mudler/LocalAI · error · ValueError

Syntax error in inline reward function '{name}': {e}

Error message

Syntax error in inline reward function '{name}': {e}

What it means

compile_inline_reward takes user-supplied Python source defining a function named _user_reward_<name> and runs compile() on it before exec'ing in a restricted namespace. A SyntaxError from compile() is re-raised as ValueError with the offending function name and compiler message, so the caller knows which spec failed.

Source

Thrown at backend/python/trl/reward_functions.py:171

    Available modules: re, math, json, string.
    """
    func_source = (
        f"def _user_reward_{name}(completions, **kwargs):\n"
        + "\n".join(f"    {line}" for line in code.splitlines())
    )

    restricted_globals = {
        "__builtins__": _SAFE_BUILTINS,
        "re": re,
        "math": math,
        "json": json,
        "string": string,
    }

    try:
        compiled = compile(func_source, f"<inline-reward-{name}>", "exec")
    except SyntaxError as e:
        raise ValueError(f"Syntax error in inline reward function '{name}': {e}")

    exec(compiled, restricted_globals)
    func = restricted_globals[f"_user_reward_{name}"]

    # Validate with a quick smoke test
    try:
        result = func(["test"], answer=["test"])
        if not isinstance(result, list):
            raise ValueError(
                f"Inline reward function '{name}' must return a list, got {type(result).__name__}"
            )
    except Exception as e:
        if "must return a list" in str(e):
            raise
        # Other errors during smoke test are acceptable (e.g. missing kwargs)
        pass

    return func

View on GitHub (pinned to 44413a9d06)

Solutions

1. Reproduce locally: python -c "compile(open('reward.py').read(), 'x', 'exec')" to find the exact line.
2. Fix the syntax and ensure the code defines _user_reward_<name> exactly (the compile step also expects that symbol to exist afterwards).
3. Check the JSON encoding of the code string — real newlines must survive as \n in the JSON payload.

Example fix

# before
code = "def _user_reward_len(c, **kw) return [float(len(x)) for x in c]"  # missing colon
# after
code = "def _user_reward_len(c, **kw):\n    return [float(len(x)) for x in c]"

def reward_code_compiles(name: str, code: str) -> bool:
    try:
        compile(code, f"<inline-reward-{name}>", "exec")
        return True
    except SyntaxError:
        return False

try:
    func = compile_inline_reward(name, code)
except ValueError as e:
    if "Syntax error" in str(e):
        return validation_error(field="reward_funcs", detail=str(e))
    raise

When it happens

Trigger: An inline reward spec whose code has a missing colon/indent, uses print without import under Python 2 habits, or was truncated by shell quoting/JSON escaping (e.g. lost newlines).

Common situations: Passing code through multiple JSON layers that eat backslash-n; writing the function body without the required def _user_reward_<name>() signature mismatch causing parse errors; copy-paste from notebooks with smart quotes.

Related errors

• Inline reward function '{name}' must return a list, got {typ
• Inline reward function '{name}' has no code
• GRPO requires at least one reward function. Specify reward_f
• reward_funcs must be a JSON array of reward function specs
• Unknown builtin reward function '{name}'. Available: {availa

AI-assisted analysis of mudler/LocalAI@44413a9d06 (2026-08-15). Data as JSON: /api/errors/b57417feb62beee0. Report an issue: GitHub.