Lossless and Near-Lossless Compression for Foundation Models
This addresses infrastructure costs for deploying large AI models, though it is incremental as it builds on traditional compression methods applied to a new context.
The paper tackles the problem of large foundation models burdening infrastructure by introducing lossless and near-lossless compression techniques, achieving over 50% size reduction for some models and estimating potential savings of over an ExaByte per month in network traffic from model hubs like HuggingFace.
With the growth of model sizes and scale of their deployment, their sheer size burdens the infrastructure requiring more network and more storage to accommodate these. While there is a vast literature about reducing model sizes, we investigate a more traditional type of compression -- one that compresses the model to a smaller form and is coupled with a decompression algorithm that returns it to its original size -- namely lossless compression. Somewhat surprisingly, we show that such lossless compression can gain significant network and storage reduction on popular models, at times reducing over $50\%$ of the model size. We investigate the source of model compressibility, introduce compression variants tailored for models and categorize models to compressibility groups. We also introduce a tunable lossy compression technique that can further reduce size even on the less compressible models with little to no effect on the model accuracy. We estimate that these methods could save over an ExaByte per month of network traffic downloaded from a large model hub like HuggingFace.