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: Qwen/QwQ-32B-Preview
4 layer_range:
5 - 0
6 - 32
7 - model: Qwen/Qwen2.5-Coder-32B-Instruct
8 layer_range:
9 - 0
10 - 32
11merge_method: slerp
12base_model: Qwen/QwQ-32B-Preview
13parameters:
14 t:
15 - filter: self_attn
16 value:
17 - 0
18 - 0.5
19 - 0.3
20 - 0.7
21 - 1
22 - filter: mlp
23 value:
24 - 1
25 - 0.5
26 - 0.7
27 - 0.3
28 - 0
29 - value: 0.5
30dtype: bfloat16
