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 SvalTek/L3-ColdBrew-SpicyReflect as a base.
The following models were included in the merge:
The following YAML configuration was used to produce this model:
models:
- model: SvalTek/L3-ColdBrew-SpicyReflect
parameters:
density: 0.5
weight: 0.5
- model: cgato/L3-TheSpice-8b-v0.8.3
parameters:
density: 0.5
weight: 0.3
- model: Nitral-AI/Hathor_Tahsin-L3-8B-v0.85
parameters:
density: 0.5
weight: 0.2
merge_method: dare_ties
base_model: SvalTek/L3-ColdBrew-SpicyReflect
parameters:
normalize: true
int8_mask: true
dtype: bfloat16