Language Models are Super Mario: Absorbing Abilities from Homologous Models as a Free Lunch
Paper
•
2311.03099
•
Published
•
30
This is a merge of pre-trained language models created using mergekit.
This model was merged using the DARE TIES merge method using redrix/wuriaee-12B-schizostock as a base.
The following models were included in the merge:
The following YAML configuration was used to produce this model:
base_model: redrix/wuriaee-12B-schizostock
merge_method: dare_ties
models:
- model: redrix/AngelSlayer-12B-Unslop-Mell-RPMax-DARKNESS-v3
parameters:
weight: 0.5
- model: redrix/GodSlayer-12B-ABYSS
parameters:
weight: 0.5
parameters:
density: 0.67
normalize: true
tokenizer_source: union
dtype: bfloat16