MLOps or ML Ops is a set of practices that aims to deploy and maintain machine learning models in production reliably and efficiently. The word is a compound of “machine learning” and the continuous…

In your journey to learn machine learning and data science, there's a high chance that you've came across "MLOps" term mentioned in some of those online tutorials and articles, but you probably thought it's just another fancy word invented to add an extra layer of complexity to the DS world. MLOps stands for Machine Learning Operations, it refers to the set of practices that improves the quality of machine learning an AI projects, and makes their lifecycle management much easier and reproducible for all stakeholders. Productionizing ML projects is difficult. The machine learning lifecycle consists of many complex components such as data ingestion, preprocessing, model training, hyperparameters tuning, model deployment, and performance monitoring... etc. It also requires collaboration and hand-offs across teams, from Data Engineering to Data Science to ML Engineering.

This blog was written in collaboration with the DeepSpeed team, the Azure ML team, and the Azure HPC team at Microsoft. Large-scale transformer-based deep learning models trained on large amounts of data have shown great results in recent years in several cognitive tasks and are behind new products and features that augment human capabilities. These models have grown several orders of magnitude in size during the last five years. Starting from a few million parameters of the original transformer model all the way to the latest 530 billion-parameter Megatron-Turing (MT-NLG 530B) model as shown in Figure 1. There is a growing need for customers to train and fine-tune large models at an unprecedented scale.

Please read the other post Red Wine Quality prediction using AzureML, AKS. This was done using machine learning techniques and not using deep learning. The same thing is accomplished here but using the deep learning framework Keras. Most of the things remain the same compared to the machine learning method, but a few steps change. I am going to highlight the changed aspects here only so that it is easy to follow.

One challenge with keeping up with bleeding edge AI and ML frameworks is that the changes can easily outpace APIs and SDKs build on top of these tools. Fortunately, since these tools are largely open-source and made available through industry common package managers, these updates can be used before the frameworks support them, albeit with some workarounds. Azure Machine Learning Studio (AzureML) is a platform service that provides full end to end management of Machine Learning and Data Science workloads. The tool provides experiment tracking, data set management, model repository, and deployment services to enable data scientists to train/validate/deploy their models in a cloud scale environment. To run TensorFlow 2.2 in AzureML is really pretty trivial, given the existing support for the TensorFlow framework.

Before I begin, let me tell you that this post is part of the Microsoft Student Partners Developer Stories initiative, and is based on the AI and ML Track. We will be exploring various Azure services - Azure Notebooks, Machine Learning Service, Container Instances and Container Registry. This post is beginner-friendly and can be used by anyone to deploy their machine learning models to Azure in a Standard format. Even high school kids are creating Machine Learning models these days, using popular machine learning frameworks like Keras, PyTorch, Caffe, etc. The model format created in one framework slightly differs with the model format created in the other.

Provide an overview of Azure Machine Learning Studio and high level architecture.

It was my favorite out of the ones our team put on and we are talking about doing a BI event in the next few months. If you interested in presenting on any BI related topic or your company would like to sponsor a full day BI training event in LA, please let me know.

This course is designed in such a way by keeping in mind the entry level of Data Scientists or having no background in programming or data science. In this course of Azure Machine Learning, you will be more excited and alsohave fun to learn, create and deploy machine learning models. This course teaches basic and also the advanced techniques of Data processing, Parameter Tuning andFeature Selection which an experienced and seasoned Data Scientist typically expands. In a very short duration, you will be able to match the results that an experienced data scientist can achieve. This course will help you to prepare for the entry to this hot career path of Machine Learning. This course goes through in details of Machine Learning and its implementation.

A key balancing act in machine learning is choosing an appropriate level of model complexity: if the model is too complex, it will fit the data used to construct the model very well but generalise poorly to unseen data (overfitting); if the complexity is too low the model won't capture all the information in the data (underfitting). In deep learning or machine learning scenarios, model performance depends heavily on the hyperparameter values selected. The goal of hyperparameter exploration is to search across various hyperparameter configurations to find a configuration that results in the best performance. Typically, the hyperparameter exploration process is painstakingly manual, given that the search space is vast and evaluation of each configuration can be expensive. The model parameters tell how to transform input data into desired output whereas, the hyperparameters are used to determine the structure of the model in use.