Notices by Emi Yusa (gnuxeava@mk.absturztau.be), page 6

:awa:

Boost if you are a hoarder too!

how do i sue Microsoft for creating teams?

watch this, free software purists

@mischievoustomato@rebased.taihou.website it is. Hosting prosody required me to open a bunch of ports scattered everywhere in the documentation, and some more by asking around on fedi. Next time i did that same but now group chat does not work. I've kept it as is because nobody talks to me

@mischievoustomato@rebased.taihou.website i remeber reading years ago how a person locked himself out of his mac by using emoji in his password

@alex@gleasonator.com do it

@puniko now i am hearing sounds i haven't heard in years

What does Firefox DevEdition mean by this?

@puniko you mean trying to pronounce absturztaube

oh devuan is memeing

@puniko

@puniko fuck around and find out

@kaia@brotka.st You can also use https://gpt.h2o.ai if you do not want to go through the hassle of setting it up

@kaia@brotka.st TheBloke does an awesome job releasing quantized version of the models. But I am not sure if those will work with llama.cpp. I use the scripts provided by llama.cpp to produce a quantized model.

https://huggingface.co/TheBloke

@kaia@brotka.st out of these, snoozy and vicuna perform the best, followed by orca mini 3b and orca mini v2. Orca was trained using a different approach. Llama is the whackier one here.

@kaia@brotka.st I run it all on llama.cpp (on CPU with 8 GB RAM)
https://huggingface.co/nomic-ai/gpt4all-13b-snoozy
https://huggingface.co/decapoda-research/llama-7b-hf
https://huggingface.co/psmathur/orca_mini_3b
https://huggingface.co/psmathur/orca_mini_v2_13b
https://huggingface.co/lmsys/vicuna-7b-v1.3
https://github.com/ggerganov/llama.cpp

@kaia@brotka.st gpt4all-snoozy

@puniko sleep

@NeonPurpleStar@outerheaven.club :nko_guns: