This is a merge of pre-trained language models created using
mergekit.
Irises are made via inbreeding to retain genetic stability. Reminded me of the model soup we have going on here.
This model was merged using the
DARE TIES merge method using
intervitens/mini-magnum-12b-v1.1 as a base.
1models:
2 - model: intervitens/mini-magnum-12b-v1.1
3 - model: nbeerbower/mistral-nemo-gutenberg-12B-v4
4 parameters:
5 weight: 0.4
6 density: 0.6
7 - model: tssst/nemo-gutenberg-12b-v2
8 parameters:
9 weight: 0.3
10 density: 0.5
11merge_method: dare_ties
12base_model: intervitens/mini-magnum-12b-v1.1
13dtype: bfloat16
14tokenizer:
15 source: "union"
