标签: 人工智能

RuntimeError: Application timeout caused pair closure

RuntimeError: Application timeout caused pair closure

该报错通常出现在 vLLM 多节点(2 节点及以上)大规模张量并行(如 TP=16)启动时的分布式初始化阶段,各 rank 在 Gloo barrier 中互相等待直至超时。优先排查是否由跨节点场景误入对称内存(symmetric memory) rendezvous 逻辑引起,可先尝试使用 --d

[Bug/Perf]: hybrid-SWA prefix caching collapses to zero for ALL requests in multi-session round-robin at ~25% pool occupancy (Gemma-4-31B; e

[Bug/Perf]: hybrid-SWA prefix caching collapses to zero for ALL requests in multi-session round-robin at ~25% pool occupancy (Gemma-4-31B; e

该问题发生在 vLLM 混合注意力(hybrid-SWA)模型的多会话轮询负载中,当 KV 缓存池占用率约达 25% 时,跨请求前缀缓存复用会突然降为零,表现为“全部命中”或“全部未命中”的断崖式切换。优先排查 SWA 组的尾部块是否被提前回收,可借助 CPU offloading(48GiB+)验

Using n8n built in AI to generate nodes leaves a description (with greater than 145 characters) on the set of grouped nodes, which then fail

Using n8n built in AI to generate nodes leaves a description (with greater than 145 characters) on the set of grouped nodes, which then fail

该报错发生在 n8n 2.36.x 版本中,使用内置 AI 生成节点分组时自动生成的描述超过 145 字符限制,导致整个工作流无法保存。优先排查方式是检查分组节点的描述字段长度,并手动缩短到 145 字符以内;升级到 n8n 2.37.0 或更高版本可以避免新生成内容触发此问题。