younggggger
younggggger
congras to me. The problem of "emoji.h" can't find is solved by me.When Telegram.sln generated,build codegen FIRST,it generate file,emoji.h;then build telegram. But YOUR building-***.md not mentioned. 
> Yeah, you don't need to do that. You have something done wrongly. i build telegram from sourcecode because i can't receive sms code from the version of official telegram...
> @younggggger the server won't send the code unless you use official Android or iPhone application. In case of Android, google services must be installed. in debug environment, server don't...
> @younggggger you mean the test server? do you use TDESKTOP_API_TEST? how to configure TDESKTOP_API_TEST?
thank you ,i take a try
> > how to configure TDESKTOP_API_TEST? > > `-D TDESKTOP_API_TEST=ON` to configure.bat i added -D TDESKTOP_API_TEST=ON to configure.bat,but still can't receive sms code
> > how to configure TDESKTOP_API_TEST? > > `-D TDESKTOP_API_TEST=ON` to configure.bat i added -D TDESKTOP_API_TEST=ON to configure.bat,but still can't receive sms code,can you help me?
> 可以自行修改`web_demo.py`以达到在colab中运行的效果。 > > 这是一个可以正常使用的colab链接: > > https://colab.research.google.com/drive/1S8w0pbOsUHU1JqJzoWQKNgvG6HBwbI6c#scrollTo=rASYNo8CxFlK > > 在我的使用过程中,回复速度在30-40s以上,内存占用在7.7G,显存占用维持在12.7G,超过20条消息后达到13.8G,属于一个能用的状态。`quantize()`和`cuda()`因为报错已经移除。 > > 至于如何进一步优化占用或榨取性能,请调整`device_map`中的值,`0`代表在GPU上运行,`cpu`代表在CPU上运行,`disk`将offload到本地存储中。 > > 参考文档: > > * https://huggingface.co/docs/accelerate/usage_guides/big_modeling > * https://github.com/THUDM/ChatGLM-6B#%E6%A8%A1%E5%9E%8B%E9%87%8F%E5%8C%96 > > PS: 感觉傻乎乎的。 我在本地运行,提示 raise...
> 爱莫能助。 colab上跑起来了,但回答问题好久不出来
> @younggggger 修好了,之前被我改坏了。现在换成了量化后的模型,不会炸内存,显存也充裕了,回复速度在10s左右。原版模型在注释里,感兴趣也可以拿出来玩玩。 可以了,谢谢啊,简单玩了下: 用户:写出10以内的质数 ChatGLM-6B:10以内的质数有2、3、5、7、11、13、17、19、23、29、31、37、41、43、47、53、59、61、67、71、73、79和83。 用户:写错了 ChatGLM-6B:非常抱歉,我的回答有误。10以内的质数是2、3、5、7、11、13、17、19、23、29、31、37、41、43、47、53、59、61、67、71、73、79和83。再次感谢您的指正。 用户:这是100以内的,不是10以内 ChatGLM-6B:非常抱歉,我的回答又有误。100以内的质数有2、3、5、7、11、13、17、19、23、29、31、37、41、43、47、53、59、61、67、71、73、79、83、89和101。再次感谢您的指正。