* Fix syntax for CUDA architecture definitions
* Extend CUDA support to GTX 10 Series to RTX 50 Series
* update cuda installer step version to install cuda 12.8.1
* Remove unsupported compute capability
* build: fix missing commit sha in macOS and Ubuntu build zip name
The build workflows for macOS and Ubuntu incorrectly check for the
"main" branch instead of "master" when retrieving the commit hash for
naming the build artifacts.
* build: correct Vulkan SDK installation condition in build workflow
* build: Enable Windows ROCm(HIP) build release
Refer to the build workflow of llama.cpp to add a Windows ROCm (HIP)
build release to the workflow.
Since there are many differences between the HIP build and other
builds, this commit add a separate "windows-latest-cmake-hip" job,
instead of enabling the ROCm matrix entry in the existing Windows
build job.
Main differences include:
- Install ROCm SDK from AMD official installer.
- Add a cache step for ROCm installation and a ccache step for build
processing, since the HIP build takes much longer time than other
builds.
- Include the ROCm/HIP artifact in the release assets.
* Fix includes and init vulkan the same as llama.cpp
* Add Windows Vulkan CI
* Updated ggml submodule
* support epsilon as a parameter for ggml_group_norm
---------
Co-authored-by: Cloudwalk <cloudwalk@icculus.org>
Co-authored-by: Oleg Skutte <00.00.oleg.00.00@gmail.com>
Co-authored-by: leejet <leejet714@gmail.com>
* move main and stb-libs to subfolder
* cmake : general additions
* ci : add simple building
---------
Co-authored-by: leejet <31925346+leejet@users.noreply.github.com>
