AMDGPU APPIMAGE asar Autoencoder BLAS clblast clinfo Consolas contextsize Cooldown cublas cuda CUDA Dolfino finetuned flashattention Flashattention friendlykobold geforce ggml gguf GGUF gpulayers hipblas kcpps kcppt koboldai KOBOLDAI koboldcpp KoboldCpp KOBOLDCPP lora lowvram Lowvram maximizable minimizable MMAP mmproj mmq multiuser noavx nocertify nocuda noheader nommq noshift nsis nvidia oldpc OLDPC opencl pkexec quantmatmul radeon remotetunnel rocm ROCM rocminfo safetensors sdclipg sdclipl sdlora sdmodel sdphotomaker sdui sdvae SDXL Segoe sonarjs SPACEBAR taskkill Tauri togglefullscreen treemap trycloudflare unquantized useclblast usecuda usemmap usevulkan vram vulkan vulkaninfo wayland websearch