Composable Function-preserving Expansions for Transformer Architectures

TL;DR

This paper introduces six composable transformations that allow incremental expansion of transformer models while preserving their functionality, potentially enabling more efficient training of larger models without starting from scratch.

Contribution

The paper proposes novel, function-preserving transformations for transformers that facilitate incremental scaling during training, a significant advancement over traditional methods.

Findings

Six transformations preserve model function during expansion

Transformations enable progressive scaling of transformer models

Potential to reduce training costs for large models

Abstract

Training state-of-the-art neural networks requires a high cost in terms of compute and time. Model scale is recognized to be a critical factor to achieve and improve the state-of-the-art. Increasing the scale of a neural network normally requires restarting from scratch by randomly initializing all the parameters of the model, as this implies a change of architecture's parameters that does not allow for a straightforward transfer of knowledge from smaller size models. In this work, we propose six composable transformations to incrementally increase the size of transformer-based neural networks while preserving functionality, allowing to expand the capacity of the model as needed. We provide proof of exact function preservation under minimal initialization constraints for each transformation. The proposed methods may enable efficient training pipelines for larger and more powerful models by progressively expanding the architecture throughout training.