roboflow/supervision · error · ValueError

Invalid vlm value: {vlm}. Must be one of {[e.value for e in

Error message

Invalid vlm value: {vlm}. Must be one of {[e.value for e in VLM]}

What it means

Raised by the VLM argument validator in supervision's VLM module when the vlm argument is a string that does not match any VLM enum value after lowercasing. Supported values mirror VLM members (e.g. 'paligemma', 'florence-2', 'qwen-2.5-vl', 'qwen-3-vl', 'deepseek-vl2'). A VLM enum instance or an exactly-matching string are the only accepted inputs.

Source

Thrown at src/supervision/detection/vlm.py:190

    """
    Validates the parameters and result type for a given Vision-Language Model (VLM).

    Args:
        vlm: The VLM enum or string specifying the model.
        result: The result object to validate (type depends on VLM).
        kwargs: Dictionary of arguments to validate against required/allowed lists.

    Returns:
        The validated VLM enum value.

    Raises:
        ValueError: If the VLM, result type, or arguments are invalid.
    """
    if isinstance(vlm, str):
        try:
            vlm = VLM(vlm.lower())
        except ValueError:
            raise ValueError(
                f"Invalid vlm value: {vlm}. Must be one of {[e.value for e in VLM]}"
            )

    if not isinstance(result, RESULT_TYPES[vlm]):
        raise ValueError(
            f"Invalid VLM result type: {type(result)}. Must be {RESULT_TYPES[vlm]}"
        )

    required_args = REQUIRED_ARGUMENTS.get(vlm, [])
    for arg in required_args:
        if arg not in kwargs:
            raise ValueError(f"Missing required argument: {arg}")

    allowed_args = ALLOWED_ARGUMENTS.get(vlm, [])
    for arg in kwargs:
        if arg not in allowed_args:
            raise ValueError(f"Argument {arg} is not allowed for {vlm.name}")

View on GitHub (pinned to 7f254d9784)

Solutions

1. Pass the enum member instead of a string: sv.detection.vlm.VLM.FLORENCE_2, which is immune to spelling issues.
2. If using a string, copy it exactly from the error message's supported list (e.g. 'florence-2').
3. Check the VLM enum in your installed version: python -c "from supervision.detection.vlm import VLM; print(VLM.list())".
4. Upgrade supervision if you expect a model whose value is not in the list.

Example fix

# before
 detections = sv.Detections.from_vlm(vlm="florence2", result=result)

# after
 from supervision.detection.vlm import VLM
 detections = sv.Detections.from_vlm(vlm=VLM.FLORENCE_2, result=result)

from supervision.detection.vlm import VLM

valid = {e.value for e in VLM}
if vlm_name.strip().lower() not in valid:
    raise ValueError(f"Unsupported VLM {vlm_name!r}; valid: {sorted(valid)}")

def is_valid_vlm_name(name: str) -> bool:
    return name.strip().lower() in {e.value for e in VLM}

try:
    detections = sv.Detections.from_vlm(vlm=vlm_name, result=result)
except ValueError as e:
    raise SystemExit(f"Configuration error: {e}") from e

When it happens

Trigger: Calling sv.Detections.from_vlm(vlm='florence2', result=...) (missing hyphen), 'Qwen2.5-VL' style names, or a made-up model name; the lookup VLM(vlm.lower()) raises ValueError internally which is re-raised with the supported list.

Common situations: Using the HuggingFace model id ('microsoft/Florence-2-large') instead of the enum value; version drift when new VLMs were added/renamed across supervision releases; copy-pasting names from older docs that used the deprecated LMM enum values.

Related errors

• Unsupported VLM value: {vlm}.
• Expected string as {task} result, got {type(result)}
• Invalid value: {value}. Must be one of {cls.list()}
• Invalid value type: {type(value)}. Must be an instance of {c
• Unsupported color lookup strategy: {color_lookup}

AI-assisted analysis of roboflow/supervision@7f254d9784 (2026-08-15). Data as JSON: /api/errors/1f9b11f0f2606b8b. Report an issue: GitHub.