sgl-project/sglang · error · ValueError

layer.layer_id={layer.layer_id} does not have an index V cac

Error message

layer.layer_id={layer.layer_id} does not have an index V cache (either dense, or in the K-only group). index_kv layers: {list(self.index_kv_layer_id_mapping.keys())}

What it means

set_index_kv_buffer on MiniMaxSparseKVPool only accepts layers registered in index_kv_layer_id_mapping. The given layer.layer_id is dense or K-only, so there is no index-V pool slot to write into and the call is rejected before touching buffers.

Source

Thrown at python/sglang/srt/mem_cache/memory_pool.py:5014

            loc,
            cache_k,
            cache_v,
            k_scale,
            v_scale,
        )

    def set_index_kv_buffer(
        self,
        layer: RadixAttention,
        loc: torch.Tensor,
        cache_idx_k: torch.Tensor,
        cache_idx_v: torch.Tensor,
        k_scale: Optional[float] = None,
        v_scale: Optional[float] = None,
    ) -> None:
        mapped_id = self.index_kv_layer_id_mapping.get(layer.layer_id)
        if mapped_id is None:
            raise ValueError(
                f"layer.layer_id={layer.layer_id} does not have an index V "
                f"cache (either dense, or in the K-only group). "
                f"index_kv layers: {list(self.index_kv_layer_id_mapping.keys())}"
            )
        self.index_kv_pool.set_kv_buffer(
            layer,
            loc,
            cache_idx_k,
            cache_idx_v,
            k_scale,
            v_scale,
            layer_id_override=mapped_id,
        )

    def set_index_k_buffer(
        self,
        layer: RadixAttention,
        loc: torch.Tensor,

View on GitHub (pinned to 0132848349)

Solutions

1. Branch per layer group: use set_index_kv_buffer only for index-KV layers and set_index_k_buffer for K-only layers
2. Verify layer.layer_id against index_kv_layer_id_mapping before the call
3. Rebuild the pool with correct sparse layer classification if the layer should be index-KV

Example fix

# before
sparse_pool.set_index_kv_buffer(layer, loc, cache_k, cache_v)  # K-only layer
# after
if layer.layer_id in sparse_pool.index_kv_layer_id_mapping:
    sparse_pool.set_index_kv_buffer(layer, loc, cache_k, cache_v)
else:
    sparse_pool.set_index_k_buffer(layer, loc, cache_k, k_scale=layer.k_scale)

if layer.layer_id not in sparse_pool.index_kv_layer_id_mapping:
    raise ValueError(f'{layer.layer_id} is not index-KV; use set_index_k_buffer or main pool')

def is_index_kv_layer(sparse_pool, layer) -> bool:
    return sparse_pool.index_kv_layer_id_mapping.get(layer.layer_id) is not None

When it happens

Trigger: Calling set_index_kv_buffer(layer, loc, cache_idx_k, cache_idx_v, ...) with a layer whose id is not in index_kv_layer_id_mapping (dense layer or K-only sparse layer).

Common situations: Model code that uses one code path for all sparse layers instead of branching between set_index_k_buffer (K-only) and set_index_kv_buffer; layer partition config mismatches after model definition edits.

Related errors

• layer.layer_id={layer.layer_id} is not in the K-only sparse
• layer_id={layer_id} does not have an index V cache (either d
• layer_id={layer_id} is not a sparse attention layer; sparse
• MHATokenToKOnlyPool does not allocate V
• MHATokenToKOnlyPool: use set_index_k_buffer on the parent Mi

AI-assisted analysis of sgl-project/sglang@0132848349 (2026-08-28). Data as JSON: /api/errors/92125304db08da63. Report an issue: GitHub.