unslothai/unsloth · error · ValueError

weighting_scheme must be one of none / bell

Error message

weighting_scheme must be one of none / bell

What it means

weighting_scheme selects the loss weighting strategy and only supports 'none' (uniform) and 'bell' (bell-shaped timestep weighting). Any other string after strip/lower normalization is rejected.

Source

Thrown at studio/backend/core/training/diffusion_train_common.py:1169

                    raise ValueError(
                        f"flow_shift must be a positive number or 'auto', got {self.flow_shift!r}"
                    ) from exc
        if not isinstance(flow_shift, str):
            flow_shift = float(flow_shift)
            # isfinite as well as positive: JSON accepts 1e309, which floats to inf and would poison every sampled sigma while progress looks normal.
            if not math.isfinite(flow_shift) or flow_shift <= 0:
                raise ValueError(
                    "flow_shift must be a finite number > 0 (1.0 disables the shift), or 'auto'"
                )
        try:
            cfg_dropout = float(self.cfg_dropout or 0.0)
        except (TypeError, ValueError) as exc:
            raise ValueError(f"cfg_dropout must be a number, got {self.cfg_dropout!r}") from exc
        if not 0.0 <= cfg_dropout <= 1.0:
            raise ValueError("cfg_dropout must be between 0 and 1")
        weighting_scheme = str(self.weighting_scheme or "none").strip().lower()
        if weighting_scheme not in ("none", "bell"):
            raise ValueError("weighting_scheme must be one of none / bell")
        # A zero/negative gamma would zero out (or invert) the min-SNR weight and silently train on a degenerate loss; None is the documented disable.
        if self.snr_gamma is not None and float(self.snr_gamma) <= 0:
            raise ValueError("snr_gamma must be > 0, or null to disable min-SNR weighting")
        # learning_rate can arrive as a string ("1e-4") from the Studio config path, so coerce it before AdamW sees it.
        try:
            learning_rate = float(self.learning_rate)
        except (TypeError, ValueError) as exc:
            raise ValueError(f"learning_rate must be a number, got {self.learning_rate!r}") from exc
        if learning_rate <= 0:
            raise ValueError("learning_rate must be > 0")
        alpha = self.lora_alpha if self.lora_alpha is not None else self.lora_rank
        targets = tuple(self.lora_target_modules) or DEFAULT_LORA_TARGETS
        # A blank Hub token (the Studio default when none is configured) must load anonymously, not as an explicit empty credential.
        token = self.hf_token.strip() if isinstance(self.hf_token, str) else self.hf_token
        from core.inference.diffusion_families import (
            _is_local_path,
            mirror_repo,
            prefer_ungated_mirror,

View on GitHub (pinned to 203007d190)

Solutions

1. Use weighting_scheme='none' or 'bell'.
2. If you ported a diffusers example config, drop its weighting_scheme value — the closest supported behavior here is 'bell'.
3. Leave it unset; the default resolves to 'none'.

Example fix

# before
cfg = DiffusionLoraConfig(weighting_scheme='zero_snr')
# after
cfg = DiffusionLoraConfig(weighting_scheme='bell')

ALLOWED_WEIGHTING = {'none', 'bell'}
ws = str(weighting_scheme or 'none').strip().lower()
assert ws in ALLOWED_WEIGHTING, f'weighting_scheme must be one of {sorted(ALLOWED_WEIGHTING)}'

def is_valid_weighting_scheme(v) -> bool:
    return str(v or 'none').strip().lower() in ('none', 'bell')

When it happens

Trigger: weighting_scheme='simsnr', 'sigma-age', 'uniform', or any typo like 'bel' / 'Bell ' with stray characters that do not normalize to the two allowed values.

Common situations: Porting configs from other trainers (diffusers examples support schemes like 'sigma' or 'zero_snr' / 'zsnr') that this trainer does not implement.

Related errors

• snr_gamma must be > 0, or null to disable min-SNR weighting
• base_precision={base_precision!r} trains in bf16 compute; se
• flow_shift must be a positive number or 'auto', got {self.fl
• cfg_dropout must be a number, got {self.cfg_dropout!r}
• cfg_dropout must be between 0 and 1

AI-assisted analysis of unslothai/unsloth@203007d190 (2026-08-15). Data as JSON: /api/errors/fd62990ecc3ddb9b. Report an issue: GitHub.