Replies: 1 comment
-
Im on unbutu debian. Make some remarkable by my know what =) |
Beta Was this translation helpful? Give feedback.
0 replies
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
-
linux, nvidia, CUDA toolkit installed (and works)
llamafile -t 12 -ngl 35 (etc) yields:
compile_nvidia: note: building ggml-cuda with nvcc -arch=native...
llamafile_log_command: /usr/bin/nvcc -arch=native --shared --forward-unknown-to-host-compiler -use_fast_math --compiler-options "-fPIC -O3 -march=native -mtune=native" -DNDEBUG -DGGML_BUILD=1 -DGGML_SHARED=1 -DGGML_CUDA_MMV_Y=1 -DGGML_MULTIPLATFORM -DGGML_CUDA_DMMV_X=32 -DK_QUANTS_PER_ITERATION=2 -DGGML_CUDA_PEER_MAX_BATCH_SIZE=128 -DGGML_USE_CUBLAS -o /home/sam/.llamafile/ggml-cuda.so.8c95tc /home/sam/.llamafile/ggml-cuda.cu -lcublas -lcuda
nvcc fatal : Value 'native' is not defined for option 'gpu-architecture'
Compile: warning: /usr/bin/nvcc returned nonzero exit status
get_nvcc_arch_flag: note: building nvidia compute capability detector
^^^^^^^^ that last line sounds promising, meanwhile I would like to know how to change "native" to "compute_86"
Do I need to compile from source?
ETA: recompiled, same errors. Re-built NVIDIA/CUDA drivers (from their website) and everything works.
Beta Was this translation helpful? Give feedback.
All reactions