sgl-project/sglang · error · ValueError

Invalid quantization method: {quantization}

Error message

Invalid quantization method: {quantization}

What it means

get_quantization_config looks up the requested quantization method in QUANTIZATION_METHODS and raises when the name is not registered. This is the standard entry point mapping a quantization string (e.g. from model config or CLI) to its QuantizationConfig class.

Source

Thrown at python/sglang/multimodal_gen/runtime/layers/quantization/__init__.py:94

    def _wrapper(quant_config_cls):
        if quantization in QUANTIZATION_METHODS:
            raise ValueError(
                f"The quantization method `{quantization}` is already exists."
            )
        if not issubclass(quant_config_cls, QuantizationConfig):
            raise ValueError(
                "The quantization config must be a subclass of " "`QuantizationConfig`."
            )
        _CUSTOMIZED_METHOD_TO_QUANT_CONFIG[quantization] = quant_config_cls
        QUANTIZATION_METHODS.append(quantization)
        return quant_config_cls

    return _wrapper


def get_quantization_config(quantization: str) -> type[QuantizationConfig]:
    if quantization not in QUANTIZATION_METHODS:
        raise ValueError(f"Invalid quantization method: {quantization}")

    method_to_config: dict[str, type[QuantizationConfig]] = {}
    # Update the `method_to_config` with customized quantization methods.
    method_to_config.update(_CUSTOMIZED_METHOD_TO_QUANT_CONFIG)

    return method_to_config[quantization]


__all__ = [
    "QuantizationMethods",
    "QuantizationConfig",
    "get_quantization_config",
    "QUANTIZATION_METHODS",
]

View on GitHub (pinned to 0132848349)

Solutions

1. Print/inspect QUANTIZATION_METHODS and correct the string to a registered name
2. If it's a custom method, ensure the module calling register_quantization is imported before lookup
3. Upgrade/downgrade to the version that supports the method you need

Example fix

// before
config = get_quantization_config("fpx")  # typo

// after
config = get_quantization_config("fp8")

from ...quantization import QUANTIZATION_METHODS
if quantization not in QUANTIZATION_METHODS:
    raise SystemExit(f"Unknown quant method {quantization!r}; valid: {QUANTIZATION_METHODS}")

try:
    cfg = get_quantization_config(name)
except ValueError as e:
    # fall back to a default or exit with a clear message
    ...

When it happens

Trigger: Calling get_quantization_config("some_method") where the string is not in QUANTIZATION_METHODS — typo'd names, unregistered custom methods, or methods only available in a newer/older version.

Common situations: Typo in a CLI --quantization flag or model config quantization_config.quant_method; using a quant method whose plugin isn't imported; version mismatch where the method was added or removed.

Related errors

• The quantization method `{quantization}` is already exists.
• The quantization config must be a subclass of `QuantizationC
• kitchen_w4a4 is inferred from per-layer checkpoint metadata;
• GGUFConfig must be constructed from a GGUF checkpoint
• f"Cannot parse checkpoint quantization for {component_name!r

AI-assisted analysis of sgl-project/sglang@0132848349 (2026-08-28). Data as JSON: /api/errors/59d675646173ec02. Report an issue: GitHub.