https://huggingface.co/jondurbin/airoboros-180b-2.2.1
#2
by
Wsdfsdf
- opened
IQ3_XXS of this model would be appreciated.
Please see if you can use an older version of llama.cpp to quantize the model with, as the perplexity might be worse for IQ3_XXS now then in the past.
https://github.com/ggerganov/llama.cpp/issues/5856
Funny concidence, airoboros is already crunching in the queue since Feb 27. I'll have a look at the issue, but going to an older version means risking corruption in other quants, so I'll probably chose the current version.
Wsdfsdf
changed discussion status to
closed