huggingface/transformers · error · ValueError

seeding_scheme has to be one of [`selfhash`, `lefthash`], bu

Error message

seeding_scheme has to be one of [`selfhash`, `lefthash`], but found {seeding_scheme}

What it means

Error "seeding_scheme has to be one of [`selfhash`, `lefthash`], but found {seeding_scheme}" thrown in huggingface/transformers.

Source

Thrown at src/transformers/generation/logits_process.py:2462

    >>> detector = WatermarkDetector(model_config=model.config, device="cpu", watermarking_config= watermarking_config)
    >>> detection_preds = detector(out)
    >>> detection_preds
    array([ True])
    ```
    """

    def __init__(
        self,
        vocab_size,
        device,
        greenlist_ratio: float = 0.25,
        bias: float = 2.0,
        hashing_key: int = 15485863,
        seeding_scheme: str = "lefthash",
        context_width: int = 1,
    ):
        if seeding_scheme not in ["selfhash", "lefthash"]:
            raise ValueError(f"seeding_scheme has to be one of [`selfhash`, `lefthash`], but found {seeding_scheme}")
        if greenlist_ratio >= 1.0 or greenlist_ratio <= 0.0:
            raise ValueError(
                f"greenlist_ratio has be in range between 0.0 and 1.0, exclusively. but found {greenlist_ratio}"
            )

        self.vocab_size = vocab_size
        self.greenlist_size = int(self.vocab_size * greenlist_ratio)
        self.bias = bias
        self.seeding_scheme = seeding_scheme
        self.rng = torch.Generator(device=device)
        self.hash_key = hashing_key
        self.context_width = context_width

        self.rng.manual_seed(hashing_key)
        self.table_size = 1_000_003
        self.fixed_table = torch.randperm(self.table_size, generator=self.rng, device=device)

    def set_seed(self, input_seq: torch.LongTensor):

View on GitHub (pinned to a597f97485)

Solutions

1. Set `seeding_scheme` to 'selfhash' or 'lefthash' in the watermarking config.
2. Use the default `WatermarkingConfig()` which already picks a valid scheme.

When it happens

Trigger: Raised in WatermarkLogitsProcessor.__init__ when seeding_scheme is neither 'selfhash' nor 'lefthash'.

Common situations: Typo or unsupported value in WatermarkingConfig.seeding_scheme when enabling text watermarking.

AI-assisted analysis of huggingface/transformers@a597f97485 (2026-08-14). Data as JSON: /api/errors/fba97ac9cd766a30. Report an issue: GitHub.