Language Models are Super Mario: Absorbing Abilities from Homologous Models as a Free Lunch
Paper • 2311.03099 • Published • 32
This is a merge of pre-trained language models created using mergekit.
This model was merged using the DARE TIES merge method using mergekit-community/mergekit-slerp-mirtnuv as a base.
The following models were included in the merge:
The following YAML configuration was used to produce this model:
models:
- model: mergekit-community/mergekit-slerp-mirtnuv
# No parameters necessary for base model
- model: mergekit-community/mergekit-della_linear-mcwqrsb
parameters:
density: 0.5
weight: 0.5
- model: mergekit-community/mergekit-sce-uinsryt
parameters:
density: 0.5
weight: 0.5
- model: mergekit-community/mergekit-slerp-rayqjvs
parameters:
density: 0.25
weight: 1
- model: mergekit-community/mergekit-slerp-lupllmg
parameters:
density: 0.25
weight: 1
merge_method: dare_ties
base_model: mergekit-community/mergekit-slerp-mirtnuv
parameters:
int8_mask: true
dtype: bfloat16