sgl-project/sglang · error · ValueError

Invalid attention metadata values.Sparsity should be in [0,

Error message

Invalid attention metadata values.Sparsity should be in [0, 1), skip_first_steps should be non-negative.Got sparsity={sparsity}, skip_first_steps={skip_first_steps}

What it means

RainFusionAttentionMetadata.build validates that sparsity is in [0, 1) and skip_first_steps is non-negative before constructing metadata; bad values would produce an invalid sparse attention pattern, so it fails immediately.

Source

Thrown at python/sglang/multimodal_gen/runtime/layers/attention/backends/rain_fusion_attn.py:74

class RainFusionAttentionMetadataBuilder(AttentionMetadataBuilder):
    def __init__(self) -> None:
        pass

    def prepare(self) -> None:
        pass

    def build(
        self,
        current_timestep: int,
        skip_first_steps: int,
        sparsity: float,
        raw_latent_shape: list[int],
        patch_size: tuple[int, int, int],
        **kwargs: dict[str, Any],
    ) -> RainFusionAttentionMetadata:
        if not (skip_first_steps >= 0 and 0.0 <= sparsity < 1.0):
            raise ValueError(
                (
                    "Invalid attention metadata values."
                    f"Sparsity should be in [0, 1), skip_first_steps should be non-negative."
                    f"Got sparsity={sparsity}, skip_first_steps={skip_first_steps}"
                )
            )

        if sparsity == 0.0:
            logger.warning(
                (
                    "Sparsity is set to 0.0, which means no tokens will be dropped."
                    "For better performance use Laser Attention or increase sparsity."
                )
            )

        latent_shape = raw_latent_shape[-3:]
        latent_shape = [latent_shape[i] // patch_size[i] for i in range(3)]

View on GitHub (pinned to 0132848349)

Solutions

Correct the values: 0.0 <= sparsity < 1.0 and skip_first_steps >= 0
Convert density to sparsity with sparsity = 1 - density if the config uses density
Validate both values at config-parse time before build() is called

Example fix

# before
meta = RainFusionAttentionMetadata.build(sparsity=1.2, skip_first_steps=0, ...)
# after
meta = RainFusionAttentionMetadata.build(sparsity=0.2, skip_first_steps=0, ...)

Defensive patterns

Strategy: validation

Validate before calling

assert 0.0 <= sparsity < 1.0 and skip_first_steps >= 0, "invalid RainFusion sparsity args"

Prevention

Validate sparse-attn config ranges where YAML is parsed
Convert density to sparsity explicitly

When it happens

Trigger: Calling RainFusionAttentionMetadata.build with sparsity >= 1.0, negative sparsity, or negative skip_first_steps.

Common situations: RainFusion (video/diffusion sparse attention) configs specifying density where sparsity is expected; negative skip_first_steps meant as 'no skip'; hand-tuned YAML with typos.

Related errors

AI-assisted analysis of sgl-project/sglang@0132848349 (2026-08-28). Data as JSON: /api/errors/d4350d45d8b56bf4. Report an issue: GitHub.