FuseChat: Knowledge Fusion of Chat Models
Paper
•
2408.07990
•
Published
•
14
This is a merge of pre-trained language models created using mergekit.
This model was merged using the SCE merge method using SvalTek/Qwen2.5-ColdBrew-Antrax as a base.
The following models were included in the merge:
The following YAML configuration was used to produce this model:
name: SvalTek/Qwen2.5-ColdBrew-Antrax
models:
- model: SvalTek/Qwen2.5-ColdBrew-Antrax
- model: open-thoughts/OpenThinker-7B
- model: bunnycore/Qwen2.5-7B-RRP-1M
merge_method: sce
base_model: SvalTek/Qwen2.5-ColdBrew-Antrax
tokenizer_source: SvalTek/Qwen2.5-ColdBrew-Antrax
parameters:
select_topk: 1.5
int8_mask: true
dtype: bfloat16