invoke-ai/InvokeAI · error · ValueError

cfg_scale list has {len(self.cfg_scale)} values but the mode

Error message

cfg_scale list has {len(self.cfg_scale)} values but the model is configured for {num_timesteps} steps. Provide one CFG value per configured step (or a single float).

What it means

The Krea-2 denoise invocation accepts cfg_scale either as a single float (broadcast to all steps) or as a per-step list. `_prepare_cfg_scale` throws when a list is provided whose length does not equal the model's configured number of timesteps, because CFG values are consumed one per denoising step.

Source

Thrown at invokeai/app/invocations/krea2_denoise.py:210

    def _get_noise(self, height: int, width: int, dtype: torch.dtype, device: torch.device, seed: int) -> torch.Tensor:
        rand_device = "cpu"
        return torch.randn(
            1,
            KREA2_LATENT_CHANNELS,
            int(height) // LATENT_SCALE_FACTOR,
            int(width) // LATENT_SCALE_FACTOR,
            device=rand_device,
            dtype=torch.float32,
            generator=torch.Generator(device=rand_device).manual_seed(seed),
        ).to(device=device, dtype=dtype)

    def _prepare_cfg_scale(self, num_timesteps: int) -> list[float]:
        if isinstance(self.cfg_scale, float):
            return [self.cfg_scale] * num_timesteps
        if isinstance(self.cfg_scale, list):
            if len(self.cfg_scale) != num_timesteps:
                raise ValueError(
                    f"cfg_scale list has {len(self.cfg_scale)} values but the model is configured for "
                    f"{num_timesteps} steps. Provide one CFG value per configured step (or a single float)."
                )
            return self.cfg_scale
        raise ValueError(f"Invalid CFG scale type: {type(self.cfg_scale)}")

    @staticmethod
    def _should_apply_cfg_for_step(cfg_scale: float, *, has_negative_conditioning: bool) -> bool:
        return has_negative_conditioning and cfg_scale > 1.0

    @staticmethod
    def _validate_effective_schedule(*, start_idx: int, end_idx: int) -> None:
        if end_idx <= start_idx:
            raise ValueError(
                "The requested denoising range does not contain any effective denoising steps at the configured "
                "step count. Increase denoising_end, decrease denoising_start, or increase steps."
            )

View on GitHub (pinned to 0b6a024f2f)

Solutions

Make the cfg_scale list length exactly equal to the `steps` value passed to the denoise invocation.
Replace the list with a single float if the same CFG should apply to every step.
Compute the list programmatically from num_timesteps (e.g. interpolate a schedule) instead of hard-coding it.

Example fix

// before
steps=20; cfg_scale=[3.0, 3.5, 4.0]
// after (option A)
steps=20; cfg_scale=3.5
// after (option B)
cfg_scale=[3.0 + 0.05*i for i in range(20)]  # len == steps

Defensive patterns

Strategy: validation

Validate before calling

if isinstance(cfg_scale, list) and len(cfg_scale) != steps:
    raise ValueError(f"cfg_scale list length {len(cfg_scale)} must equal steps {steps}")

Type guard

def is_valid_cfg_scale(cfg_scale, steps: int) -> bool:
    if isinstance(cfg_scale, float):
        return True
    return isinstance(cfg_scale, list) and len(cfg_scale) == steps and all(isinstance(v, float) for v in cfg_scale)

Try / catch

try:
    out = invoke_krea2_denoise(cfg_scale=cfg_scale, steps=steps)
except ValueError as e:
    if "cfg_scale list has" in str(e):
        cfg_scale = float(cfg_scale[0]) if isinstance(cfg_scale, list) else cfg_scale
        out = invoke_krea2_denoise(cfg_scale=cfg_scale, steps=steps)
    else:
        raise

Prevention

Derive per-step CFG lists from the steps value programmatically, never hard-code.
Update CFG schedules whenever the steps setting changes.
Prefer a single float unless per-step scheduling is intentional.

When it happens

Trigger: Setting the invocation's cfg_scale input to a list like [3.0, 4.0] while running with steps=20, or changing `steps` after authoring a per-step CFG list sized for the old step count.

Common situations: Users copying a per-step CFG schedule from an example with a different steps value; workflow authors tweaking step count without updating the CFG list; programmatic graph generation computing the list before the final step count is known.

Related errors

AI-assisted analysis of invoke-ai/InvokeAI@0b6a024f2f (2026-08-29). Data as JSON: /api/errors/dc8b6320a98814fe. Report an issue: GitHub.