This is a merge of pre-trained language models created using
mergekit.
This model was merged using the
DARE TIES merge method using
mistralai/Mistral-7B-v0.1 as a base.
1models:
2 - model: mistralai/Mistral-7B-Instruct-v0.1
3 parameters:
4 density: 0.9
5 weight: 1.0
6 - model: mistralai/Mistral-7B-v0.1
7 parameters:
8 density: 0.33
9 weight:
10 - filter: mlp
11 value: 0.5
12 - value: 0
13merge_method: dare_ties
14base_model: mistralai/Mistral-7B-v0.1
15parameters:
16 normalize: true
17 int8_mask: true
18dtype: float16
