MaziyarPanahi/Llama-Nemotron-Post-Training-Dataset-v1-ShareGPT

Llama-Nemotron-Post-Training-Dataset-v1是一个转换为ShareGPT格式的数据集，保留了原始的数据划分和列。数据集来源于NVIDIA的Llama-Nemotron-Post-Training-Dataset-v1，包含角色为用户和助手的对话。它保留了原始的划分（代码、数学、科学、聊天、安全）和所有原始列，并在ShareGPT格式中添加了messages字段。数据集的各个部分行数如下：代码部分约1.56百万行，数学部分约13.1百万行，科学部分约48.4万行，聊天部分约3.98万行，安全部分约3.14万行。

Llama-Nemotron-Post-Training-Dataset-v1 is a dataset converted into ShareGPT format, preserving the original splits and columns. It is sourced from NVIDIAs Llama-Nemotron-Post-Training-Dataset-v1 and includes conversations with roles of user and assistant. It maintains the original splits (code, math, science, chat, safety) and all original columns, with an added messages field in the ShareGPT format. The dataset has the following row counts for each section: code section with approximately 1.56 million rows, math section with approximately 13.1 million rows, science section with approximately 484,000 rows, chat section with approximately 39.8 thousand rows, and safety section with approximately 31.4 thousand rows.