diff --git a/clip_benchmark/datasets/builder.py b/clip_benchmark/datasets/builder.py index f8dd95d..14c1e56 100644 --- a/clip_benchmark/datasets/builder.py +++ b/clip_benchmark/datasets/builder.py @@ -759,7 +759,7 @@ def read_txt(fname): if not cache_dir or not isinstance(cache_dir, str): cache_dir = None dataset = ( - wds.WebDataset(filepattern, cache_dir=cache_dir) + wds.WebDataset(filepattern, cache_dir=cache_dir, nodesplitter=lambda src: src) .decode(wds.autodecode.ImageHandler("pil", extensions=["webp", "png", "jpg", "jpeg"])) ) # Load based on classification or retrieval task