diff --git a/backends/trtllm/lib/backend.cpp b/backends/trtllm/lib/backend.cpp index a3ac05343b7..b41933859e8 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; }