AbstractPhil/geometric-vocab

这是一个包含几何词汇嵌入的完整统一集合，具有优化的碎片大小以避免速率限制并提高加载性能。数据集以19种不同的维度提供，并且包含Unicode和WordNet数据。每个令牌都被嵌入为一个5顶点单形，并在n维空间中表示。数据集以Parquet文件格式存储，并可以通过Hugging Face的`load_dataset`函数进行加载。该数据集旨在帮助研究人员和开发者构建和训练几何词汇模型。

This dataset is a complete unified collection of all geometric vocabulary embeddings with optimized shard sizes to avoid rate limits and improve loading performance. It includes embeddings in various dimensions, such as Unicode and WordNet data. Each token is embedded as a 5-vertex simplex in n-dimensional space. The dataset is stored in Parquet files and can be loaded using the `load_dataset` function from Hugging Face. It is designed to assist researchers and developers in building and training geometric vocabulary models.