expose shutdown function at ffi layer

backends/trtllm/include/backend.h

@@ -92,6 +92,11 @@ namespace huggingface::tgi::backends {
          * @return Global number of generated tokens for this request id
          */
         uint32_t Stream(RequestId reqId, std::function<TokenStreamingCallback> &cb);
+
+        /***
+         * Stop the underlying executor
+         */
+        void Shutdown();
     };
 }
 

backends/trtllm/lib/backend.cpp

@@ -136,3 +136,9 @@ uint32_t huggingface::tgi::backends::TensorRtLlmBackend::Stream(const tle::IdTyp
 std::vector<tle::Response> huggingface::tgi::backends::TensorRtLlmBackend::Poll(const tle::IdType requestId) {
     return executor.awaitResponses(requestId);
 }
+
+
+void huggingface::tgi::backends::TensorRtLlmBackend::Shutdown() {
+    SPDLOG_INFO("Shutting down executor");
+    executor.shutdown();
+}

backends/trtllm/src/ffi.cpp

@@ -7,8 +7,6 @@
 #include <filesystem>
 #include <vector>
 
-//#include "rust/cxx.h"
-//#include "../include/ffi.h"
 #include "backends/trtllm/include/ffi.h"
 
 

backends/trtllm/src/lib.rs

@@ -1,4 +1,4 @@
-pub use backend::TrtLLmBackend;
+pub use backend::TensorRtLlmBackend;
 
 use crate::backend::GenerationContext;
 
@@ -58,5 +58,8 @@ mod ffi {
             request_id: u64,
             callback: fn(Box<GenerationContext>, u32, u32, bool),
         ) -> u32;
+
+        #[rust_name = "shutdown"]
+        fn Shutdown(self: Pin<&mut TensorRtLlmBackendImpl>);
     }
 }