Nemotron-RL-Agentic-SWE-Pivot-v1
收藏资源简介:
SWE-RL 数据集提供了用于在 NeMo Gym 的 OpenHands 环境中训练和验证现实世界软件工程代理的 GitHub 问题。该数据集是对 SWE-Gym、R2E-Gym 和 SWE-Bench-Verified 数据集的重构版本,以支持 NeMo Gym 输入格式。数据集包含 6436 个训练样本,具有两个顶级特征(responses_create_params 和 agent_ref),训练数据大小约为 4.25GB。该数据集作为 NVIDIA NeMo Gym 的一部分发布,旨在用于大型语言模型(LLMs)的后训练。数据集采用 Creative Commons Attribution 4.0 International (CC-BY 4.0) 许可,适用于商业用途。
The SWE-RL dataset provides GitHub issues for training and validating real-world software engineering agents in the OpenHands environment of NeMo Gym. This dataset is a reconstructed version of the SWE-Gym, R2E-Gym, and SWE-Bench-Verified datasets to support the NeMo Gym input format. The dataset contains 6,436 training samples with two top-level features: responses_create_params and agent_ref, with a total training data size of approximately 4.25 GB. Released as part of NVIDIA NeMo Gym, this dataset is intended for post-training of Large Language Models (LLMs). The dataset is licensed under Creative Commons Attribution 4.0 International (CC-BY 4.0) and permits commercial use.



