牛哥精选 · 所有

📋 全部 ☁️ 云服务 🤖 AI 平台 🔗 API 中转 🔐 安全/认证 💳 支付 📧 通讯 📊 数据分析 🖼 媒体处理 🌐 域名/DNS

📝 深度技术 arXiv 机器学习 2026-06-30

One-Step Gradient Delay is Not a Barrier for Large-Scale Asynchronous Pipeline Parallel LLM Pretraining

打破异步训练瓶颈，一步延迟不再是障碍，大模型预训练提速新思路。

arXiv:2606.30634v1 Announce Type: new Abstract: Modern large-scale LLM pretraining benefits from utilizing Pipeline Parallelism; however, synchronous …

异步训练流水线并行 llm预训练梯度延迟大模型

📝 深度技术 Hacker News LLM 2026-06-20

Pipeline-parallel LLM inference across GPUs on separate machines

744B超大规模模型通过分布式流水线并行在多个消费级GPU上高效推理，速度达30 tok/s，展示前沿部署方案。

Article URL: https://github.com/leyten/shard Comments URL: https://news.ycombinator.com/item?id=48602121 Points: 1 # Comments: 0

744b参数分布式推理流水线并行消费级gpu glm-5.2

📝 深度技术 arXiv 机器学习 2026-05-26

A Tabular Schedule Abstraction for Communication-Aware Evaluation of Pipeline-Parallel LLM Training

针对大模型流水线并行训练，提出一种表格化调度抽象，有效评估通信感知性能，提升训练效率。

arXiv:2605.24006v1 Announce Type: cross Abstract: Pipeline parallelism is a key technique for distributed training of large language models because it…

llm训练流水线并行通信优化调度抽象论文

📅 日期

2026-05-20 2026-05-19

🐂 牛哥精选

One-Step Gradient Delay is Not a Barrier for Large-Scale Asynchronous Pipeline Parallel LLM Pretraining

Pipeline-parallel LLM inference across GPUs on separate machines

A Tabular Schedule Abstraction for Communication-Aware Evaluation of Pipeline-Parallel LLM Training

📅 日期