End-to-End Sample Project: Deep Learning Recommender with TensorFlow
This tutorial shows by example how to use Gradient Notebooks and Workflows, and the interaction between them. The project is an end-to-end demonstration of building a recommender system then adding deep learning to it.
Many real data science projects begin with an exploratory phase where the data are investigated and machine learning models trained, then if value is found the team wishes to put the model into production. While it is not a requirement to be the case, often the exploratory phase will correspond to using Notebooks, and the production (or production-like) phase to using Workflows.
Many online demonstrations also either focus on the model, neglecting the data preparation and deployment into production, or they focus on the engineering and neglect the data science. Here, we show both.
Tutorial contents
The tutorial consists of
Running the Tutorial
Assuming you are signed up for Gradient and logged in, to run the tutorial, run the Notebook. To do this using Gradient:
    In the Gradient GUI, create a Notebook with the following settings:
      Name = Deep Learning Recommender (or any allowed name)
      Select a runtime = TensorFlow 2.4.1
      Select a machine = C3 (C5, P4000, etc., will also work)
      Public/private = set to preferred option
      Under Advanced options, change the Workspace URL field from https://github.com/gradient-ai/TF2.4.1.git to https://github.com/gradient-ai/Deep-Learning-Recommender-TF to point to this repository. The other options can remain the same.
      Start the Notebook
    Once the Notebook has started, clickdeep_learning_recommender_tf.ipynb to run in the usual way by clicking Run under each cell in turn.
Notebook creation can also be done on the command line if desired, via gradient notebooks create. For more details on Notebooks, see the documentation.
Workflow
To run the Workflow section of the Notebook, section 4.3, requires some extra steps. These are detailed in the Notebook.
Target Audience
We assume the readers of the tutorial are generally technical, but are not necessarily experts in data science, engineering, recommender systems, or Gradient.
Last modified 1mo ago
Copy link