RCreddit.com·
暂不在当前实时榜单
Trained my first small language model
A developer trained a small language model, approximately 50MB, to address a user experience issue with an internal application. This new model, which runs locally on an older Threadripper CPU, achieves a 97% accuracy rate and responds in 0.06 seconds. This significantly improves upon the previous Gemini Flash-based tool, which had a 0.9-1.2 second response time. The developer plans to deploy the model over the weekend, anticipating that its 0.1-0.2 second latency will be sufficient without requiring client-side inference.
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25日 23:01 UTC
- 收录
- 2026年9月25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 →