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 mllm-dev/merge_diff_data_DROID as a base.
The following models were included in the merge:
The following YAML configuration was used to produce this model:
base_model: mllm-dev/merge_diff_data_DROID
dtype: float16
merge_method: dare_ties
slices:
- sources:
- layer_range: [0, 12]
model: mllm-dev/merge_diff_data_DROID
- layer_range: [0, 12]
model: mllm-dev/merge_diff_data_YELP
parameters:
density: 0.5
weight: 0.5