From 58773af70883a9c7cfb75e5b4b336715a18e15f2 Mon Sep 17 00:00:00 2001 From: lianyibo Date: Sat, 20 Dec 2025 20:47:30 +0800 Subject: [PATCH] [Fix] Delete pooling redundant code (#4940) ### What this PR does / why we need it? Remove redundant code in #3122. - vLLM version: v0.12.0 - vLLM main: https://github.com/vllm-project/vllm/commit/ad32e3e19ccf0526cb6744a5fed09a138a5fb2f9 Signed-off-by: lianyibo --- vllm_ascend/worker/model_runner_v1.py | 9 +-------- 1 file changed, 1 insertion(+), 8 deletions(-) diff --git a/vllm_ascend/worker/model_runner_v1.py b/vllm_ascend/worker/model_runner_v1.py index 19e8a310..1b32e359 100644 --- a/vllm_ascend/worker/model_runner_v1.py +++ b/vllm_ascend/worker/model_runner_v1.py @@ -1173,14 +1173,7 @@ class NPUModelRunner(GPUModelRunner): def _build_attn_state(self, num_reqs, num_scheduled_tokens, num_valid_tokens): - if self.model_config.runner_type == "pooling": - if isinstance( - self.kv_cache_config.kv_cache_groups[0].kv_cache_spec, - EncoderOnlyAttentionSpec): - attn_state = AscendAttentionState.PrefillNoCache - else: - attn_state = AscendAttentionState.PrefillCacheHit - elif np.array_equal(self.seq_lens.np[:num_reqs], num_scheduled_tokens): + if np.array_equal(self.seq_lens.np[:num_reqs], num_scheduled_tokens): attn_state = AscendAttentionState.PrefillNoCache # We assume it is the decode stage, where prefill occurs but only one token is not hit in cache. elif np.all(num_scheduled_tokens == 1):