Thomas Wolf· @Thom_Wolf · X·· 6 小时前精选AI 评分68
AI 导读
DeepSeek 宣布将支撑 DeepSeek-V4.1-Flash 大规模推理的自研框架 DeepInfer 完全开源,以 Apache 2.0 协议发布,代码在 http://github.com/deepseek-ai/DeepInfer,称可帮助社区降低算力成本和能耗。Thomas Wolf 转发并列举其此前开源的 FlashMLA(H800 上 580 TFLOPS attention)、DeepEP(MoE 通信)、DeepGEMM(1350+ FP8 TFLOPS matmul)和 3FS 文件系统,称其开源节奏令人敬佩。
推荐理由
作者梳理了 DeepSeek 从 FlashMLA 到 DeepInfer 的开源脉络,读者可以借此看清其生产级推理栈的完整版图。
正文 · 原文
DeepSeek's open-source streak is getting absurd:
- FlashMLA: 580 TFLOPS attention on H800
- DeepEP: MoE comms
- DeepGEMM: 1350+ FP8 TFLOPS matmul
- 3FS: they built an entire file system
- and today: the full inference engine behind V4.1-Flash, Apache 2.0
Mad respect
🚀 Introducing DeepInfer, our in-house inference framework, now fully open source! Powering DeepSeek-V4.1-Flash at scale with high throughput and low latency. We’re sharing our production stack to help the community lower compute costs, reduce energy use, and make AI more affordable. 🔗 http://github.com/deepseek-ai/DeepInfer 1/4在 X 查看被引用的帖子
来源:Thomas Wolf · x.com