I am trying to compile the ds4 glm5.2 branch on Dgx spark using make cuda-spark command after cloning the current repository on my machine.
This is the issue:
make -B ds4 ds4-server ds4-bench ds4-eval ds4-agent CUDA_ARCH=
make[1]: Verzeichnis „/home/nilab/apps/ds4-glm52“ wird betreten
cc -O3 -ffast-math -g -march=native -Wall -Wextra -std=c99 -D_GNU_SOURCE -fno-finite-math-only -c -o ds4_cli.o ds4_cli.c
cc -O3 -ffast-math -g -march=native -Wall -Wextra -std=c99 -D_GNU_SOURCE -fno-finite-math-only -c -o ds4_help.o ds4_help.c
cc -O3 -ffast-math -g -march=native -Wall -Wextra -std=c99 -D_GNU_SOURCE -fno-finite-math-only -c -o linenoise.o linenoise.c
cc -O3 -ffast-math -g -march=native -Wall -Wextra -std=c99 -D_GNU_SOURCE -fno-finite-math-only -c -o ds4.o ds4.c
cc -O3 -ffast-math -g -march=native -Wall -Wextra -std=c99 -D_GNU_SOURCE -fno-finite-math-only -c -o ds4_distributed.o ds4_distributed.c
cc -O3 -ffast-math -g -march=native -Wall -Wextra -std=c99 -D_GNU_SOURCE -fno-finite-math-only -c -o ds4_ssd.o ds4_ssd.c
/usr/local/cuda/bin/nvcc -O3 -g -lineinfo --use_fast_math -Xcompiler -march=native -Xcompiler -pthread -c -o ds4_cuda.o ds4_cuda.cu
ds4_cuda.cu(12963): error: more than one instance of overloaded function "ds4_gpu_routed_moe_one_tensor" has "C" linkage
extern "C" int ds4_gpu_routed_moe_one_tensor(ds4_gpu_tensor *out, ds4_gpu_tensor *gate, ds4_gpu_tensor *up, ds4_gpu_tensor *mid, ds4_gpu_tensor *down, const void *model_map, uint64_t model_size, uint64_t gate_offset, uint64_t up_offset, uint64_t down_offset, uint32_t gate_type, uint32_t down_type, uint64_t gate_expert_bytes, uint64_t gate_row_bytes, uint64_t down_expert_bytes, uint64_t down_row_bytes, uint32_t expert_in_dim, uint32_t expert_mid_dim, uint32_t out_dim, const ds4_gpu_tensor *selected, const ds4_gpu_tensor *weights, uint32_t n_total_expert, uint32_t n_expert, float clamp, const ds4_gpu_tensor *x, uint32_t layer_index) {
^
1 error detected in the compilation of "ds4_cuda.cu".
make[1]: *** [Makefile:219: ds4_cuda.o] Fehler 2
make[1]: Verzeichnis „/home/nilab/apps/ds4-glm52“ wird verlassen
make: *** [Makefile:98: cuda-spark] Fehler 2
I also verify with command git pull origin glm5.2 if there are fix on the repository:
Von https://github.com/antirez/ds4
- branch glm5.2 -> FETCH_HEAD
Bereits aktuell.
How to fix this ?
I am trying to compile the ds4 glm5.2 branch on Dgx spark using make cuda-spark command after cloning the current repository on my machine.
This is the issue:
make -B ds4 ds4-server ds4-bench ds4-eval ds4-agent CUDA_ARCH=
make[1]: Verzeichnis „/home/nilab/apps/ds4-glm52“ wird betreten
cc -O3 -ffast-math -g -march=native -Wall -Wextra -std=c99 -D_GNU_SOURCE -fno-finite-math-only -c -o ds4_cli.o ds4_cli.c
cc -O3 -ffast-math -g -march=native -Wall -Wextra -std=c99 -D_GNU_SOURCE -fno-finite-math-only -c -o ds4_help.o ds4_help.c
cc -O3 -ffast-math -g -march=native -Wall -Wextra -std=c99 -D_GNU_SOURCE -fno-finite-math-only -c -o linenoise.o linenoise.c
cc -O3 -ffast-math -g -march=native -Wall -Wextra -std=c99 -D_GNU_SOURCE -fno-finite-math-only -c -o ds4.o ds4.c
cc -O3 -ffast-math -g -march=native -Wall -Wextra -std=c99 -D_GNU_SOURCE -fno-finite-math-only -c -o ds4_distributed.o ds4_distributed.c
cc -O3 -ffast-math -g -march=native -Wall -Wextra -std=c99 -D_GNU_SOURCE -fno-finite-math-only -c -o ds4_ssd.o ds4_ssd.c
/usr/local/cuda/bin/nvcc -O3 -g -lineinfo --use_fast_math -Xcompiler -march=native -Xcompiler -pthread -c -o ds4_cuda.o ds4_cuda.cu
ds4_cuda.cu(12963): error: more than one instance of overloaded function "ds4_gpu_routed_moe_one_tensor" has "C" linkage
extern "C" int ds4_gpu_routed_moe_one_tensor(ds4_gpu_tensor *out, ds4_gpu_tensor *gate, ds4_gpu_tensor *up, ds4_gpu_tensor *mid, ds4_gpu_tensor *down, const void *model_map, uint64_t model_size, uint64_t gate_offset, uint64_t up_offset, uint64_t down_offset, uint32_t gate_type, uint32_t down_type, uint64_t gate_expert_bytes, uint64_t gate_row_bytes, uint64_t down_expert_bytes, uint64_t down_row_bytes, uint32_t expert_in_dim, uint32_t expert_mid_dim, uint32_t out_dim, const ds4_gpu_tensor *selected, const ds4_gpu_tensor *weights, uint32_t n_total_expert, uint32_t n_expert, float clamp, const ds4_gpu_tensor *x, uint32_t layer_index) {
^
1 error detected in the compilation of "ds4_cuda.cu".
make[1]: *** [Makefile:219: ds4_cuda.o] Fehler 2
make[1]: Verzeichnis „/home/nilab/apps/ds4-glm52“ wird verlassen
make: *** [Makefile:98: cuda-spark] Fehler 2
I also verify with command git pull origin glm5.2 if there are fix on the repository:
Von https://github.com/antirez/ds4
Bereits aktuell.
How to fix this ?