This is a merge of pre-trained language models created using
mergekit.
This model was merged using the SLERP merge method.
1models:
2 - model: Qwen/Qwen2.5-Coder-7B-Instruct
3 - model: Qwen/Qwen2.5-Coder-7B-Instruct
4merge_method: slerp
5tokenizer_merge_method: slerp
6tokenizer_parameters:
7base_model: Qwen/Qwen2.5-Coder-7B-Instruct
8dtype: bfloat16
9parameters:
10 t: [0, 0.3, 0.6, 0.8, 0.6, 0.3, 0] # Curva che favorisce leggermente il centro del merge
11 temp: 1.0 # Temperatura aumentata per un merge più audace
12density:
13 - threshold: 0.05
14 t: 0.8
15 - threshold: 0.3
16 t: 0.6
17 - threshold: 0.7
18 t: 0.4
19 - threshold: 0.95
20 t: 0.2
21noise: 0.1 # Aggiunto rumore per ulteriore variabilità
22
