将真实模型转换为模拟器可读格式，让系统研究者轻松测试实际模型。
ModTrans: Translating Real-world Models for Distributed Training Simulator
- 开发翻译工具ModTrans，实现真实模型到ASTRA-sim的格式转换
- 实验表明转换开销可忽略不计，不影响模拟精度
- 适合想用真实模型做系统研究的机器学习研究人员
近年来，大规模分布式训练在产业与学术界均成为研究热点。然而，缺乏物理设备与资源使得实验难以开展。分布式训练模拟器（如ASTRA-sim）是可行方案，但现有工具无法导入真实世界开发的模型，限制了机器学习研究者使用。针对此问题，本文提出ModTrans，一个支持任意真实模型向ASTRA-sim输入格式转换的翻译器，打通了机器学习专家与机器学习系统研究者之间的壁垒。实验结果表明，ModTrans的开销可忽略不计。
原文摘要 · Abstract (English)
Large-scale distributed training has been a research hot spot in machine learning systems for industry and academia in recent years. However, conducting experiments without physical machines and corresponding resources is difficult. One solution is to leverage distributed training simulators, but current ones like ASTRA-sim do not support importing real-world developed models, which poses challenges for ML researchers seeking to use them. Based on this challenge, we developed ModTrans, a translator supporting format translation from any real-world model to the ASTRA-sim simulator's input, removing the barrier between machine learning experts and machine learning system researchers. The experiment results show that ModTrans's cost is negligible.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。