diff --git a/backends/trtllm/lib/backend.cpp b/backends/trtllm/lib/backend.cpp index e7a5b969..fc5d4594 100644 --- a/backends/trtllm/lib/backend.cpp +++ b/backends/trtllm/lib/backend.cpp @@ -16,6 +16,10 @@ tle::ExecutorConfig huggingface::tgi::backends::GetExecutorConfig(const json &co tle::OrchestratorConfig(true, workerPath) )); + + // TODO : Need to check for >= sm_80 (ampere) + // execConfig.setEnableChunkedContext(true) + execConfig.setKvCacheConfig(tle::KvCacheConfig(true)); return execConfig; }