精选70Chubby♨️行业动态
OpenAI 用 GPT-Astra 为 Jalapeño 芯片优化内核
interesting detail in OpenAIs blogpost to "Jalapeño":
interesting detail in OpenAIs blogpost to "Jalapeño":
OpenAI关于“Jalapeño”的博客文章中一个有趣的细节：
GPT-Astra used Codex to write and optimize low-level kernels that brought three previously unplanned open-weight models to high performance on OpenAI’s Jalapeño chip within two months.
GPT-Astra使用Codex编写并优化了低级内核，使得三个此前未计划的开放权重模型在两个月内就在OpenAI的Jalapeño芯片上达到了高性能。
For selected attention and mixture-of-experts blocks, its implementations ran 1.5–1.8× faster than existing human-expert-written code.
对于选定的注意力机制和混合专家模块，其实现比现有的人类专家编写的代码运行速度快了1.5至1.8倍。
looks like astra is a really really good model.
看来astra确实是一个非常非常好的模型。
更进一步：量化金融体系
看懂新闻只是起点——沿量化金融路径，把它变成能交付的工程能力