Metrics

TODO:

• [ ] Add page-specific notes for ai-and-models/training-pipeline/metrics.md after inspecting the real project files.
• [ ] Cross-link Metrics to the most relevant code, data, testing, and safety pages.
• [ ] Document the exact training or inference file related to this page.
• [ ] List relevant command-line flags and their current intended values.
• [ ] Document input tensor shape, label format, and output steering range where relevant.
• [ ] Explain what changed between model versions if this page covers a model.
• [ ] Add the dataset names used and whether any labels are historical or current.
• [ ] Add offline metrics that belong on the page once retested.
• [ ] Add field behavior that must be checked before trusting the model.
• [ ] Add known failure cases and what data would improve them.
• [ ] Reference code/ai_models_data/train_autonomy_v2.py where relevant.
• [ ] Reference code/ai_models_data/steering_corrections.json where relevant.
• [ ] List model file path, version, preprocessing path, output scale, and known field behavior.
• [ ] Add current vs historical metrics once retesting is complete.
• [ ] Add the full training command with every flag.