[Bugfix] Avoid global config lookup in sparse indexer forward (#54400)

Signed-off-by: zjy0516 <riverclouds.zhu@qq.com>
Co-authored-by: OpenAI Codex <codex@openai.com>
This commit is contained in:
Jiangyun Zhu
2026-08-30 03:31:47 -07:00
committed by GitHub
co-authored by OpenAI Codex
parent 2c7d7dd64a
commit 8c51b92654
@@ -769,6 +769,7 @@ class SparseAttnIndexer(CustomOp):
# during model construction) and pass them into the custom op, rather
# than threading them through per-step metadata.
parallel_config = get_current_vllm_config().parallel_config
self._parallel_config = parallel_config
self.dcp_world_size = parallel_config.decode_context_parallel_size
self.dcp_rank = get_dcp_group().rank_in_group if self.dcp_world_size > 1 else 0
self.use_pcp = parallel_config.prefill_context_parallel_size > 1
@@ -787,9 +788,7 @@ class SparseAttnIndexer(CustomOp):
(it's set up in Worker.initialize_from_config, ahead of warmup/serving).
"""
if self._cp_kv_cache_interleave_size is None:
value = (
get_current_vllm_config().parallel_config.cp_kv_cache_interleave_size
)
value = self._parallel_config.cp_kv_cache_interleave_size
if isinstance(get_forward_context().attn_metadata, dict):
self._cp_kv_cache_interleave_size = value
return value