From 381f0f4d5a70ecf9a9ed354f1ef995400865d6c8 Mon Sep 17 00:00:00 2001 From: wucong25 Date: Mon, 1 Dec 2025 11:47:08 +0800 Subject: [PATCH] add attn_groups judge Signed-off-by: wucong25 --- vllm_ascend/worker/model_runner_v1.py | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) diff --git a/vllm_ascend/worker/model_runner_v1.py b/vllm_ascend/worker/model_runner_v1.py index aec2ebfc0c4..3c359d6b5f1 100644 --- a/vllm_ascend/worker/model_runner_v1.py +++ b/vllm_ascend/worker/model_runner_v1.py @@ -2697,7 +2697,8 @@ def initialize_kv_cache(self, kv_cache_config: KVCacheConfig) -> None: self.kv_cache_config = kv_cache_config self.may_add_encoder_only_layers_to_kv_cache_config() # NOTE(cmq): initialize_attn_backend must before using self.attn_groups - self.initialize_attn_backend(kv_cache_config) + if not self.attn_groups: + self.initialize_attn_backend(kv_cache_config) self.use_hybrid_blocks = (len(self.attn_groups) > 1) # NOTE: Currently, we determine whether we need `num_accepted_tokens` through `MambaSpec`. self.need_accepted_tokens = any([