nemo-evaluator 0.2.7

Common utilities for NVIDIA evaluation frameworks

NeMo Evaluator

For complete documentation, please see: docs/nemo-evaluator/index.md

Custom Benchmarks with BYOB

Create custom evaluation benchmarks in ~12 lines of Python using the BYOB (Bring Your Own Benchmark) framework:

from nemo_evaluator.contrib.byob import benchmark, scorer

@benchmark(name="my-qa", dataset="data.jsonl", prompt="Q: {question}\nA:", target_field="answer")
@scorer
def check(response: str, target: str, metadata: dict) -> dict:
    return {"correct": target.lower() in response.lower()}

# Compile and run
nemo-evaluator-byob my_benchmark.py
nemo-evaluator run_eval --eval_type byob_my_qa.my-qa --model_url http://localhost:8000 --model_id my-model

See the BYOB quickstart guide for full documentation, built-in scorers, and examples.