[https://nvbugs/5601682][fix] Fix cacheTransceiver hang (NVIDIA#9311)

Signed-off-by: Iman Tabrizian <10105175+tabrizian@users.noreply.github.com>
Signed-off-by: Mike Iovine <6158008+mikeiovine@users.noreply.github.com>

Author: Tabrizian

cpp/tensorrt_llm/nanobind/batch_manager/cacheTransceiver.cpp

@@ -88,8 +88,10 @@ void tb::CacheTransceiverBindings::initBindings(nb::module_& m)
         .def("respond_and_send_async", &BaseCacheTransceiver::respondAndSendAsync)
         .def("request_and_receive_sync", &BaseCacheTransceiver::requestAndReceiveSync)
         .def("request_and_receive_async", &BaseCacheTransceiver::requestAndReceiveAsync)
-        .def("check_context_transfer_status", &BaseCacheTransceiver::checkContextTransferStatus)
-        .def("check_gen_transfer_status", &BaseCacheTransceiver::checkGenTransferStatus)
+        .def("check_context_transfer_status", &BaseCacheTransceiver::checkContextTransferStatus,
+            nb::call_guard<nb::gil_scoped_release>())
+        .def("check_gen_transfer_status", &BaseCacheTransceiver::checkGenTransferStatus,
+            nb::call_guard<nb::gil_scoped_release>())
         .def("check_gen_transfer_complete", &BaseCacheTransceiver::checkGenTransferComplete)
         .def("cancel_request", &BaseCacheTransceiver::cancelRequest);
 

cpp/tensorrt_llm/pybind/batch_manager/cacheTransceiver.cpp

@@ -84,8 +84,10 @@ void tb::CacheTransceiverBindings::initBindings(py::module_& m)
         .def("respond_and_send_async", &BaseCacheTransceiver::respondAndSendAsync)
         .def("request_and_receive_sync", &BaseCacheTransceiver::requestAndReceiveSync)
         .def("request_and_receive_async", &BaseCacheTransceiver::requestAndReceiveAsync)
-        .def("check_context_transfer_status", &BaseCacheTransceiver::checkContextTransferStatus)
-        .def("check_gen_transfer_status", &BaseCacheTransceiver::checkGenTransferStatus)
+        .def("check_context_transfer_status", &BaseCacheTransceiver::checkContextTransferStatus,
+            py::call_guard<py::gil_scoped_release>())
+        .def("check_gen_transfer_status", &BaseCacheTransceiver::checkGenTransferStatus,
+            py::call_guard<py::gil_scoped_release>())
         .def("check_gen_transfer_complete", &BaseCacheTransceiver::checkGenTransferComplete)
         .def("cancel_request", &BaseCacheTransceiver::cancelRequest);