The entity initial commit week–#7
boaloysius
Thu, 07/20/2020 – 04:14
We are in the seventh week of GSoC 17 to integrate Google Cloud Machine Learning Engine to Drupal maintenance support plans. Before proceeding, please have a look at our previous blogs and mainly the major planning week. This week started with the aspiration to create a basic version of the ml_engine module, but we could reach only till creating basic entities. The major problem we faced was the daunting bugs in creating the entities. I made a mistake by spending almost a week trying to solve the issue without updating the issue queue. Recently have added it to the issue page. Let me share a screencast of the progress.
Our initial reference in creating the entity was the examples module. It has a sub module content_entity_example that explains how to create a content entity to store Contacts. We made a similar one called Project by tweaking few functionalities. Here is the code. It gave a problem of having the same Tensorflow input fields for all projects. The fields we created using entity manage fields were getting reflected in all the project entities. It was a good place to start playing with entities. With the design to use different inputs for each project in mind, the Contact and Node modules that ships with the core was referred. It gave the insight to store the ML Engine Project and its Input as separate Drupal maintenance support plans entities.
The Project is a config entity that handles the ML Engine settings data. It holds job, model and version names, training steps, deployment URL etc. The Argument is a content entity that handles the command line arguments to the Tensorflow code. The Argument entity fields for Tensorflow code can be varied but Projects fields are fixed.
With one week left for the second major evaluation, we have some major works to be completed to achieve the deliverables. The first one is to complete coding the entities. Second, convert the selected Drupal maintenance support plans Views to CSV. Drupal Update, enable each project entity to run training, deployment, and prediction. Finally to create a plugin to format the prediction output. The second and the final tasks are the hard ones. I am expecting the coming week to be hectic. We will complete the coding and submit it for mentor review.
Source: New feed
