From c870eb9e0f001ed5d07e6b6e2eb5e500a080a717 Mon Sep 17 00:00:00 2001 From: Xin Yang <105740670+xyang16@users.noreply.github.com> Date: Mon, 23 Feb 2026 23:17:53 -0800 Subject: [PATCH] [LoRA] Update LoRA expand kernel block_n calculation (#32621) Signed-off-by: Xin Yang --- vllm/lora/ops/triton_ops/utils.py | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/vllm/lora/ops/triton_ops/utils.py b/vllm/lora/ops/triton_ops/utils.py index 39c175f30..c7ac5914b 100644 --- a/vllm/lora/ops/triton_ops/utils.py +++ b/vllm/lora/ops/triton_ops/utils.py @@ -251,7 +251,7 @@ def get_lora_op_configs( else: default = { "block_m": 64, - "block_n": max(64, next_power_of_2(128 // num_slices)), + "block_n": 64 if num_slices > 1 else 128, "block_k": 16, "num_warps": 4, "num_ctas": 1,