Deep Learning never ceases to amaze us and that is why in this session we propose to see the latest advances and the state of the art of Deep learning. To do this, we will use as the mainstay of the session one of the latest published models: "Dall-e", which is based on a "transformers" model, being able to generate images from text descriptions. Here we leave you the link where you can know more about the model that we are commenting on: https://openai.com/blog/dall-e/

In this session we will focus on explaining how a training pipeline of the Dall-e model would be developed so that you learn about geometric shapes using various Azure services. The most important of all this is not only how Dalle trains, but at the same time learning to organize our project and how to train the most current Deep Learning models in Azure.

For this, we will use Databricks, it will be in charge of managing our training and validation of the model, the reason and curiosities of the use of said service will be discussed. The corresponding metrics for a model of this type will be registered in Azure Machine Learning Service. The metrics will be explained accordingly to better understand the reason for the selected model. Finally, the model will be saved in the Azure Machine Learning Service model registry.

Once we have trained our model and there is a stable version in the Azure Machine Learning registry, We will be able to consume our model using a small virtual assistant, making use of the bot framework, which will help and behave as a teacher towards the students so that they can learn about geometric shapes!