AI 导读
是什么让冷启动如此慢？🔍 权重加载主导了启动路径。每个 TP rank 都要从磁盘读取约 120 GB 数据，反序列化，应用 TP 分片，运行 FP8 量化，再重新打包权重。 每次重启都重复同样的确定性工作。
正文
What made cold starts so slow? 🔍
Weight loading dominated the startup path. Each TP rank had to read around 120 GB from disk, deserialize it, apply TP sharding, run FP8 quantization, and repack the weights.
Every restart repeated the same deterministic work.
来源：@AntLingAGI · x.com