sgl-project/sglang · error · ValueError

Multi-output conditioning requires at least one prompt.

Error message

Multi-output conditioning requires at least one prompt.

What it means

Thrown by ConditionExpansion.from_batch when the computed prompt_batch_size is <= 0, i.e. batch.prompt is an empty list, so there is nothing to expand even though multi-output conditioning was requested.

Source

Thrown at python/sglang/multimodal_gen/runtime/utils/condition_expansion.py:30

    prompt_batch_size: int
    sample_batch_size: int

    @classmethod
    def from_batch(cls, batch):
        num_outputs = int(batch.num_outputs_per_prompt or 1)
        if num_outputs <= 1:
            return None
        if isinstance(batch.prompt, list):
            prompt_batch_size = len(batch.prompt)
        elif batch.prompt is not None:
            prompt_batch_size = 1
        else:
            raise ValueError(
                "Multi-output conditioning requires prompt text so the prompt "
                "batch size is unambiguous."
            )
        if prompt_batch_size <= 0:
            raise ValueError("Multi-output conditioning requires at least one prompt.")
        return cls(prompt_batch_size, prompt_batch_size * num_outputs)

    def _expand_tensor(self, value: torch.Tensor, name: str) -> torch.Tensor:
        current_batch_size = value.shape[0]
        if current_batch_size == self.sample_batch_size:
            return value
        if current_batch_size != self.prompt_batch_size:
            raise ValueError(
                f"{name} has batch dim {current_batch_size} (shape "
                f"{tuple(value.shape)}); expected {self.prompt_batch_size} "
                f"(per-prompt) or {self.sample_batch_size} (per-sample)."
            )
        repeats = self.sample_batch_size // self.prompt_batch_size
        return value.repeat_interleave(repeats, dim=0)

    def _expand_tensors(self, value, name: str):
        """Expand a tensor or each tensor in a list, preserving its container."""
        if value is None:

View on GitHub (pinned to 0132848349)

Solutions

1. Ensure the batch contains at least one prompt before calling expansion
2. Guard upstream: skip processing when the batch is empty
3. Fix the upstream filter/slicing that produced an empty prompt list

Example fix

# before
batch.prompt = []
# after
batch.prompt = ["prompt text"]

if isinstance(batch.prompt, list):
    assert len(batch.prompt) > 0, "empty prompt list"

def has_prompts(batch) -> bool:
    return batch.prompt is not None and (not isinstance(batch.prompt, list) or len(batch.prompt) > 0)

When it happens

Trigger: batch.prompt == [] with num_outputs > 1 passed into expand_conditioning_to_sample_batch.

Common situations: Empty batch constructed by mistake (filtered to zero items upstream); test fixture with an empty prompt list.

Related errors

• Multi-output conditioning requires prompt text so the prompt
• kv-canary: RealKvSource.read_bytes must be a positive multip
• top_p must be scalar or have one value per row, got {top_ps.
• top_p values must be in (0, 1]
• top_k must be scalar or have one value per row, got {top_ks.

AI-assisted analysis of sgl-project/sglang@0132848349 (2026-08-28). Data as JSON: /api/errors/8bd5a8a08f4df954. Report an issue: GitHub.