All projects

Feed-Forward Assisted Transformers

Why fine-tune the whole model when you can train a tiny network on top?

January 2024
Northwestern University
ProjectX 2024 · 3rd place
Feed-forward assisted transformers for efficient fine-tuning

I'm grateful to my teammates Blake Hu, Julian Baldwin, Stephen Cheng, Marko Veljanovski, and Michelle Zhang for their dedication, creativity, and collaborative spirit, and to our advisor Prof. Zach Wood-Doughty, whose feedback shaped both the research direction and the way we presented it.

This one taught me how to balance a theoretically interesting idea against practical efficiency constraints — and put me through a bit of light CUDA hazing along the way.

Abstract

Fine-tuning is the standard way to adapt a pre-trained language model to a downstream task, but the time and energy needed to update every parameter become prohibitive as models grow. Recent parameter-efficient methods shrink the number of parameters that get updated without meaningfully shrinking training time or energy consumption.

We propose a time-efficient alternative based on feature extraction: treat an off-the-shelf language model as a fixed source of embeddings and train a small feed-forward network on top of it for each downstream task. Averaged across the GLUE NLI benchmark with distilRoBERTa, this trains 124× faster than full fine-tuning and 101× faster than parameter-efficient fine-tuning, while reaching 81.9% and 85.0% of their respective performance.

Resources

This was a competition project; the write-up isn't publicly hosted, but I'm happy to share it — just email me.