Primary sourcePyTorch
View sourceEvent ·Multi-source
PyTorchCon Demo Theater 13 场演示:从 Trainium 强化学习到 MoE 推理
PyTorchCon 周三 Demo Theater 安排 13 场演示,覆盖训练、推理、硬件适配和智能体工作流。Baseten 展示 Wan 2.2 生成五秒视频用时不到三秒,较基线提速 53.6 倍,靠时间步蒸馏、定制内核和 4-bit 量化实现。Verda Cloud 演示在 NVIDIA NVL72 上跑 WideEP 专家并行 MoE 推理,Anyscale 讲 Ray Core 扩展到 10,000 节点集群的后训练方案。其他演示包括华为的 CRCR 跨仓库 CI 测试、Clockwork 的训练断链恢复,以及 Lightning AI 上自动复现 ML 研究的智能体。
Current read
PyTorchCon 一天 13 场硬核演示,Wan 2.2 视频生成提速 53.6 倍、万节点 Ray 后训练都在列,做基建的值得扫一遍日程。
11 sources46° AI heatLast updated 2026/10/21 20:01:22
Evidence
11 sourcesRelated source 1rohanpaul_ai
View sourceRelated source 2techcrunch
View sourceRelated source 3VTV Công nghệ
View sourceRelated source 4coderabbitai
View sourceRelated source 5IT之家
View sourceRelated source 6kimmonismus
View sourceRelated source 7Teortaxes
View sourceRelated source 8Tech in Asia
View sourceRelated source 9NVIDIA
View sourceRelated source 10CoreWeave
View sourceConflict check
Dedup data only shows that these sources discuss the same event, not that they agree. There is no structured supporting or refuting evidence yet, so no conflict is inferred automatically.