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/Qwen2.5-72B-Instruct
4 layer_range:
5 - 0
6 - 80
7 - model: MaziyarPanahi/calme-2.1-qwen2.5-72b
8 layer_range:
9 - 0
10 - 80
11merge_method: slerp
12base_model: Qwen/Qwen2.5-72B-Instruct
13parameters:
14 t:
15 - filter: self_attn
16 value: 0.5
17 - filter: mlp
18 value: 0.5
19 - value: 0.5
20dtype: bfloat16
21tensor_parallel_devices: auto
22allow_different_heads: true
