Runtime definitions to deploy ML/AI models on DaaS (Deployment-as-a-Service)
You can easily create your own runtime image running on DaaS, besides of your own libraries, the requirements-service.txt is required to install for the web service runtime.
Dockerfiles
- Common AI on CPU: Dockerfile, Instructions
- ONNX on GPU: Dockerfile, Instructions
- Pytorch on GPU: Dockerfile, Instructions
- Tensorflow on GPU: Dockerfile, Instructions
An out-of-box runtime includes all most popular open source machine learning and deep learning libraries running on CPU
Supported build arguements with default values:
base=18.04
CONDA_VERSION=py37_4.8.2
SPARK_VERSION=2.4.8
HADOOP_VERSION=2.7
Build the docker image from the Dockerfile in this repository, use --build-arg to specify arguements
docker build --build-arg base="20.04" -t ai-cpu -f Dockerfile .
ONNX Runtime on GPU
The default base image of ONNX Runtime:
base=v1.5.2-cuda10.2-cudnn8
Build the docker image from the Dockerfile in this repository.
docker build -t onnx-gpu -f Dockerfile.onnx .
Pytroch on GPU
The default base image of Pytorch:
base=1.7.0-cuda11.0-cudnn8-runtime
Build the docker image from the Dockerfile in this repository.
docker build -t pytorch-gpu -f Dockerfile.pytorch .
Tensorflow on GPU
The default base image of Tensorflow:
base=2.4.0-gpu
Build the docker image from the Dockerfile in this repository.
docker build -t tensorflow-gpu -f Dockerfile.tensorflow .
CONDA_MIRROR
APACHE_MIRROR
PIP_MIRROR
APT_MIRROR
Build the common AI image using mirrors, for example:
docker build --build-arg CONDA_MIRROR="mirrors.ustc.edu.cn/anaconda" --build-arg APACHE_MIRROR="mirrors.ustc.edu.cn/apache" --build-arg PIP_MIRROR="pypi.mirrors.ustc.edu.cn/simple/" --build-arg APT_MIRROR="mirrors.ustc.edu.cn/ubuntu" -t ai-cpu -f Dockerfile .