Model Evaluation Reference#
Lookup pages for eval/model_eval.
For the section overview, refer to About Model Evaluation.
For procedural walk-throughs, refer to Model Evaluation How-To Guides.
Configuration Reference
YAML schema for default.yaml and tiny_chat.yaml, field by field.
CLI Reference
nemotron steps run eval/model_eval flags and Hydra overrides.
Output Artifacts
The eval_results contract and the on-disk directory layout.
Benchmarks Catalog
Benchmark identifiers grouped by family, with endpoint-type guidance.
Troubleshooting
Named error modes from step.toml, with the most common cause and the recovery for each.