microsoft/VibeVoice · error · ValueError

Must pass exactly one of `num_inference_steps` or `timesteps

Error message

Must pass exactly one of `num_inference_steps` or `timesteps`.

What it means

`set_timesteps()` requires exactly one of `num_inference_steps` or `timesteps`; calling with neither (both None) leaves the scheduler with no timestep grid, so it raises ValueError up front. This is an API-contract check copied from diffusers — you must tell the scheduler how many steps (or exactly which timesteps) to run.

Source

Thrown at vibevoice/schedule/dpm_solver.py:341

        num_inference_steps: int = None,
        device: Union[str, torch.device] = None,
        timesteps: Optional[List[int]] = None,
    ):
        """
        Sets the discrete timesteps used for the diffusion chain (to be run before inference).

        Args:
            num_inference_steps (`int`):
                The number of diffusion steps used when generating samples with a pre-trained model.
            device (`str` or `torch.device`, *optional*):
                The device to which the timesteps should be moved to. If `None`, the timesteps are not moved.
            timesteps (`List[int]`, *optional*):
                Custom timesteps used to support arbitrary timesteps schedule. If `None`, timesteps will be generated
                based on the `timestep_spacing` attribute. If `timesteps` is passed, `num_inference_steps` and `sigmas`
                must be `None`, and `timestep_spacing` attribute will be ignored.
        """
        if num_inference_steps is None and timesteps is None:
            raise ValueError("Must pass exactly one of `num_inference_steps` or `timesteps`.")
        if num_inference_steps is not None and timesteps is not None:
            raise ValueError("Can only pass one of `num_inference_steps` or `custom_timesteps`.")
        if timesteps is not None and self.config.use_karras_sigmas:
            raise ValueError("Cannot use `timesteps` with `config.use_karras_sigmas = True`")
        if timesteps is not None and self.config.use_lu_lambdas:
            raise ValueError("Cannot use `timesteps` with `config.use_lu_lambdas = True`")

        if timesteps is not None:
            timesteps = np.array(timesteps).astype(np.int64)
        else:
            # Clipping the minimum of all lambda(t) for numerical stability.
            # This is critical for cosine (squaredcos_cap_v2) noise schedule.
            clipped_idx = torch.searchsorted(torch.flip(self.lambda_t, [0]), self.config.lambda_min_clipped)
            last_timestep = ((self.config.num_train_timesteps - clipped_idx).numpy()).item()

            # "linspace", "leading", "trailing" corresponds to annotation of Table 2. of https://arxiv.org/abs/2305.08891
            if self.config.timestep_spacing == "linspace":
                timesteps = (

View on GitHub (pinned to 94da20d98b)

Solutions

Pass a step count: `scheduler.set_timesteps(30)`.
If you intended custom timesteps, pass `timesteps=[... , device=...)` list instead.
Trace where your pipeline computes num_inference_steps and give it a concrete default when config omits it.

Example fix

# before
scheduler.set_timesteps()  # ValueError

# after
scheduler.set_timesteps(num_inference_steps=30)

Defensive patterns

Strategy: validation

Validate before calling

if num_inference_steps is None and timesteps is None:
    num_inference_steps = 30  # your default
scheduler.set_timesteps(
    num_inference_steps=num_inference_steps, timesteps=timesteps
)

Prevention

Always give num_inference_steps a concrete default in config-driven pipelines.
Fail fast when the config key is missing instead of letting None flow to set_timesteps.
set_timesteps is mandatory setup — pair scheduler creation and set_timesteps in one function.

When it happens

Trigger: `scheduler.set_timesteps()` with no arguments, or code that conditionally computes `num_inference_steps` and the variable ends up None (e.g., a config key miss), or passing `num_inference_steps=None` explicitly to 'use defaults'.

Common situations: Wrapping the scheduler in a generic pipeline where step count comes from config and the key is absent; refactoring that moves the step-count argument into kwargs and forgets to forward it.

Related errors

AI-assisted analysis of microsoft/VibeVoice@94da20d98b (2026-08-15). Data as JSON: /api/errors/d20ac73666ada434. Report an issue: GitHub.