TiC-CLIP: Continual Training of CLIP Models

arXiv:2310.16226v2 Announce Type: replace-cross
Abstract: Keeping large foundation models up to date on latest data is inherently expensive. To avoid the prohibitive costs of constantly retraining, it is imperative to \emph{continually} train these models. This problem is exacerbated by the lack of any large scale continual learning benchmarks or baselines. We introduce the first set of web-scale Time-Continual (TiC) benchmarks for training vision-language models: TiC-DataComp, TiC-YFCC, and TiC-Redcaps. TiC-DataComp, our largest dataset, contains over 12.7B timestamped image-text pairs spanning 9 …

arxiv clip continual cs.cl cs.cv cs.lg training type