Exploiting a Zoo of Checkpoints for Unseen Tasks
This addresses the challenge for practitioners in efficiently leveraging the abundance of available models for unseen tasks, though it is incremental as it builds on existing checkpoint and Gaussian process methods.
The paper tackles the problem of selecting effective pre-trained model checkpoints for new tasks by modeling task relationships with a Gaussian process and using a submodular objective to identify representative checkpoints, resulting in superior generalization performance across computational linguistics and computer vision applications.
There are so many models in the literature that it is difficult for practitioners to decide which combinations are likely to be effective for a new task. This paper attempts to address this question by capturing relationships among checkpoints published on the web. We model the space of tasks as a Gaussian process. The covariance can be estimated from checkpoints and unlabeled probing data. With the Gaussian process, we can identify representative checkpoints by a maximum mutual information criterion. This objective is submodular. A greedy method identifies representatives that are likely to "cover" the task space. These representatives generalize to new tasks with superior performance. Empirical evidence is provided for applications from both computational linguistics as well as computer vision.