Skip to content

[https://nvbugs/6480574][fix] mpi_session: guard MpiPoolSession.shutdown against partial init #16613

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Closed
lowsfer wants to merge 1 commit into from
Closed

[https://nvbugs/6480574][fix] mpi_session: guard MpiPoolSession.shutdown against partial init #16613

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
11 changes: 7 additions & 4 deletions tensorrt_llm/llmapi/mpi_session.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -61,6 +61,7 @@ def is_initialized() -> bool:

def external_mpi_comm_available(model_world_size: int) -> bool:
"""Check if the current process is launched by mpirun and does not use MPIPoolExecutor to spawn processes.

e.g. mpirun -np 4 python script.py
"""
if ENABLE_MULTI_DEVICE:
Expand Down Expand Up @@ -241,7 +242,9 @@ def __init__(self,
n_workers: int,
wait_shutdown: bool = False,
env_overrides: Optional[Dict[str, str]] = None):
"""Args:
"""Create a pool session that spawns and manages MPI worker processes.

Args:
n_workers: number of MPI workers to spawn.
wait_shutdown: when True, ``shutdown()`` blocks until the spawned
worker processes have actually exited. ``MPIPoolExecutor.shutdown``
Expand Down Expand Up @@ -292,12 +295,12 @@ def shutdown(self, wait=True):
wait = False
if self.mpi_pool is not None:
logger.info(
f"MpiPoolSession.shutdown: joining {self.n_workers} worker(s) "
f"(wait={wait})")
"MpiPoolSession.shutdown: joining "
f"{getattr(self, 'n_workers', '?')} worker(s) (wait={wait})")
self.mpi_pool.shutdown(wait=wait)
logger.info("MpiPoolSession.shutdown: done")
self.mpi_pool = None
if self._wait_shutdown:
if getattr(self, '_wait_shutdown', False):
self._wait_workers_exit()

def _collect_worker_identities(self) -> Tuple:
Expand Down
1 change: 0 additions & 1 deletion tests/integration/test_lists/waives.txt
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -414,7 +414,6 @@ unittest/_torch/visual_gen/test_attention_integration.py::test_sage_attention_se
unittest/_torch/visual_gen/test_attention_integration.py::test_sage_attention_self_attention[int8-2-1560] SKIP (https://nvbugs/6198760)
unittest/auto_deploy/multigpu/custom_ops SKIP (https://nvbugs/6403920)
unittest/disaggregated/test_kv_transfer.py::test_transfer_worker_v2[tp4_pp1_to_tp2_pp2] SKIP (https://nvbugs/6426834)
unittest/executor/test_proxy_fast_death.py::test_pool_session_shutdown_never_blocks_after_release SKIP (https://nvbugs/6480574)
unittest/executor/test_rpc.py::TestRpcCorrectness::test_incremental_task_async SKIP (https://nvbugs/5741476)
unittest/executor/test_rpc_proxy.py SKIP (https://nvbugs/5605741)
unittest/executor/test_rpc_worker.py SKIP (https://nvbugs/5605741)
Expand Down
Loading