天空вℓσи∂ [Deprecated edition] @skyblond@social.cleverlibre.org

关于我为什么想要极力避免C和C++

And I'm thinking how I can update its knowledge without doing a full training. Unfortunately, I don't have access to several A100 cards.

One way I came up is let it tell me what it need using prompt. Something like this:

[User]: What time is it?
[LLaMa]: [Request] current time
[System]: 2023-07-27T08:00:29
[LLaMa]: [Answer] Now is 8 AM.

But it seems like LLaMa did not have enough logic power to make request of what it needs to finish the task. I don't know if it's my prompt not good enough, or it's just simply not possible to do it.

Also, one observation of LLaMa 2, where I have to exit almost all programs to free up as much memory as I can.

It supports multiple languages including Chinese, but facing some questions that it doesn't know how to handle, the language becomes broken and creepy (while I know it's generated, but seeing those texts randomly on the internet kinda feels creepy). However, English is generally fine. No matter what I ask, it can at least generate proper human language.

Also, WTF is Google's WEI?

"a way for browser clients to establish trust with a server through a third party (eg, Google Play) that presents a token attesting to the integrity of the client environment."

They just want to control your PC like how they control Android.

Well, only the Nintendo users will figure their own way out...

Maybe those handheld PCs are going to change this?
----
I cant believe I beat Nintendo to it...
by Zac Builds
https://www.youtube.com/watch?v=KboTw3NBuuk

Tried llamacpp for 70b model, for now it can only run on cpu, thus very slow. But can I tell from the texr it generated, it's even better than the 13b model.

13b model feels like a robot, it answers what you ask, but 70b feels more like talk, it will talk about the questions you give, feels like chatgpt.

And it can run on my local hardware (despite very slow)