Docker for MlOps
What is Docker ?
Docker is an open source platform used to build, deploy and manage applications containing content known as a platform with features that enable developers to create containers with integrated applications that can continue to work anywhere. However containers can be built without Docker but Docker makes it easier, safer and easier to build, move and manage containers.
Docker was introduced by Solmon Hykes in 2013 as a set of PAAS products using OS level virtualization, a time when containers were popular
What is MLOps ?
MLOps stand for Machine Learning Operations, are a set of best practises for anyone to run ML by providing softwares and cloud services
MLOps is the communication between data scientists and the performance or production team. It is deeply intertwined in nature, designed to eliminate waste, make automation possible, and produce a rich, consistent understanding of machine learning. ML could be a game changer, but without some form of programming, it could go into scientific testing. MLOps follow the same pattern as DevOps. Practices that drive seamless integration between your development cycle and your overall performance can also change how your organization handles big data. As DevOps shortens the production life cycles by building better products with each iteration, MLOps carries information you can trust and play with very fast.
Docker with MLOps
Docker has some advantages over the increase compared to the Anaconda. If you make a pipe in a container after that, with a little modification of your Dockerfile, your pipe is ready for production. For example, you may need to delete the features of the Dockerfile used for upgrades (Jupyter library, SSH configuration), enter your pipeline script as a command, and then install your favorite preferences (write to database, work over REST). After that you can sort the containers, use your pipeline with tools like Kubernetes in the same road collection. Alternatively, if you train in a container (as we do below) you can run the containers evenly in the hyperparameter tuning collection.