mirror of
https://github.com/microsoft/mscclpp.git
synced 2026-05-11 17:00:22 +00:00
15 lines
1.2 KiB
Bash
Executable File
15 lines
1.2 KiB
Bash
Executable File
|
|
module load mpi/hpcx #mpi/hpcx-mrc #mpi/hpcx-mrc-2.23.1
|
|
|
|
MPI_ARGS=""
|
|
MPI_ARGS+="-x CUDA_VISIBLE_DEVICES=0,2 --mca coll ^ucc,hcoll -mca coll_hcoll_enable 0 --mca btl tcp,vader,self --mca pml ob1 --mca oob_tcp_if_include enP22p1s0f1 --mca btl_tcp_if_include enP22p1s0f1 "
|
|
MPI_ARGS+="-x MSCCLPP_IBV_SO=/opt/microsoft/mrc/Azure-Compute-AI-HPC-Perf-verbs-mrc/libibverbs.so -x UCX_NET_DEVICES=enP22p1s0f1 -x LD_LIBRARY_PATH=/opt/microsoft/mrc/Azure-Compute-AI-HPC-Perf-verbs-mrc/mrc-header-lib:$LD_LIBRARY_PATH"
|
|
MPI_ARGS+=" -x MSCCLPP_IBV_MODE=host -x MSCCLPP_SOCKET_IFNAME=enP22p1s0f1 -x MSCCLPP_IBV_MODE=host-no-atomic -x VMRC_LIBMRC_SO=/opt/mellanox/doca/lib/aarch64-linux-gnu/libnv_mrc.so"
|
|
MPI_ARGS+=" -x VMRC_LIBIBVERBS_SO=/lib/aarch64-linux-gnu/libibverbs.so.1 -x MSCCLPP_HCA_DEVICES=mlx5_1,mlx5_3 -x PATH=/home/azhpcuser/binyli/mscclpp_venv/bin/:$PATH "
|
|
MPI_ARGS+=" -x MSCCLPP_LOG_LEVEL=ERROR -x MSCCLPP_DEBUG=ERROR -x MSCCLPP_IB_GID_INDEX=3"
|
|
MPI_ARGS+=" /home/azhpcuser/binyli/mscclpp_venv/bin/python3 /home/azhpcuser/binyli/mscclpp/python/test/executor_test.py -path /home/azhpcuser/binyli/mscclpp/test.json"
|
|
|
|
|
|
|
|
mpirun -np 2 $MPI_ARGS --size 4K --n_iters 500 --n_graph_iters 100
|