of images into a LoRA training dataset. The tool automatically:
- Analyzes images using local LLM (Gemma via LM Studio)
- Generates subject-only captions (removes all style references)
- Creates training pairs (image + caption describing content only)
- Organizes dataset with proper structure and documentation