[Bugfix] GLM-5.3-Flash: launch the kpool paged MQA logits in the varlen mode its schedule was built with (#55270)
Signed-off-by: Yifan Qiao <yifanqiao@inferact.ai> Co-authored-by: Claude Fable 5.1 <noreply@anthropic.com>
Yifan Qiao committed
Sep 24, 2026 at 17:13 UTC
39724758e7b74e7ca2c4416b51dd02d1da6d5ad8
1 file changed
+1
vllm/models/glm5next/nvidia/sparse_indexer.py
+1
@@ -558,6 +558,7 @@ def sparse_attn_indexer_kpool(
558
decode_metadata.schedule_metadata,
559
max_model_len=max_pool_len,
560
clean_logits=False,
561
+ indices=decode_metadata.indices,
562
)
563
num_rows = logits.shape[0]
564
# kpool: logits are pool-granular -> select topk_tokens//kpool pools,