eatcosmos

Results 33 comments of eatcosmos

> Unfortunately, the free API is not operational at the moment. The current cost for using the API is $155.72 per day, and it seems unfeasible to sustain the API...

is there debug environment to Recurrence this problem?

> @keith2018 I've tried tapping but it didn't help hmmm. Gonna give pefdlog a try and see if there's anything suspicious. @eatcosmos hmmm nope it also happens in the release...

https://github.com/iree-org/iree ![image](https://github.com/triton-lang/triton/assets/83876321/ac2251c3-ac08-4f21-8309-50c0ae367b9e)

Hi @kyr0 Translated by ChatGPT,maybe with some mistake: Sorry for the late reply; I am still getting used to GitHub's notification system. I believe that many developments in technology are...

![image](https://github.com/liucongg/ChatGLM-Finetuning/assets/83876321/2bf11122-110e-45e1-84c8-50413617619e) 你好,这个什么时候停止呢?epoch=0/micro_step=560/global_step=560 这边global_step=3800,训练了好几个小时了,为什么还不停止呢? 是直接运行的 finetuning_pt.py,能帮讲解下是不是代码里要修改什么参数?

> 我这边1441条数据,训练5epoch、A40-48G,需要112min 你好 ,1441 条数据 训练4个小时了,还在epoch=0怎么回事呢?没有训练结束的迹象 下载后,直接运行的 finetuning_pt.py ` python finetuning_pt.py --train_path data/spo_0.json --model_dir ../pretrainmodel/THUDM/ChatGLM-6B_ChatGLM-Finetuning/ ` ![image](https://github.com/liucongg/ChatGLM-Finetuning/assets/83876321/5213b226-4a7c-4bed-98e2-2b6a35ebcc79)

@Barenboim 大佬好,哈哈距离上次使用好久了,技术水平还是原来的,发现还是做独立开发更有机会用workflow,上班没啥机会用到,想把workflow用在独立开发的场景里是不是可以降维打击, 用workflow做服务器端功能的话,是不是一般用nginx做前端解析https,然后转发给workflow的http,因为在workflow文档里没搜到https,所以我想应该是workflow做后端然后nginx做前端这个架构,早知道把计算机知识系统学一下了。 现在各种agent相关协议变多了,高性能的实现还很少,感觉传统的计算机要求还是很多,主要是算法和任务流规划环节增加了AI,数据本身的传输还是要高性能

能不能推荐个nginx web服务迁移到workflow全栈的教程,搭建网站(现在AI生成网站很火,所以搭建网站的需求变多了)平时都是照着网上的教程,教程几乎都是配置nginx.conf,因为确实计算机背景知识这块然后就懵了,不知道怎么迁移配置成workflow(是否需要把nginx.conf参数一一对应到workflow),要是全部用workflow应该会省事很多,用nginx反向代理到workflow的各个端口可能会损失性能