sgl-project/sglang · error · ValueError

Either num_inference_steps, sigmas, or timesteps must be…

Error message

Either num_inference_steps, sigmas, or timesteps must be provided

What it means

Raised by the flow-match Euler scheduler's set_timesteps when none of num_inference_steps, sigmas, or timesteps is provided, leaving no way to build a schedule. Pass num_inference_steps (and optionally sigmas/timesteps).

Solutions

1. Pass num_inference_steps explicitly: scheduler.set_timesteps(50)
2. Or pass a sigmas/timesteps list from which the count is inferred
3. Audit refactor leftovers where the steps variable became None

Example fix

# before
scheduler.set_timesteps()
# after
scheduler.set_timesteps(num_inference_steps=50)

assert num_inference_steps is not None or sigmas is not None or timesteps is not None, "provide steps or a schedule"

When it happens

Trigger: Calling scheduler.set_timesteps() with no arguments (relying on a previously stored default that does not exist in this implementation).

Common situations: Porting code from schedulers that default num_inference_steps=50; refactors that moved the step count into a variable that evaluates to None.

Understand the failure class

Background: "missing required argument" and "the following required arguments were not provided": what required-argument errors mean and how to fix them — this error's family across 20 libraries.

Related errors

• denoising_strength must be positive
• --disagg-server-addr is required for --disagg-role
• items must be provided
• kv-canary: launch_canary_plan_kernels_torch_reference…
• kv-canary: max_prefill_tokens must be positive, got

AI-assisted analysis of sgl-project/sglang@0132848349 (2026-08-28). Data as JSON: /api/errors/8aa396e9e3d2e1fb. Report an issue: GitHub.

Appendix: source

Thrown at python/sglang/multimodal_gen/runtime/models/schedulers/scheduling_flow_match_euler_discrete.py:326

            and timesteps is not None
            and len(sigmas) != len(timesteps)
        ):
            raise ValueError("`sigmas` and `timesteps` should have the same length")

        if num_inference_steps is not None:
            if (sigmas is not None and len(sigmas) != num_inference_steps) or (
                timesteps is not None and len(timesteps) != num_inference_steps
            ):
                raise ValueError(
                    "`sigmas` and `timesteps` should have the same length as num_inference_steps, if `num_inference_steps` is provided"
                )
        else:
            if sigmas is not None:
                num_inference_steps = len(sigmas)
            elif timesteps is not None:
                num_inference_steps = len(timesteps)
            else:
                raise ValueError(
                    "Either num_inference_steps, sigmas, or timesteps must be provided"
                )

        self.num_inference_steps = num_inference_steps

        # 1. Prepare default sigmas
        is_timesteps_provided = timesteps is not None

        timesteps_array: np.ndarray | None = None
        if is_timesteps_provided:
            assert timesteps is not None
            timesteps_array = np.array(timesteps).astype(np.float32)

        sigmas_array: np.ndarray
        if sigmas is None:
            if timesteps_array is None:
                timesteps_array = np.linspace(
                    self._sigma_to_t(self.sigma_max),

View on GitHub (pinned to 0132848349)