DLOps - MNIST Convolutional Neural Network

In this project, a basic convolutional neural network model for the binary classification of handwritten digits is developed using the MNIST dataset and Pytorch Lightning. In addition, all the flow associated with MLOps or DLOps is performed. According to this, an own dataset is generated from MNIST, performing version control through GIT, DVC and Firebase. Subsequently, the data modules and the model are built using Pytorch Lightning, monitoring the training using Weights & Biases. Afterwards, the best trained model is optimized by ONNX to be able to be used in different environments. Finally, an API is built using FastAPI to expose the model and be able to be consumed by sending an image, in this way, the Docker is also defined to be able to be deployed in different environments. This project was inspired by content generated by Sensio.