Skip to content

Commit 4180417

Browse files
authored
[https://nvbugs/5601682][fix] Fix cacheTransceiver hang (#9311)
Signed-off-by: Iman Tabrizian <10105175+tabrizian@users.noreply.github.com>
1 parent 838df92 commit 4180417

File tree

2 files changed

+8
-4
lines changed

2 files changed

+8
-4
lines changed

cpp/tensorrt_llm/nanobind/batch_manager/cacheTransceiver.cpp

Lines changed: 4 additions & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -81,8 +81,10 @@ void tb::CacheTransceiverBindings::initBindings(nb::module_& m)
8181
.def("respond_and_send_async", &BaseCacheTransceiver::respondAndSendAsync)
8282
.def("request_and_receive_sync", &BaseCacheTransceiver::requestAndReceiveSync)
8383
.def("request_and_receive_async", &BaseCacheTransceiver::requestAndReceiveAsync)
84-
.def("check_context_transfer_status", &BaseCacheTransceiver::checkContextTransferStatus)
85-
.def("check_gen_transfer_status", &BaseCacheTransceiver::checkGenTransferStatus)
84+
.def("check_context_transfer_status", &BaseCacheTransceiver::checkContextTransferStatus,
85+
nb::call_guard<nb::gil_scoped_release>())
86+
.def("check_gen_transfer_status", &BaseCacheTransceiver::checkGenTransferStatus,
87+
nb::call_guard<nb::gil_scoped_release>())
8688
.def("check_gen_transfer_complete", &BaseCacheTransceiver::checkGenTransferComplete);
8789

8890
nb::enum_<executor::kv_cache::CacheState::AttentionType>(m, "AttentionType")

cpp/tensorrt_llm/pybind/batch_manager/cacheTransceiver.cpp

Lines changed: 4 additions & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -77,8 +77,10 @@ void tb::CacheTransceiverBindings::initBindings(py::module_& m)
7777
.def("respond_and_send_async", &BaseCacheTransceiver::respondAndSendAsync)
7878
.def("request_and_receive_sync", &BaseCacheTransceiver::requestAndReceiveSync)
7979
.def("request_and_receive_async", &BaseCacheTransceiver::requestAndReceiveAsync)
80-
.def("check_context_transfer_status", &BaseCacheTransceiver::checkContextTransferStatus)
81-
.def("check_gen_transfer_status", &BaseCacheTransceiver::checkGenTransferStatus)
80+
.def("check_context_transfer_status", &BaseCacheTransceiver::checkContextTransferStatus,
81+
py::call_guard<py::gil_scoped_release>())
82+
.def("check_gen_transfer_status", &BaseCacheTransceiver::checkGenTransferStatus,
83+
py::call_guard<py::gil_scoped_release>())
8284
.def("check_gen_transfer_complete", &BaseCacheTransceiver::checkGenTransferComplete);
8385

8486
py::enum_<executor::kv_cache::CacheState::AttentionType>(m, "AttentionType")

0 commit comments

Comments
 (0)