Qiyu (Allen) Zhong

Results 8 comments of Qiyu (Allen) Zhong

> ![image](https://user-images.githubusercontent.com/45348354/142579840-b1f17c62-c287-4e8f-a491-fb95c7f7f36e.png) 首次安装 先导入服务器就可以打开了 也可以在不配置服务器的情况下完成操作

您好!请问您是北京大学校友吗~ On Tue, Jun 21, 2022 at 17:42 dependabot[bot] ***@***.***> wrote: > Bumps numpy from 1.18.2 to 1.22.0. > Release notes > > *Sourced from numpy's releases .* > >...

You might want to check the CUDA maximum memory allocated.

I think the devs can add their own branches to the llama.cpp repo or huggingface.co? The 2.5 version of InternVL also got released..can take it a try for transfer as...

> 已经按照文档配置好了讯飞对接信息,并且从讯飞控制台已经能看到API次数消耗。 > > 问题:不定时失败,返回【收到未预期数据:】,只要失败前面的进度全部中断 > > 希望:能不能把转换失败的语音忽略,对应文字留空,不要影响前面已经转换成功的数据,这对API授权额度造成很大的浪费。 对每条要遍历抽取数据做的语音标记一下, 其他再做转换

https://github.com/KaihuaTang/Scene-Graph-Benchmark.pytorch This might be a good one to check, as nowadays we apply instruction following and tuning to work on the iterations.

https://github.com/LiuHC0428/LAW-GPT

有可能3或者3.5B能实现吗? 2的话和1.8,甚至1.9差别不大(参见Index-1.9B之类的? 同理5和6也是和7接近,虽然可能少一点翻译的纠错空间 如果依据黄金分割和scaling law,在1.8和7之间取可能是这两个尺寸了。