Code analysis of the SDNet model
Abstract
Previous chapters have introduced the design principles of the network architecture for machine reading comprehension (MRC) models. In practice, there are many important implementation details, such as establishing a word dictionary, build training data in batches, and padding text of unequal lengths. This chapter details these implementation techniques through analysis of the source code of an MRC model, SDNet. Many of these techniques are transferrable to other models. In addition, since SDNet utilizes the pretrained model Bidirectional Encoder Representations from Transformers (BERT), the analysis will showcase how to efficiently use BERT in natural language processing tasks.
Keywords
SDNet; ...
Get Machine Reading Comprehension now with the O’Reilly learning platform.
O’Reilly members experience books, live events, courses curated by job role, and more from O’Reilly and nearly 200 top publishers.
