eightHundreds
eightHundreds
我其实是个白嫖党, 不喜永爱发电
我们来搞个awesome-gpt-rss
the solution ``` diff const Config = require('webpack-chain'); const myPlugin = require('my-plugin/webpack'); const config = new Config(); config .plugin('xxx') - .use(myPlugin); + .use(myPlugin()); ```
open ai也出问题, 出问题的地方各不相同, open ai的响应码是200,但是内容是报错的 
因为供应商的情况各种各样, 评测数据的也是各种各样. 我还是建议, 直接丢弃有问题的评测数据, 在规模量大的情况下, 少几条对结果没太大影响 如果靠开发者在异常回调那边手动补模型返回结果那代码是写不完的
> Hi, [@eightHundreds](https://github.com/eightHundreds). Thank you for the report. Can I know what your motivation was for passing return_outputs=True and return_all_scores=True as eval_kwargs? The evaluation result is treated as a single...
我这边是用evalscope app 看报告