Event ·Multi-source

TorchTitan 更新：支持 DeepSeek-V4 等 5D 并行，新增 TitanRL 框架

TorchTitan 在 PyTorch 原生框架中加入完整 5D 并行支持，覆盖 DeepSeek-V4、Qwen3.5、Kimi K3 和 Muse Spark Open 等模型。新发布的 TitanRL 框架统一了后训练与智能体工作负载的流程。FlexShard 可将分布式 Muon 优化器耗时最多降低 79%，torch_checkpointing 在 512 张 GPU 上将 checkpoint 阻塞时间压缩到 34 毫秒。

Current read

PyTorch 官方的 TorchTitan 大更新，DeepSeek-V4 和 Qwen3.5 都能跑 5D 并行了，Muon 优化器还能提速 79%，训大模型的得看看。

5 sources68° AI heatLast updated 2026/10/21 20:00:12

Evidence

5 sources

Primary sourcePyTorch

2026/10/21 20:00:12View source

Related source 1arXiv: DeepSeek

2026/10/21 11:50:17View source

Related source 2DAIR.AI

2026/10/21 02:15:02View source

Related source 3AWS Machine Learning Blog

2026/10/21 20:00:22View source

Related source 4IT之家

2026/10/22 09:19:52View source

Conflict 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.

Open the story