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 mistralai/Mistral-7B-v0.1 as a base.
The following models were included in the merge:
The following YAML configuration was used to produce this model:
models:
- model: augmxnt/shisa-gamma-7b-v1
parameters:
weight: 0.5
density: 0.4
- model: WizardLMTeam/WizardMath-7B-V1.1
parameters:
weight: 0.5
density: 0.4
- model: GAIR/Abel-7B-002
parameters:
weight: 0.5
density: 0.4
merge_method: dare_ties
base_model: mistralai/Mistral-7B-v0.1
parameters:
normalize: true
dtype: bfloat16