/opt/dtk-23.10/cuda/bin/nvcc -fPIC   -DWMMA -fPIC -fPIC -O2 -g -DNDEBUG -O3 -Wl,--no-as-needed,--version-script /data/norm/lmdeploy/build/src/turbomind/triton_backend/libtriton_fastertransformer.ldscript  -fuse-ld=gold -shared -Wl,-soname,libtriton_turbomind.so -o ../../../lib/libtriton_turbomind.so CMakeFiles/triton-turbomind-backend.dir/libfastertransformer.cc.o   -L/opt/mpi/lib  -L/usr/local/mpi/lib  -L/opt/dtk-23.10/cuda/targets/x86_64-linux/lib/stubs  -Wl,-rpath,"\${ORIGIN}" ../../../lib/libtritonserver.so ../../../lib/libtritonbackendutils.a ../../../lib/libtransformer-shared.so -lcublas -lcudart -lcurand /opt/dtk-23.10/cuda/lib64/libnccl.so /opt/mpi/lib/libmpi.so -lcudart ../../../lib/libtritonasyncworkqueue.a ../../../lib/libkernel-library-new.a /opt/dtk-23.10/cuda/lib64/libcudart_static.a -pthread -ldl /usr/lib64/librt.so -lmpi -lcublas -lcudart -lcurand /opt/dtk-23.10/cuda/lib64/libnccl.so -ldl 
