Google

The training of BERT models follows a standard iterative optimization procedure used for deep neural networks. First, a large collection of training data is gathered. During each iteration, a random batch of these samples is selected, and the cumulative loss, $$\mathrm{Loss}_{\mathrm{BERT}}$$, is computed over the batch. Next, the model's parameters are updated to minimize this loss using an optimization algorithm like gradient descent or one of its variants. This cycle continues until a specific stopping condition is met, such as the convergence of the training loss.

BERT Training Process

A data scientist is describing a single iterative step in the training process for a large language model that uses two distinct pre-training objectives. Which of the following descriptions accurately portrays the correct sequence of operations within that single step?

A large language model is being trained on a massive text corpus using an iterative optimization procedure. Arrange the following key operations into the correct sequence for a single training iteration.

A machine learning engineer is training a large language model on a text corpus. After several iterations, they observe that the model's total loss value is fluctuating but not showing a consistent downward trend. Assuming the loss calculation itself is correct, which core component of the iterative optimization procedure is most likely misconfigured or malfunctioning, and why does this lead to the observed behavior?

Learn Before

Related