Capacity Matters: a Proof-of-Concept for Transformer Memorization on Real-World Data
This work provides insights into transformer memory mechanisms for optimizing model design with structured real-world data, though it appears incremental in scope.
The paper investigates how model architecture and data configurations affect the memorization capacity of generative transformers on synthetic text datasets derived from SNOMED, finding that embedding size primarily determines learning speed and capacity, while additional layers offer limited benefits and Softmax activation provides greater stability.
This paper studies how the model architecture and data configurations influence the empirical memorization capacity of generative transformers. The models are trained using synthetic text datasets derived from the Systematized Nomenclature of Medicine (SNOMED) knowledge graph: triplets, representing static connections, and sequences, simulating complex relation patterns. The results show that embedding size is the primary determinant of learning speed and capacity, while additional layers provide limited benefits and may hinder performance on simpler datasets. Activation functions play a crucial role, and Softmax demonstrates greater stability and capacity. Furthermore, increasing the complexity of the data set seems to improve the final memorization. These insights improve our understanding of transformer memory mechanisms and provide a framework for optimizing model design with structured real-world data.