From 94fbb09894a00533a41ce2d976d9aa2f06e7e000 Mon Sep 17 00:00:00 2001
From: namgyu-youn <namgyu.dev@gmail.com>
Date: Tue, 7 Apr 2026 03:05:39 +0900
Subject: [PATCH] [EASY] Drop duplicate KV-cache initialization (#38799)

Signed-off-by: namgyu-youn <namgyu.dev@gmail.com>
---
 vllm/model_executor/layers/attention/attention.py | 3 ---
 1 file changed, 3 deletions(-)

diff --git a/vllm/model_executor/layers/attention/attention.py b/vllm/model_executor/layers/attention/attention.py
index 3ff4ec62a..a49415a3d 100644
--- a/vllm/model_executor/layers/attention/attention.py
+++ b/vllm/model_executor/layers/attention/attention.py
@@ -131,9 +131,6 @@ def _init_kv_cache_quant(
         quant_config: Optional quantization configuration.
         prefix: Layer name prefix for quantization method lookup.
     """
-    quant_method = (
-        quant_config.get_quant_method(layer, prefix=prefix) if quant_config else None
-    )
 
     # Note [Register q/k/v/prob scales in state dict]
     # When calling model.to(device), only parameters/buffers in state dict are