Summary

Implement a hierarchical configuration management system for StreetScene using OmegaConf/Hydra-style YAMLs. Add task-level and experiment configs, a CLI entry point to merge/validate overrides, and automatic resolution artifact saving to improve reproducibility.

Details

• Introduced ConfigManager in PyTorch/StreetScene/src/common/config.py to load defaults, task configs, and experiment configs; apply CLI and environment overrides; and produce a resolved config artifact.
• Added configs:
  • defaults.yaml
  • tasks/detection/pedestrian.yaml
  • tasks/detection/vehicle.yaml
  • tasks/classification/vehicle.yaml
  • experiments/baseline.yaml
  • experiments/large_scale.yaml
• Implemented CLI entry point scripts/run_experiment.py to merge configs, apply CLI/env overrides, validate required fields, and emit a resolved YAML artifact under outputs/.
• Implemented environment override handling using APP_ prefix with nested keys using double underscores, with safety checks to avoid unknown keys leading to spurious fields.
• Ensured resolved configs are stored for reproducibility with a run-name derived artifact.
• Updated documentation: README.md and IMPLEMENTATION.md provide usage and design details.

Warning: Task VM test is not passing, cto.new will perform much better if you fix the setup