OpenCoder-LLM/opc-fineweb-math-corpus

OpenCoder数据集由多个子集组成，其中包括opc-fineweb-math-corpus子集。该子集是从Fineweb中提取的数学相关数据，专门用于OpenCoder的预训练。通过fastText进行三轮迭代，最终召回了一个包含55B代码和数学相关数据的数据集。

The opc-fineweb-math-corpus is a subset of the OpenCoder dataset, containing math-related data recalled from Fineweb. This data was specifically used in the pre-training of OpenCoder. The data was recalled using fastText in three iterative rounds, resulting in a dataset of 55B code and math-related data. Code-related data can be found in OpenCoder-LLM/fineweb-code-corpus. The dataset includes features such as url, tag, text, file_path, dump, file_size_in_byte, and line_count. The train split has 5241900 examples and a size of 18159796472 bytes. The dataset is licensed under odc-by.