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 SvdH/RPLament-22B as a base.
The following models were included in the merge:
The following YAML configuration was used to produce this model:
merge_method: dare_ties
base_model: SvdH/RPLament-22B
parameters:
int8_mask: true
dtype: bfloat16
models:
- model: SvdH/RPLament-22B
use_cache: true
parameters:
weight: 0.60
density: 0.78
- model: TheDrummer/Cydonia-22B-v1.3
use_cache: true
parameters:
weight: 0.20
density: 0.66
- model: concedo/Beepo-22B
use_cache: true
parameters:
weight: 0.20
density: 0.54