# 混元文本/多模态预训练框架研发工程师（深圳/北京/上海/杭州）

External opening: you cannot apply through AI Native Match; apply on the official site.

Apply through the official channel. The full external receipt stays in your workspace. When signed in, three minimal event types explicitly recorded through the CLI are reported by default: selection, self-reported application and channel observation. Browsing or saving a role does not report an event; external applications do not appear under My applications.

- Company: 腾讯 (Stage unknown)
- Locations: 深圳
- Role: Engineering
- Apply on the official site: https://careers.tencent.com/jobdesc.html?postId=1958073219999223808

## Role description from source

1.参与开发优化大模型训练框架，支持单任务万卡以上规模高效稳定训练；
2.参与NLP、多模态大模型结构设计，并联合业务进行模型训练效率和效果验证；
3.参与文生图、文生视频、文生3D等业务的训练性能加速；
4.参与低精度训练性能优化和业务推广、参与大窗口训练性能优化。
职位要求
1.熟练使用主流大模型训练框架DeepSpeed、Megatron，掌握3D并行、ZeRO机制、Flash-Attn等的原理、使用场景、优劣势以及可优化方向；
2.有ViT、SD、DiT模型训练性能优化经验者优先；
3.熟练掌握CUDA性能优化手段，有算子编写优化项目经验者优先；
4.对大模型前沿技术比较敏锐者优先；
5.有实际大模型的训练调参和效果评测项目经验的优先；
6.良好的沟通能力、解决问题能力。
工作年限（官方顶部字段）
一年以上工作经验

---
opening_id: 1ab924b3-f006-47f9-834d-c4268306147c

`anm openings save --id 1ab924b3-f006-47f9-834d-c4268306147c`