This is a merge of pre-trained language models created using
mergekit.
This model was merged using the SLERP merge method.
1models:
2 - model: nbeerbower/flammen15-gutenberg-DPO-v1-7B
3 layer_range: [0, 32]
4 - model: nbeerbower/slerp-bophades-truthy-math-mistral-7B
5 layer_range: [0, 32]
6merge_method: slerp
7base_model: nbeerbower/flammen15-gutenberg-DPO-v1-7B
8parameters:
9 t:
10 - filter: self_attn
11 value: [0, 0.5, 0.3, 0.7, 1]
12 - filter: mlp
13 value: [1, 0.5, 0.7, 0.3, 0]
14 - value: 0.5
15dtype: bfloat16
16
