When i run curl -fsSL https://raw.githubusercontent.com/ericcurtin/vllm-vulkan/main/install.sh | bash it fails due to it trying to compile for cuda and not for cpu, as described in the readme. It also starts way to many jobs that takes 1-2 gb ram and almost crashes the system.
I assume that incorrect env is set, and that it doesent ask the system for how many jobs to actually run in parrarel.
I assume that these environment variables are missing
export VLLM_PRECOMPILED_WHEEL_VARIANT=cpu VLLM_TARGET_DEVICE=cpu
I also get a larage warning that obscures the real errors
CMake Deprecation Warning at .deps/onednn-src/CMakeLists.txt:36 (cmake_policy):
The OLD behavior for policy CMP0146 will be removed from a future version
of CMake.
The cmake-policies(7) manual explains that the OLD behaviors of all
policies are deprecated and that a policy should be set to OLD only under
specific short-term circumstances. Projects should be ported to the NEW
behavior and not rely on setting a policy to OLD.
It then crashes bacause it needs numa.h. Im continuing to try to fix errors as i go and report them. Hope this is helping the project.
When i run
curl -fsSL https://raw.githubusercontent.com/ericcurtin/vllm-vulkan/main/install.sh | bashit fails due to it trying to compile for cuda and not for cpu, as described in the readme. It also starts way to many jobs that takes 1-2 gb ram and almost crashes the system.I assume that incorrect env is set, and that it doesent ask the system for how many jobs to actually run in parrarel.
I assume that these environment variables are missing
I also get a larage warning that obscures the real errors
It then crashes bacause it needs
numa.h. Im continuing to try to fix errors as i go and report them. Hope this is helping the project.