Closed
Description
What happened?
After #10320, the docker image ci for llama-server-cuda failed.
This bug could be reproduced locally.
Name and Version
After Commit 467576b, docker ci failed.
What operating system are you seeing the problem on?
No response
Relevant log output
#13 [8/8] RUN if [ "default" != "default" ]; then export CMAKE_ARGS="-DCMAKE_CUDA_ARCHITECTURES=default"; fi && cmake -B build -DGGML_CUDA=ON -DLLAMA_CURL=ON ${CMAKE_ARGS} -DCMAKE_EXE_LINKER_FLAGS=-Wl,--allow-shlib-undefined . && cmake --build build --config Release -j$(nproc) && cp build/bin/* .
#13 0.130 -- The C compiler identification is GNU 11.4.0
#13 0.182 -- The CXX compiler identification is GNU 11.4.0
#13 0.189 -- Detecting C compiler ABI info
#13 0.253 -- Detecting C compiler ABI info - done
#13 0.261 -- Check for working C compiler: /usr/bin/cc - skipped
#13 0.262 -- Detecting C compile features
#13 0.262 -- Detecting C compile features - done
#13 0.265 -- Detecting CXX compiler ABI info
#13 0.338 -- Detecting CXX compiler ABI info - done
#13 0.346 -- Check for working CXX compiler: /usr/bin/c++ - skipped
#13 0.346 -- Detecting CXX compile features
#13 0.347 -- Detecting CXX compile features - done
#13 0.351 -- Found Git: /usr/bin/git (found version "2.34.1")
#13 0.496 -- Looking for pthread.h
#13 0.565 -- Looking for pthread.h - found
#13 0.565 -- Performing Test CMAKE_HAVE_LIBC_PTHREAD
#13 0.633 -- Performing Test CMAKE_HAVE_LIBC_PTHREAD - Success
#13 0.634 -- Found Threads: TRUE
#13 0.636 -- Warning: ccache not found - consider installing it for faster compilation or disable this warning with GGML_CCACHE=OFF
#13 0.653 -- CMAKE_SYSTEM_PROCESSOR: x86_64
#13 0.876 -- Found OpenMP_C: -fopenmp (found version "4.5")
#13 0.957 -- Found OpenMP_CXX: -fopenmp (found version "4.5")
#13 0.958 -- Found OpenMP: TRUE (found version "4.5")
#13 0.958 -- OpenMP found
#13 0.958 -- Using llamafile
#13 0.959 -- x86 detected
#13 0.959 -- Using runtime weight conversion of Q4_0 to Q4_0_x_x to enable optimized GEMM/GEMV kernels
#13 0.959 -- Including CPU backend
#13 0.960 -- Using AMX
#13 0.960 -- Including AMX backend
#13 0.976 -- Found CUDAToolkit: /usr/local/cuda/include (found version "12.6.20")
#13 0.997 -- CUDA Toolkit found
#13 0.997 -- Using CUDA architectures: native
#13 1.019 CMake Error at /usr/share/cmake-3.22/Modules/CMakeDetermineCompilerId.cmake:726 (message):
#13 1.019 Compiling the CUDA compiler identification source file
#13 1.019 "CMakeCUDACompilerId.cu" failed.
#13 1.019
#13 1.019 Compiler: /usr/local/cuda/bin/nvcc
#13 1.019
#13 1.019 Build flags:
#13 1.019
#13 1.019 Id flags: --keep;--keep-dir;tmp;-gencode=arch=compute_,code=sm_ -v
#13 1.019
#13 1.019
#13 1.019
#13 1.019 The output was:
#13 1.019
#13 1.019 1
#13 1.019
#13 1.019 nvcc fatal : Unsupported gpu architecture 'compute_'
#13 1.019
#13 1.019
#13 1.019
#13 1.019
#13 1.019
#13 1.019 Call Stack (most recent call first):
#13 1.019 /usr/share/cmake-3.22/Modules/CMakeDetermineCompilerId.cmake:6 (CMAKE_DETERMINE_COMPILER_ID_BUILD)
#13 1.019 /usr/share/cmake-3.22/Modules/CMakeDetermineCompilerId.cmake:48 (__determine_compiler_id_test)
#13 1.019 /usr/share/cmake-3.22/Modules/CMakeDetermineCUDACompiler.cmake:298 (CMAKE_DETERMINE_COMPILER_ID)
#13 1.019 ggml/src/ggml-cuda/CMakeLists.txt:25 (enable_language)
#13 1.019
#13 1.019
#13 1.020 -- Configuring incomplete, errors occurred!
#13 1.020 See also "/app/build/CMakeFiles/CMakeOutput.log".
#13 1.020 See also "/app/build/CMakeFiles/CMakeError.log".
#13 ERROR: process "/bin/sh -c if [ \"${CUDA_DOCKER_ARCH}\" != \"default\" ]; then export CMAKE_ARGS=\"-DCMAKE_CUDA_ARCHITECTURES=${CUDA_DOCKER_ARCH}\"; fi && cmake -B build -DGGML_CUDA=ON -DLLAMA_CURL=ON ${CMAKE_ARGS} -DCMAKE_EXE_LINKER_FLAGS=-Wl,--allow-shlib-undefined . && cmake --build build --config Release -j$(nproc) && cp build/bin/* ." did not complete successfully: exit code: 1