[mac_ai] testing: remove the kompute flavor (#790)

see ggml-org/llama.cpp#14501

Author: kpouget

projects/mac_ai/testing/config.yaml

@@ -290,7 +290,6 @@ prepare:
             LLAMA_CPP_CMAKE_FLAGS: -DLLAMA_CURL=OFF
             BUILD_FLAVOR: # set at runtime with the name of the flavor (one from below)
           flavors:
-            kompute: -DGGML_KOMPUTE=ON -DGGML_NATIVE=OFF -DGGML_METAL=OFF
             vulkan: -DGGML_VULKAN=ON -DGGML_NATIVE=OFF -DGGML_METAL=OFF
             remoting: -DGGML_REMOTINGFRONTEND=ON -DGGML_CPU_ARM_ARCH=native
 
@@ -396,9 +395,7 @@ prepare:
 test:
   platform: *all_platforms
 
-  platforms_to_skip:
-  # - macos/llama_cpp/upstream_bin # always built, but no need to test it every time
-  - podman/llama_cpp/no-gpu/kompute # slow, not interesting atm
+  platforms_to_skip: []
 
   inference_server:
     port: 11434
@@ -531,5 +528,5 @@ __platform_check:
     no_gpu: no-gpu
 
   flavors:
-    llama_cpp: [metal, vulkan, upstream_bin, kompute, remoting]
+    llama_cpp: [metal, vulkan, upstream_bin, remoting]
     ramalama: [remoting, null]

projects/mac_ai/testing/images/llama_cpp.containerfile

@@ -38,8 +38,7 @@ ARG LLAMA_CPP_CMAKE_BUILD_FLAGS="--parallel 4"
 
 RUN git clone "${LLAMA_CPP_REPO}" src \
  && git -C src fetch origin ${LLAMA_CPP_VERSION} \
- && git -C src reset --hard FETCH_HEAD \
- && git -C src submodule update --init ggml/src/ggml-kompute/kompute
+ && git -C src reset --hard FETCH_HEAD
 
 RUN mkdir -p build \
  && echo "Version: ${LLAMA_CPP_VERSION}" > build/build.flags.log \

projects/mac_ai/testing/prepare_llama_cpp.py

@@ -194,10 +194,6 @@ def prepare_from_source(base_work_dir, platform):
         artifact_dir_suffix="_llama_cpp",
     )
 
-    # for the Kompute build
-    cmd = f"sed -i.bu s/-Werror//g {dest}/ggml/src/ggml-kompute/kompute/CMakeLists.txt"
-    remote_access.run_with_ansible_ssh_conf(base_work_dir, cmd)
-
     src_dir = dest
     cmake_parallel = config.project.get_config("prepare.llama_cpp.source.cmake.parallel")