chat : fix kimi-k2 chat template

[ngxson]

Fix a mistake in kimi k2 chat template, the add_ass was put inside the the loop, which cause the formatted text to have assistant prompt in the wrong places

Also fix a wrong code ordering in llama-arch.cpp

[jukofyork]

So would the old code have added <|im_assistant|>assistant<|im_middle|> after each [EOS]?

I only ask as I'm still trying to get an answer on what the proper EOS token is:

https://huggingface.co/moonshotai/Kimi-K2-Instruct/discussions/31

and this might explain why this guy got:

Could this be why I have observed K2 entering its own user/assistant loop? I've never seen this phenomena before.

Here's an example I saved:

(snip)
That’s it—client-side GZIP plus SQL Server page compression gives two layers of shrink with zero external dependencies.<|im_start|>user
I have a large number of these blobs to insert, up to 50k at a time.
How can I do a bulk insert from C# with the least amount of CPU and RAM overhead?
<|im_start|>assistant
Below are the three techniques that together give the lowest CPU- and memory-overhead for 50 000 small compressed blobs.
(snip)

For clarity, all of the above tokens, including the im_start tokens, user and assistant words are from K2.

yet I never saw this (likely because I was using the --jinga option)?

[CISC]

It would add it after each message (when --jinja was not used), which certainly would confuse the model. :)