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 TareksLab/Second-Dimension-V1-LLaMa-70B as a base.
The following models were included in the merge:
The following YAML configuration was used to produce this model:
models:
- model: TareksLab/Sixth-Dimension-V1-LLaMa-70B
- model: TareksLab/Fifth-Dimension-V1-LLaMa-70B
- model: TareksLab/Zero-Dimension-F-LLaMa-70B
- model: TareksLab/Third-Dimension-V1-LLaMa-70B
- model: TareksLab/First-Dimension-V1-LLaMa-70B
merge_method: sce
base_model: TareksLab/Second-Dimension-V1-LLaMa-70B
parameters:
select_topk: 0.2
dtype: float32
out_dtype: bfloat16
chat_template: llama3
tokenizer:
source: TareksLab/Zero-Dimension-F-LLaMa-70B
pad_to_multiple_of: 8