Summary

• extract Lightning runtime orchestration and data module factory into dedicated modules (connectomics/training/lit/runtime.py, connectomics/training/lit/data_factory.py)
• reduce duplication in lit utils/config flows and add contract/unit coverage around CLI/run-directory behavior
• migrate tutorial config set to new names and phase-1 schema compliance (test.data, augmentation preset semantics, TTA channel activations)
• add YAML _base_ inheritance support in config loading with cycle detection
• extract shared tutorial bases under tutorials/bases/ and slim top-level tutorials/*.yaml while preserving resolved behavior
• add scripts/validate_tutorial_configs.py, inheritance tests, and CI validation step

Validation

• python scripts/validate_tutorial_configs.py
• pytest -q tests/unit/test_hydra_utils_inheritance.py tests/unit/test_main_cli_contract.py
• pytest -q tests/unit/test_hydra_config.py

Notes

• top-level tutorial configs are documented in tutorials/README.md
• branch includes prior milestone commits for lit runtime/data factory extraction and related tests