bytedance/byteps

How did you get the horovod & bytePS performance

Open

#68 创建于 2019年7月18日

在 GitHub 查看
 (12 评论) (2 反应) (0 负责人)Python (3,483 star) (476 fork)batch import
distributeddocumentationgood first issue

描述

I have the same hardware envs, same network, but I could not get the result as you, almost half as you. Any best practices and experience? thanks very much! for bytePS with 1 instance and 8 GPU, I have similar testing result.

贡献者指南

技术栈
python
领域
machine learningperformance
议题类型
performance
难度面向新贡献者的预计实现难度，1 表示很小改动，5 表示专家级工作。
4
预计时间有经验贡献者完成调查、实现、测试并准备 pull request 的粗略时间范围。
over 1 week
活动状态议题当前的可参与程度：新鲜、活跃、陈旧、阻塞或等待维护者输入。
stale
清晰度议题是否清楚说明期望改动、验收标准和下一步。
unclear
前置要求
distributed trainingBytePSHorovodGPU cluster
新手友好度1-100 的估计分数，表示该议题对首次贡献者的友好程度。
10
研究方向
The issue is a open ended query about reproducing BytePS performance benchmarks. To assist, first clarify the user's exact hardware, software versions, and network topology. Then, review BytePS configuration files and compare with published benchmarks. Check for common pitfalls like NCCL settings, network latency, or GPU affinity. Provide step by step tuning guidance and recommend running a simple allreduce test to isolate the bottleneck.