Training Adaptive Computation for Open-Domain Question Answering with Computational Constraints

TL;DR

This paper introduces Adaptive Passage Encoder, a novel adaptive computation method for open-domain question answering that enhances efficiency and accuracy while requiring minimal computational resources for training.

Contribution

It presents a new AC approach that trains efficiently on a single GPU, fixing the base model parameters and optimizing computation dynamically.

Findings

Improves accuracy over state-of-the-art ODQA models

Requires only a single GPU for training

Outperforms previous AC methods in efficiency

Abstract

Adaptive Computation (AC) has been shown to be effective in improving the efficiency of Open-Domain Question Answering (ODQA) systems. However, current AC approaches require tuning of all model parameters, and training state-of-the-art ODQA models requires significant computational resources that may not be available for most researchers. We propose Adaptive Passage Encoder, an AC method that can be applied to an existing ODQA model and can be trained efficiently on a single GPU. It keeps the parameters of the base ODQA model fixed, but it overrides the default layer-by-layer computation of the encoder with an AC policy that is trained to optimise the computational efficiency of the model. Our experimental results show that our method improves upon a state-of-the-art model on two datasets, and is also more accurate than previous AC methods due to the stronger base ODQA model. All source code and datasets are available at https://github.com/uclnlp/APE.