vector-index-bench/vibe

该数据集包含用于VIBE(Vector Index Benchmark for Embeddings)的多个子数据集，涵盖了文本、图像、文本到图像和代码等多种类型的数据。每个数据集以HDF5格式存储，包含了用于构建向量索引的嵌入向量、测试查询嵌入向量、每个测试查询的100个最近邻的ID和距离等信息。This dataset contains multiple sub-datasets for VIBE (Vector Index Benchmark for Embeddings), covering various types of data such as text, images, text-to-image, and code. Each dataset is stored in HDF5 format, including embedding vectors for building vector indexes, test query embedding vectors, IDs and distances for the 100 nearest neighbors of each test query, and more.