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 /teamspace/studios/this_studio/work_dir/intermediate_stages/balanced_creative as a base.
The following models were included in the merge:
The following YAML configuration was used to produce this model:
base_model: /teamspace/studios/this_studio/work_dir/intermediate_stages/balanced_creative
dtype: bfloat16
merge_method: dare_ties
models:
- model: /teamspace/studios/this_studio/work_dir/intermediate_stages/balanced_creative
parameters:
density: 0.75
weight: 0.52
- model: /teamspace/studios/this_studio/work_dir/intermediate_stages/intelligence_fusion
parameters:
density: 0.7
weight: 0.48
name: FINAL_UnhingedScribe_12B
parameters:
int8_mask: true
tokenizer_source: base