sgl-project/sglang · error · ComponentCheckpointUnsupportedError

Cannot configure checkpoint quantization for {component_name

Error message

Cannot configure checkpoint quantization for {component_name!r}: {error}

What it means

Wrapper error: the underlying checkpoint-quantization parser (get_quantization_config path) raised KeyError/NotImplementedError/TypeError/ValueError while deriving a quant config for the component, and it is re-raised as ComponentCheckpointUnsupportedError with the component name and original error chained. The root cause is in {error}.

Source

Thrown at python/sglang/multimodal_gen/runtime/loader/component_loaders/text_encoder_loader.py:283

            )
        # Preserve model-owned formats such as Ideogram's bitsandbytes state.
        # Those models parse metadata, construct layers, and attach quant states
        # themselves; running the generic lifecycle as well would process twice.
        return

    _delegate_standard_bnb4_to_transformers(
        component_config,
        component_name,
    )
    try:
        quant_config = _get_encoder_quant_config(
            component_config,
            component_model_path,
            component_weights_path,
            model_cls,
        )
    except (KeyError, NotImplementedError, TypeError, ValueError) as error:
        raise ComponentCheckpointUnsupportedError(
            f"Cannot configure checkpoint quantization for {component_name!r}: {error}"
        ) from error
    model_config.quant_config = quant_config
    if explicit_quantization is not None:
        if quant_config is not None:
            raise ComponentCheckpointUnsupportedError(
                f"{component_name!r} already declares checkpoint quantization; "
                "drop the explicit online quantization override"
            )
        if explicit_quantization not in _ONLINE_ENCODER_QUANTIZATIONS:
            raise ComponentCheckpointUnsupportedError(
                f"Online quantization {explicit_quantization!r} is not supported "
                f"for native encoders; choose one of "
                f"{sorted(_ONLINE_ENCODER_QUANTIZATIONS)}"
            )
        from sglang.multimodal_gen.runtime.layers.quantization import (
            get_quantization_config,
        )

View on GitHub (pinned to 0132848349)

Solutions

1. Read the chained {error} text to identify the underlying exception and fix that (e.g. correct the quantization_config field)
2. Regenerate or re-download the checkpoint config so the quantization_config matches a supported format
3. If the format is genuinely unsupported, convert/dequantize the checkpoint or use a supported quantization variant
4. As a last resort remove quantization_config to load unquantized weights

Example fix

// before
{"quantization_config": {"quant_method": "my_custom_fmt"}}

// after
{"quantization_config": {"quant_method": "fp8", "fmt": "e4m3"}}

import json
cfg = json.loads((model_path / "config.json").read_text())
qc = cfg.get("quantization_config")
assert qc is None or qc.get("quant_method") in SUPPORTED_METHODS, f"unsupported: {qc}"

try:
    _configure_encoder_quantization(...)
except ComponentCheckpointUnsupportedError as e:
    if "Cannot configure checkpoint quantization" in str(e):
        logger.error("bad quant metadata: %s", e.__cause__)
        raise

When it happens

Trigger: Loading a component whose config/weights metadata is malformed or uses an unrecognized quantization format: missing keys in quantization_config (KeyError), unknown quant method (NotImplementedError), wrong config types (TypeError), or invalid values (ValueError) inside _configure_encoder_quantization.

Common situations: Checkpoint saved by a newer/older library version with a quantization_config schema this loader can't parse; hand-edited config.json; exotic quant formats (e.g. compressed-tensors variants) not registered; truncated weight-file metadata.

Related errors

• Cannot parse checkpoint quantization for {component_name!r}:
• The SRT encoder checkpoint adapter supports only serialized
• Serialized quantized component weights cannot use a stacked
• A GGUF encoder checkpoint cannot be combined with a second q
• {component_name!r} manages its own checkpoint quantization a

AI-assisted analysis of sgl-project/sglang@0132848349 (2026-08-28). Data as JSON: /api/errors/76050f6c7695b6e6. Report an issue: GitHub.