最新

The standard GPU utilization metric reported by nvidia-smi, nvtop, Weights &amp; Biases, Amazon CloudWatch, Google Cloud Monitoring, and Azure Monitor is highly misleading. It reports the fraction of time that any kernel is running on the GPU, which means a GPU can report 100% utilization even if only a small portion of its compute capacity is actually being used. In practice, we&#x27;ve seen workloads with ~1–10% real compute throughput while dashboards show 100%.<p>This becomes a problem when teams rely on that metric for capacity planning or optimization decisions, it can make underutilized systems look saturated.<p>We&#x27;re releasing an open-source (Apache 2.0) tool, Utilyze, to measure GPU utilization differently. It samples hardware performance counters and reports compute and memory throughput relative to the hardware&#x27;s theoretical limits. It also estimates an attainable utilization ceiling for a given workload.<p>GitHub link: <a href="https:&#x2F;&#x2F;github.com&#x2F;systalyze&#x2F;utilyze" rel="nofollow">https:&#x2F;&#x2F;github.com&#x2F;systalyze&#x2F;utilyze</a><p>We&#x27;d love to hear your thoughts!

赠阅文章：<a href="https:&#x2F;&#x2F

我一直在尝试构建一个完全本地的 RAG 流程：使用 Weaviate 进行向量存储和混合搜索，配合 Node.js 脚本，以及 Ollama 上的 Qwen 3.5 模型。<p>我发现，大部分挑战都集中在检索和文本分块上，而不是大语言模型本身。一个好的文本分块策略加上混合搜索的适当平衡，比使用更大、更昂贵的模型更有效。

多年来，我一直对那些要求你将工作方式扭曲以适应其系统的项目管理工具感到沮丧。Chimerical 反其道而行之；描述你正在做的事情，工作空间就会围绕你的答案自动组装。基于 16 年的领域知识，我用 6 周时间构建了它，今天发布。