Part 3 Advanced deep networks for complex problems

Deep learning has come a long way since the introduction of models like convolutional neural networks, LSTMs, and so on. Transformer-based models encompassing billions of parameters have out-performed the aforementioned models across the board. Another topic that has gained popularity, due to the demand for better models and agility in developing ML models, is tracking and productionizing ML models.

In part 3, we will first discuss a more complex variant of RNN based models known as sequence-to-sequence models. Then we will discuss Transformer-based models in more detail and see firsthand how they can be used for tasks like spam classification and question answering. You will also learn how you can leverage high-level libraries like Hugging Face’s Transformers to implement solutions quickly.

Then, you will learn how to use the TensorBoard to track the performance of models. You will learn about easily visualizing model performance over time and advanced capabilities like performance profiling. Finally, we introduce TFX, a library that standardizes the productionization of ML models. You will develop an end-to-end pipeline that manages the ML workflow end to end, from data to the deployment.