VanillaKD: Revisit the Power of Vanilla Knowledge Distillation from Small Scale to Large Scale

TL;DR

This paper revisits vanilla knowledge distillation, demonstrating its effectiveness on large-scale datasets like ImageNet-1K when combined with strong data augmentation and training strategies, achieving state-of-the-art results.

Contribution

It reveals the small data pitfall in previous KD methods and shows vanilla KD's potential in large-scale scenarios with simple techniques.

Findings

Vanilla KD performs strongly on large datasets like ImageNet-1K.

Stronger data augmentation reduces the gap between vanilla and advanced KD methods.

State-of-the-art accuracy achieved with vanilla KD on multiple models.

Abstract

The tremendous success of large models trained on extensive datasets demonstrates that scale is a key ingredient in achieving superior results. Therefore, the reflection on the rationality of designing knowledge distillation (KD) approaches for limited-capacity architectures solely based on small-scale datasets is now deemed imperative. In this paper, we identify the \emph{small data pitfall} that presents in previous KD methods, which results in the underestimation of the power of vanilla KD framework on large-scale datasets such as ImageNet-1K. Specifically, we show that employing stronger data augmentation techniques and using larger datasets can directly decrease the gap between vanilla KD and other meticulously designed KD variants. This highlights the necessity of designing and evaluating KD approaches in the context of practical scenarios, casting off the limitations of small-scale datasets. Our investigation of the vanilla KD and its variants in more complex schemes, including stronger training strategies and different model capacities, demonstrates that vanilla KD is elegantly simple but astonishingly effective in large-scale scenarios. Without bells and whistles, we obtain state-of-the-art ResNet-50, ViT-S, and ConvNeXtV2-T models for ImageNet, which achieve 83.1\%, 84.3\%, and 85.0\% top-1 accuracy, respectively. PyTorch code and checkpoints can be found at https://github.com/Hao840/vanillaKD.