diff --git a/vllm/models/glm5next/nvidia/sparse_indexer.py b/vllm/models/glm5next/nvidia/sparse_indexer.py index c80e442015..00f51cc6a7 100644 --- a/vllm/models/glm5next/nvidia/sparse_indexer.py +++ b/vllm/models/glm5next/nvidia/sparse_indexer.py @@ -558,6 +558,7 @@ def sparse_attn_indexer_kpool( decode_metadata.schedule_metadata, max_model_len=max_pool_len, clean_logits=False, + indices=decode_metadata.indices, ) num_rows = logits.shape[0] # kpool: logits are pool-granular -> select topk_tokens//kpool pools,