Overview

This is a tool to quickly save your machine learning models and deploy them on k8s or docker-compose for serving purpose.

Support

• Serve Tensorflow .pb models
• Serve Pytorch models
• Serve MXNet models
• Serve SKLearn models
• Serve other machine learning models saved in .onnx form

Functions:

1:MLModelMgmt:

Build an image, convert and save your trained models to this serviceable image

apis：

• MLModelMgmt.build_image
• MLModelMgmt.push_image
• MxnetToOnnx.build_image
• PytorhcToOnnx.build_image
• SklToOnxx.build_image

You can check the api documents for more information about these apis

2:FeedMeImages:

Use the model saving images and a serving image to serve your model on k8s or docker-compose

apis:

• MxnetServing.k8sYaml
• MxnetServing.composeYaml
• OnnxServing.k8sYaml
• OnnxServing.composeYaml
• TfServing.k8sYaml
• TfServing.composeYaml

You can check the api documents for more information about these apis

Pre-request:

• Python3.6+
• Docker-ce
• onnx
• tensorflow
• torch
• torchvision
• mxnet
• skl2onnx
• minikube or kubernetes(if you want to serve it on k8s)
• docker-compose(if you want to serve it with docker-compose)

Example:

Install the package:

git clone <url> 

cd feedmeimages-master

python setup.py install