Reply is generated twice
#1
by SandInTheDunes - opened
For some reason the model reply is generated once, then underneath the same reply would be generated once more in the /* */ section. I see that this model uses a manual prompt template (Alpaca) instead of Jinja (in LM Studio).
Interesting, some other chats instances I tried did not have this issue.
It should not use alpaca, but the hf to gghuf I used didn't apparenlly added the jinja correctly, at least nor for LM studio so this falls back to alpaca. I have to revisit this again, don't know where are all the bits and bobs from this model on my disk.