Language Models are Super Mario: Absorbing Abilities from Homologous Models as a Free Lunch
Paper
•
2311.03099
•
Published
•
30
ChaosRose-24B is a merge of pre-trained language models created using mergekit.
Notes: Very chatty and descriptive.
This model was merged using the DARE TIES merge method using arcee-ai/Arcee-Blitz as a base.
The following models were included in the merge:
The following YAML configuration was used to produce this model:
base_model: arcee-ai/Arcee-Blitz
merge_method: dare_ties
dtype: bfloat16
models:
- model: arcee-ai/Arcee-Blitz
parameters:
weight: 0.38
density: 0.92
- model: LatitudeGames/Harbinger-24B
parameters:
weight: 0.31
density: 0.86
- model: Vortex5/ChaosFlowerRP-24B
parameters:
weight: 0.31
density: 0.86
tokenizer:
source: arcee-ai/Arcee-Blitz
chat_template: auto
parameters:
normalize: true