MagicSource

Yes, but I still struggle on a new issue: if set client to re-connect auto true, when the network is poor, it will reconnect endlessly and make the UI stuck,...

Awesome! looking forward to it!

@mMrBun 我发现虽然代码里面有streaming相关的逻辑，但实际请求的时候，时一次性打印的，而不是stream的方式，我的请求代码如下： ``` stream_mode = True if stream_mode: # create a chat completion for chunk in openai.ChatCompletion.create( model="gpt-3.5-turbo", messages=[{ "role": "user", "content": "列举出中国所有的省市以及对应的省会" }], stream=True, ): content = chunk["choices"][0].get("delta", {}).get("content")...

@hiyouga 我测试了一下，好像可以stream。 不过看起来是一行一行stream的，有办法一个字一个字的stream吗（或者一个token一个token）

奇怪哎，为啥我是一行一行输出的

Will this effect performance on chat model? Just saw many other frameworks all using padding right

@ltdrdata thanku! very helpful. I setted clip skip, but still can not really get a approximate result as the model announced, this is I got:  I have using 30...

@ltdrdata hi, how to set the `token weights` can u elebrate? Do u mean cfg? from the webiste it says using 7.0, does there any a calculate formate between A1111...

@ltdrdata You might know magic, bro! 

Same