KingOmini_14b_v1 is a merge of the following models using
mergekit:
1merge_method: dare_ties
2base_model: Qwen/Qwen2.5-14B-Instruct # Still use ORIGINAL base for task vectors
3
4models:
5 # Stage 1 result (treat as single model)
6 - model: kainatq/KingOmini_14b_v1_M1
7 parameters:
8 weight: 1.0
9
10 # Medium priority: Literary creativity
11 - model: v000000/Qwen2.5-14B-Gutenberg-1e-Delta
12 parameters:
13 weight: 0.8
14
15 # Lower priority: Security/coding (avoid over-restriction)
16 - model: clouditera/secgpt
17 parameters:
18 weight: 0.5
19
20parameters:
21 density: 0.33
22 int8_mask: true
23 normalize: true```
