From d4ff2681acd315cb88d18b3e0d7178385dccfb8a Mon Sep 17 00:00:00 2001
From: Hermes Agent service account
 <hermes@carousel-of-progress.local.mk-labs.cloud>
Date: Thu, 6 Aug 2026 09:24:42 -0500
Subject: [PATCH] llm-inference-multimodel: fix qwen unit -- llama.cpp requires
 --flash-attn <on|off|auto>, not bare flag

---
 .../templates/llama-server-qwen.service.j2                      | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/ansible/roles/llm-inference-multimodel/templates/llama-server-qwen.service.j2 b/ansible/roles/llm-inference-multimodel/templates/llama-server-qwen.service.j2
index 8f23346..76a11ae 100644
--- a/ansible/roles/llm-inference-multimodel/templates/llama-server-qwen.service.j2
+++ b/ansible/roles/llm-inference-multimodel/templates/llama-server-qwen.service.j2
@@ -14,7 +14,7 @@ ExecStart={{ llm_binary_path }} \
     --port {{ llm_qwen_port }} \
     --n-gpu-layers {{ llm_qwen_gpu_layers }} \
     --ctx-size {{ llm_qwen_ctx_size }} \
-    --flash-attn \
+    --flash-attn on \
     --cache-type-k q8_0 --cache-type-v q8_0 \
     --batch-size {{ llm_qwen_batch_size }} --ubatch-size {{ llm_qwen_ubatch_size }} \
     --jinja \