返回探索

Jev 与本地决策模型:延迟优先的思路

作者认为让决策模型在本地运行是更优方案,并指出这类模型可做得足够小以适配多数硬件,同时应优先优化延迟以提升每秒决策数。

Jev opened the box, but I believe that having your decisioning model run locally is the winning recipe. Decisioning models can certainly be made small enough for most hardware and you really want to maximize for latency with these models, as a higher decisions-per-second may be

· 1 次赞在 X 打开