This is a merge of pre-trained language models created using
mergekit.
This model was merged using the SLERP merge method.
1slices:
2- sources:
3 - model: jaspionjader/bh-2
4 layer_range:
5 - 0
6 - 32
7 - model: jaspionjader/Kosmos-EVAA-immersive-sof-v44-8B
8 layer_range:
9 - 0
10 - 32
11merge_method: slerp
12base_model: jaspionjader/bh-2
13parameters:
14 t:
15 - filter: self_attn
16 value:
17 - 0.09
18 - 0.05
19 - 0.07
20 - 0.08
21 - 0.06
22 - filter: mlp
23 value:
24 - 0.06
25 - 0.08
26 - 0.07
27 - 0.05
28 - 0.09
29 - value: 0.07
30dtype: bfloat16
31
