This is a merge of pre-trained language models created using
mergekit.
This model was merged using the
TIES merge method using
CultriX/NeuralTrix-7B-dpo as a base.
1models:
2 - model: CultriX/NeuralTrix-7B-dpo # no parameters necessary for base model
3 - model: mlabonne/AlphaMonarch-7B
4 parameters:
5 density: 0.5 # fraction of weights in differences from the base model to retain
6 weight: # weight gradient
7 - filter: mlp
8 value: 0.5
9 - value: 0
10 - model: bardsai/jaskier-7b-dpo-v5.6
11 parameters:
12 density: 0.5
13 weight: 0.5
14merge_method: ties
15base_model: CultriX/NeuralTrix-7B-dpo
16parameters:
17 normalize: true
18 int8_mask: true
19dtype: float16
Detailed results can be found
here
