[Distributed] Add opt-in FlashInfer PCIe IPC all-reduce backend (#53576)

Signed-off-by: Zihua Wu <13583761+lucifer1004@users.noreply.github.com>
Co-authored-by: Yongye Zhu <zyy1102000@gmail.com>
Co-authored-by: mergify[bot] <37929162+mergify[bot]@users.noreply.github.com>
This commit is contained in:
Gabriel Wu
2026-09-01 03:34:56 +00:00
committed by GitHub
co-authored by Yongye Zhu mergify[bot] <37929162+mergify[bot]@users.noreply.github.com>
parent e16b5e518d
commit 446c769482
7 changed files with 450 additions and 4 deletions
+1
View File
@@ -396,6 +396,7 @@ def test_flashinfer_all_reduce_precedes_nccl(monkeypatch: pytest.MonkeyPatch) ->
fi_ar_comm.all_reduce.return_value = output
communicator = CudaCommunicator.__new__(CudaCommunicator)
communicator.fi_ar_comm = fi_ar_comm
communicator.fi_pcie_ipc_ar_comm = None
communicator.pynccl_comm = Mock(world_size=8)
communicator.qr_comm = None
nccl_selector = Mock(return_value=True)