1. Data Delivery Based on Object Storage As shown above, model R&D mainly involves three sub-platforms: Data platform The data platform is mainly responsible for data-related management, such as data ingestion and data processing, ultimately producing the data needed for training. The data platform stores raw data in object storage. During processing, it fetches data from object storage, processes it, and stores the result back in object storage.

1. Data Parallelism Training steps: The master device loads the model and copies the model parameters to every worker device The master device splits the training data along the batch dimension and passes each batch to every worker device Each worker device performs training The master device aggregates the gradients from each worker device and updates the model parameters The master device broadcasts the model parameters to every worker device, preparing for the next batch of training Core idea:

As requirements are gradually confirmed, the system’s DB model is gradually confirmed as well. Sometimes you use a tool like MySQL Workbench to draw an E-R diagram and then generate MySQL database tables, and at that point you need to reverse-generate Django Models; other times you write the Django Models directly, but then you need to look at the E-R diagram.

Define metadata for a model through an inner class “class Meta”, like this: 1 2 3 4 class Foo(models.Model): bar = models.CharField(max_length=30) class Meta: #...... Model metadata means: any data that is not a field such as ordering options, admin options, and so on. Below are all the Meta options you might use.