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 bunnycore/Qwen-2.5-7b-S1k as a base.
The following models were included in the merge:
The following YAML configuration was used to produce this model:
models:
# Pivot model
- model: bunnycore/Qwen-2.5-7b-S1k
# Target models
- model: arcee-ai/Arcee-Maestro-7B-Preview
- model: NovaSky-AI/Sky-T1-mini
- model: gz987/qwen2.5-7b-cabs-v0.3
merge_method: sce
base_model: bunnycore/Qwen-2.5-7b-S1k
parameters:
select_topk: 1.3
dtype: bfloat16