Quantization made by Richard Erkhov.
This repo contains a merge of pre-trained language models created using
mergekit.
In my testing, this model feels smart and attentive to context, balanced between creative and instruction-following.
Tested at temp=1 and minP=0.01.
This model was merged using the SLERP merge method.
1models:
2 - model: grimjim/Magot-v2-Gemma2-8k-9B
3 - model: grimjim/Magnolia-v1-Gemma2-8k-9B
4merge_method: slerp
5base_model: grimjim/Magot-v2-Gemma2-8k-9B
6parameters:
7 t:
8 - value: 0.2
9dtype: bfloat16
10
