RCreddit.com·
暂不在当前实时榜单
Simulating fault tolerance with stage skipping in pipeline-parallel training [R]
Templar's recent work explores fault tolerance in Crucible, their distributed pre-training platform, aiming to maintain training with healthy workers when a pipeline stage fails. Simulations with a 178M model, eight replicas, and four stages per replica showed that a 1% per-replica failure probability per global step, where each outage removed a stage for six global steps, resulted in validation loss staying close to the no-failure baseline. Each configuration was compared against its own no-failure run.
This report uniquely details how Templar's simulations, unlike prior work, specifically test fault tolerance by skipping stages with fixed projections.
时间与来源
时间显示为 UTC
显示时区:UTC
本地时区尚不可用,暂时显示 UTC。
收录当时偏移:UTC+02026年9月22日 23:01 UTC
- 收录
- 2026年9月22日 23:01
- 来源类型
- 开发者社区
本站未收录正文。
前往源站阅读 →