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 Qwen/Qwen3-30B-A3B-Base as a base.
The following models were included in the merge:
The following YAML configuration was used to produce this model:
base_model: Qwen/Qwen3-30B-A3B-Base
models:
- model: allura-forge/q3-30b-ft-ep2-merged
parameters:
select_topk: 0.75
- model: Gryphe/Pantheon-Proto-RP-1.8-30B-A3B
parameters:
select_topk: 0.4
- model: Qwen/Qwen3-30B-A3B
parameters:
select_topk: 0.25
merge_method: sce
dtype: bfloat16