gpo.zugaina.org

Search Portage & Overlays:

dev-cpp/llama-cpp

Port of Facebook's LLaMA model in C/C++

Screenshots

  • llama-cpp-9999
    ~amd64
    blas cublas lto tests tools rocm +amdgpu_targets_gfx906 +amdgpu_targets_gfx908 +amdgpu_targets_gfx90a +amdgpu_targets_gfx1030 amdgpu_targets_gfx803 amdgpu_targets_gfx900 amdgpu_targets_gfx1010 amdgpu_targets_gfx1011 amdgpu_targets_gfx1012 amdgpu_targets_gfx1031 amdgpu_targets_gfx1100 amdgpu_targets_gfx1101 amdgpu_targets_gfx1102

    View      Download      Browse     License: MIT   
    Overlay: displacer
  • llama-cpp-0_pre4676
    ~amd64 ~x86 ~x86-macos ~x64-macos ~arm ~arm64 ~arm64-macos ~riscv ~loong
    static lto test examples +server curl hbm android msvc +accelerate blas blis +llamafile cann musa cuda cuda_force_mmq cuda_force_cublas +cuda_unified_memory cuda_f16 cuda_no_peer_copy cuda_no_vmm cuda_fa_all_quants +cuda_graphs hip hip_graphs +hip_no_vmm hip_uma vulkan vulkan_check_results vulkan_debug vulkan_memory_debug vulkan_shader_debug_info vulkan_perf vulkan_validate vulkan_run_tests kompute +openmp rpc opencl opencl_profiling +opencl_embed_kernels +opencl_use_adreno_kernels metal metal_use_bf16 metal_ndebug metal_shader_debug +metal_embed_library +cpu cpu_native cpu_flags_x86_avx cpu_flags_x86_avx_vnni cpu_flags_x86_avx2 cpu_flags_x86_avx512 cpu_flags_x86_avx512_vbmi cpu_flags_x86_avx512_vnni cpu_flags_x86_avx512_bf16 cpu_flags_x86_fma cpu_flags_x86_f16c cpu_flags_x86_amx_tile cpu_flags_x86_amx_int8 cpu_flags_x86_amx_bf16 cpu_flags_x86_sse cpu_flags_x86_sse2 cpu_flags_x86_sse3 cpu_flags_x86_sse4 cpu_flags_x86_sse4a cpu_flags_x86_sse41 cpu_flags_x86_sse42 cpu_flags_x86_ssse3 cpu_flags_loong_lasx cpu_flags_loong_lsx cpu_flags_riscv_rvv abi_x86_32 abi_x86_64 abi_x86_x32 abi_mips_n32 abi_mips_n64 abi_mips_o32 abi_s390_32 abi_s390_64 +amdgpu_targets_gfx906 +amdgpu_targets_gfx908 +amdgpu_targets_gfx90a +amdgpu_targets_gfx942 +amdgpu_targets_gfx1030 +amdgpu_targets_gfx1100 amdgpu_targets_gfx803 amdgpu_targets_gfx900 amdgpu_targets_gfx940 amdgpu_targets_gfx941 amdgpu_targets_gfx1010 amdgpu_targets_gfx1011 amdgpu_targets_gfx1012 amdgpu_targets_gfx1031 amdgpu_targets_gfx1101 amdgpu_targets_gfx1102

    View      Download      Browse     License: MIT   
    Overlay: ex_repo

ChangeLog

commit 93aa9392fb115eb38a8aba252406457cdf081536
Author: Arniiiii <lg3dx6fd@gmail.com>
Date: Sun Feb 9 10:46:49 2025 +0000

add some more docs

commit 6e01bd63e5f4f810647105b4071309ec9555cd7b
Author: Arniiiii <lg3dx6fd@gmail.com>
Date: Sun Feb 9 10:40:23 2025 +0000

attempt to fix tests for cuda and hip

commit cc6d3ca33e5258624f7a6f43e4830a9235b492e5
Author: Arniiiii <lg3dx6fd@gmail.com>
Date: Sun Feb 9 09:30:06 2025 +0000

update + fix `examples` use flag + new `cpu-native` use flag

commit 0b1ac3cfa55f10d7b2004cdd7c0e6b1e07650eaa
Author: Arniiiii <lg3dx6fd@gmail.com>
Date: Sun Feb 9 08:47:32 2025 +0000

make use of `cuda_src_prepare` only if USE `cuda`

commit a8975595cda53d58453ef9cd96796119ac0fbe1f
Author: Arniiiii <lg3dx6fd@gmail.com>
Date: Sun Feb 9 08:45:53 2025 +0000

make llama-cpp ebuild use `cuda` eclass

Now it requires manual overriding `NVCC_CCBIN=/usr/bin/gcc-13` or to
something else.

commit 6ef0fd7cedcad7cef374ed9caf2adc8a53d80321
Author: Arniiiii <lg3dx6fd@gmail.com>
Date: Sat Feb 8 12:25:48 2025 +0000

add new half-working ebuild for llama.cpp