English

Efficient Self-supervised Learning with Contextualized Target Representations for Vision, Speech and Language

Machine Learning 2023-06-16 v2 Computation and Language Sound Audio and Speech Processing

Abstract

Current self-supervised learning algorithms are often modality-specific and require large amounts of computational resources. To address these issues, we increase the training efficiency of data2vec, a learning objective that generalizes across several modalities. We do not encode masked tokens, use a fast convolutional decoder and amortize the effort to build teacher representations. data2vec 2.0 benefits from the rich contextualized target representations introduced in data2vec which enable a fast self-supervised learner. Experiments on ImageNet-1K image classification show that data2vec 2.0 matches the accuracy of Masked Autoencoders in 16.4x lower pre-training time, on Librispeech speech recognition it performs as well as wav2vec 2.0 in 10.6x less time, and on GLUE natural language understanding it matches a retrained RoBERTa model in half the time. Trading some speed for accuracy results in ImageNet-1K top-1 accuracy of 86.8\% with a ViT-L model trained for 150 epochs.

Keywords

Cite

@article{arxiv.2212.07525,
  title  = {Efficient Self-supervised Learning with Contextualized Target Representations for Vision, Speech and Language},
  author = {Alexei Baevski and Arun Babu and Wei-Ning Hsu and Michael Auli},
  journal= {arXiv preprint arXiv:2212.07525},
  year   = {2023}
}
R2 v1 2026-06-28T07:35:31.876Z