From cdba16fd23e42d2b77b3e47f9422eeaa52a5fd27 Mon Sep 17 00:00:00 2001
From: Morgan Funtowicz <funtowiczmo@gmail.com>
Date: Mon, 21 Oct 2024 23:40:54 +0200
Subject: [PATCH] chore(trtllm): ensure max throughput scheduling policy is
 selected

---
 backends/trtllm/lib/backend.cpp | 1 +
 1 file changed, 1 insertion(+)

diff --git a/backends/trtllm/lib/backend.cpp b/backends/trtllm/lib/backend.cpp
index 9e8e50c6..a80e9c5d 100644
--- a/backends/trtllm/lib/backend.cpp
+++ b/backends/trtllm/lib/backend.cpp
@@ -53,6 +53,7 @@ tle::ExecutorConfig huggingface::tgi::backends::GetExecutorConfig(const json &co
     // Define some configuration variables
     execConfig.setKvCacheConfig(tle::KvCacheConfig(true));
     execConfig.setEnableChunkedContext(computeCapabilities.isPostAmpere());
+    execConfig.setSchedulerConfig(tle::SchedulerConfig(tle::CapacitySchedulerPolicy::kMAX_UTILIZATION));
     return execConfig;
 }