FuseChat: Knowledge Fusion of Chat Models
Paper
•
2408.07990
•
Published
•
14
A merge of a merge of a merge. I gathered some awesome models which were merges many of the same models just in different ways and merged them together as an experiment. Came out solid, I think.
This is a merge of pre-trained language models created using mergekit.
This model was merged using the SCE merge method using nbeerbower/Llama-3.1-Nemotron-lorablated-70B 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: Steelskull/L3.3-MS-Nevoria-70b
# Target models
- model: Nohobby/L3.3-Prikol-70B-v0.3
- model: Tarek07/Progenitor-V1.2-LLaMa-70B
- model: Tarek07/Progenitor-V1.1-LLaMa-70B
- model: sophosympatheia/Nova-Tempus-70B-v0.1
merge_method: sce
base_model: nbeerbower/Llama-3.1-Nemotron-lorablated-70B
parameters:
select_topk: 1.0
dtype: bfloat16