Empirical Capacity Model for Self-Attention Neural Networks
This work addresses the problem of inefficient model design for researchers and practitioners by providing a tool to optimize transformer parameters based on memorization needs, though it is incremental as it builds on existing capacity analysis.
The paper tackled the gap between theoretical and practical memorization capacity in large transformer models by developing an empirical capacity model (ECM) based on synthetic data, enabling optimal parameter design for task-specific memorization.
Large pretrained self-attention neural networks, or transformers, have been very successful in various tasks recently. The performance of a model on a given task depends on its ability to memorize and generalize the training data. Large transformer models, which may have billions of parameters, in theory have a huge capacity to memorize content. However, the current algorithms for the optimization fall short of the theoretical capacity, and the capacity is also highly dependent on the content. In this paper, we focus on the memory capacity of these models obtained using common training algorithms and synthetic training data. Based on the results, we derive an empirical capacity model (ECM) for a generic transformer. The ECM can be used to design task-specific transformer models with an optimal number of parameters in cases where the target memorization capability of the task can be defined.