ms-2k
ms-2k
/ cuda-rocm-vulkan.Dockerfile
Last active
September 19, 2025 09:33
Dockerfile for llama.cpp built with CUDA, ROCm, and Vulkan backends for RTX 3090 + AMD Instinct MI50 (or any RTX 30-series + gfx906). Requires running ROCm Offline Installer Creator first (I couldn't figure out package manager installation method without it conflicting with CUDA base images yet; still working on it)
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| ARG UBUNTU_VERSION=24.04 | |
| ARG CUDA_VERSION=12.8.1 | |
| # Target the CUDA build image | |
| ARG BASE_CUDA_DEV_CONTAINER=nvidia/cuda:${CUDA_VERSION}-devel-ubuntu${UBUNTU_VERSION} | |
| ARG BASE_CUDA_RUN_CONTAINER=nvidia/cuda:${CUDA_VERSION}-runtime-ubuntu${UBUNTU_VERSION} | |
| # Build-time ROCm base (CUDA devel + ROCm) | |
| FROM ${BASE_CUDA_DEV_CONTAINER} AS build-rocm-base |