sgl-project/sglang · error · ValueError

Unknown pair_postprocess name

Error message

Unknown pair_postprocess name: {name}

What it means

set_pair_postprocess_by_name only recognizes a fixed set of built-in pair postprocess names (e.g. 'dual_sigma_shift'); any other string reaches the trailing raise. The name is matched before this line, so the value passed is simply not one of the implemented options.

Solutions

  1. Use the exact supported name, e.g. 'dual_sigma_shift'
  2. If you need custom behavior, pass a callable via set_pair_postprocess(fn) instead of a name
  3. Inspect the if/elif chain above line 442 to enumerate valid names

Example fix

# before
sched.set_pair_postprocess_by_name("dual_shift")
# after
sched.set_pair_postprocess_by_name("dual_sigma_shift")
Defensive patterns

Strategy: validation

Validate before calling

VALID = {"dual_sigma_shift"}
if name not in VALID:
    raise ValueError(f"unsupported pair_postprocess {name!r}; valid: {VALID}")

Type guard

def is_valid_pair_postprocess(name: str) -> bool:
    return name == "dual_sigma_shift"

Prevention

When it happens

Trigger: Calling set_pair_postprocess_by_name('sigma_shift') or any typo'd/unimplemented name; forward() also routes to this method when configured with a name string.

Common situations: Renamed or upstream-removed postprocess names after a version change; copy-pasting a name from another scheduler codebase.

Related errors


AI-assisted analysis of sgl-project/sglang@0132848349 (2026-08-28). Data as JSON: /api/errors/013b0ffca88a5475. Report an issue: GitHub.

Appendix: source

Thrown at python/sglang/multimodal_gen/runtime/models/schedulers/flow_match_pair.py:442

                        scale_factor = one_minus_z[-1] / (1 - self.shift_terminal)
                        if scale_factor != 0:
                            base = 1 - (one_minus_z / scale_factor)

                    if self.reverse_sigmas:
                        base = 1 - base

                    if source == "timesteps":
                        return base * self.num_train_timesteps
                    return base

                col0 = _build_column(visual_shift, visual_denoising_strength, visual_mu)
                col1 = _build_column(audio_shift, audio_denoising_strength, audio_mu)
                return torch.stack([col0, col1], dim=1)

            _dual_sigma_shift._requires_source = True
            self.set_pair_postprocess(_dual_sigma_shift)
            return
        raise ValueError(f"Unknown pair_postprocess name: {name}")

    def _make_pairs_from_vector(self, vec: torch.Tensor) -> torch.Tensor:
        if vec.ndim != 1:
            raise ValueError("vec must be 1D")
        return torch.stack([vec, vec], dim=1)

    def get_pairs(self, source: str = "timesteps") -> torch.Tensor:
        if source == "timesteps":
            if self.pair_timesteps is None:
                self._refresh_pair_cache()
            return self.pair_timesteps
        if source == "sigmas":
            if self.pair_sigmas is None:
                self._refresh_pair_cache()
            return self.pair_sigmas
        raise ValueError("source must be 'timesteps' or 'sigmas'")

    def timestep_to_sigma(self, timestep: torch.Tensor | float) -> torch.Tensor:

View on GitHub (pinned to 0132848349)