This is a merge of pre-trained language models created using
mergekit.
This model was merged using the SLERP merge method.
1
2
3slices:
4
5 - sources:
6
7 - model: LDCC/LDCC-SOLAR-10.7B
8
9 layer_range: [0, 48]
10
11 - model: hyeogi/SOLAR-10.7B-dpo-v1
12
13 layer_range: [0, 48]
14
15merge_method: slerp
16tokenizer_source: base
17base_model: LDCC/LDCC-SOLAR-10.7B
18
19embed_slerp: true
20
21parameters:
22
23 t:
24
25 - filter: self_attn
26
27 value: [0, 0.5, 0.3, 0.7, 1]
28
29 - filter: mlp
30
31 value: [1, 0.5, 0.7, 0.3, 0]
32
33 - value: 0.5
34
35dtype: bfloat16
36
37
Finetuned using LoRA with
kyujinpy/OpenOrca-KO
