rohitg00/ai-engineering-from-scratch · error · ValueError

batch_size must be > 0, got {batch_size}

Error message

batch_size must be > 0, got {batch_size}

What it means

Error "batch_size must be > 0, got {batch_size}" thrown in rohitg00/ai-engineering-from-scratch.

Source

Thrown at phases/19-capstone-projects/49-lm-eval-harness/code/main.py:297

        for ex in examples:
            f.write(json.dumps({
                "id": ex.id,
                "prompt": ex.prompt,
                "targets": ex.targets,
                "metric": ex.metric,
                **({"extras": ex.extras} if ex.extras else {}),
            }) + "\n")


def run_task(
    task_name: str,
    examples: List[Example],
    adapter: ModelAdapter,
    *,
    batch_size: int = 8,
) -> TaskResult:
    if batch_size <= 0:
        raise ValueError(f"batch_size must be > 0, got {batch_size}")
    if not examples:
        return TaskResult(task=task_name, metric="none", score=0.0, correct=0, total=0)
    metric = examples[0].metric
    assert all(ex.metric == metric for ex in examples), f"task {task_name} mixes metrics"
    metric_fn = METRIC_FNS[metric]
    per_example: List[Dict[str, object]] = []
    correct_sum = 0.0
    total = 0
    start = time.perf_counter()
    for i in range(0, len(examples), batch_size):
        chunk = examples[i:i + batch_size]
        prompts = [ex.prompt for ex in chunk]
        outputs = adapter.generate(prompts)
        if len(outputs) != len(chunk):
            raise ValueError(
                f"adapter returned {len(outputs)} outputs for {len(chunk)} prompts in task {task_name}"
            )
        for ex, out in zip(chunk, outputs, strict=True):

View on GitHub (pinned to 39ea8a1c6d)

When it happens

Trigger: Thrown at phases/19-capstone-projects/49-lm-eval-harness/code/main.py:297 when the library encounters an invalid state.

Common situations: See trigger scenarios.

AI-assisted analysis of rohitg00/ai-engineering-from-scratch@39ea8a1c6d (2026-08-26). Data as JSON: /api/errors/6480d7af5aef14c6. Report an issue: GitHub.