This repository contains Dockerfiles and related configurations for building and managing containerized AI services and tools. The focus is on leveraging AMD ROCm for GPU acceleration and providing a streamlined environment for various AI applications.
- Base Images: Located in
dockerfiles/base/, Foundational images, Currently there's one single image,ubuntu_rocm_pytorch.dockerfile, which includes AMD ROCm, PyTorch, and other essential tools. Good enough. Check the dockerfile for versions. - Service Images: Found in
dockerfiles/services/, Dockerfiles for building specific AI services:ollama.dockerfile: For Ollama.comfyui.dockerfile: For ComfyUI.kokoro-tts-fastapi.dockerfile: For Kokoro TTS with FastAPI.forge.dockerfile: For WebUI Forge, AUTOMATIC1111's fork. [NOT BEING UPDATED]a1111.dockerfile: For AUTOMATIC1111's Stable Diffusion WebUI. [NOT BEING UPDATED]openedai-speech.dockerfile: For speech-related tasks. [NOT BEING UPDATED]zonos.dockerfile: For Zonos. [NOT BEING UPDATED]
- Located in
compose/services/. Docker Compose configurations for deploying services. Examples include:valkey.yaml: Configuration for the Valkey service.tika.yaml: Configuration for Apache Tika.open-webui-server.yaml: Configuration for Open WebUI Server.ollama.yaml: Configuration for Ollama.kokoro-tts.yaml: Configuration for Kokoro TTS.forge.yaml: Configuration for Forge.comfyui.yaml: Configuration for ComfyUI.a1111.yaml: Configuration for AUTOMATIC1111.
build.sh: A shell script for building and managing Docker images.build.py: An attempt to migratebuild.shto python. Currently put on hold.
- The
volumes/directory is used for storing persistent data.
Use the build.sh or build.py scripts to build Docker images. For example:
./build.sh allUse the Docker Compose files in compose/ to deploy services. For example:
docker-compose -f compose/services/ollama.yaml up